Online advertising technology for artificial intelligence (AGI) and superintelligence (SI)

Abstract

Currently, online advertising systems are primarily used to attract human attention and monetize it through the sale of goods and services. However, this traditional use of online advertising significantly undervalues ​​human attention, a valuable resource. A superior way to monetize human attention is to focus it on solving valuable problems. Even better is to acquire specific human expertise through online advertising, utilize that expertise to train advanced AI systems, and then use that expertise to solve valuable problems. This technology demonstrates that Alphabet, Google, Meta, Amazon, Alibaba, ByteDance, TikTok, Microsoft, Apple, TenCent, Baidu, Twitter, X, Spotify, PubMatic, Pinterest, Snap, and other online advertisers can significantly increase their online advertising revenue. This technology discloses a system and method for monetizing human attention far more efficiently and effectively than traditional online advertising systems, by enabling existing online advertising technologies to power AI, Artificial General Intelligence (AGI), and superintelligence systems. Several preferred embodiments are discussed, including specific scenarios related to the above-mentioned companies. This technology concludes with a discussion of how the invention can be used to improve AI safety and maximize the potential for human survival and prosperity in the age of AI.

Description

Technical Field

[0001] In some aspects, the present technology relates to online advertising techniques related to artificial intelligence (AI), artificial general intelligence (AGI), and superintelligence (SI), and is utilized to overcome the data bottleneck currently faced by AI researchers in order to improve the intelligence of AI / AGI systems.

[0002] Existing online advertising models are experiencing diminishing returns due to the decreasing additional effect of improving ad targeting. In some other aspects, the present technology relates to methods related to more effectively monetizing human interests than existing online advertising models.

[0003] In yet other aspects, all activities described in this patent specification as being performed on an external network are also feasible within a single computerized intelligent system. In such external networks, multiple intelligent entities participate in collaborative problem-solving. When implemented within a single computerized intelligent system, all of these intelligent entities are either computerized entities existing within that system or AI agents.

Background Art

[0004] The most rapid and secure paths in the development of AGI and SI are described in previous invention disclosures. Methods and catalysts for improving the intelligence of general AI systems, as well as the development of AGI and personalized superintelligence (PSI), have also been previously disclosed. Accordingly, the following U.S. Provisional Patent Applications (PPAs) are hereby incorporated by reference herein.

[0005] This application incorporates, by reference, the entire content of Patent Document 1, titled "Advanced Autonomous Artificial Intelligence (AAAI) System and Methods," which was filed with the USPTO on 28 February 2023 and accepted.

[0006] This application incorporates, by reference, the entire content of Patent Document 2, titled "System and Methods for Ethical and Safe Artificial General Intelligence (AGI) Including Scenarios with Technology from Meta, Amazon, Google, DeepMind, YouTube, TikTok, Microsoft, OpenAI, Twitter, Tesla, Nvidia, Tencent, Apple, and Anthropic," which was filed with and accepted by the USPTO on March 17, 2023.

[0007] This application incorporates, by reference, the entire content of Patent Document 3, titled "System and Methods for Human-Centered AGI," filed with the USPTO on March 24, 2023.

[0008] This application incorporates, by reference, the entire content of Patent Document 4, titled "System and Methods for Safe, Scalable, Artificial General Intelligence," filed with the USPTO on 18 July 2023.

[0009] This application incorporates, by reference, the entire content of Patent Document 5, titled "Safe Personalized Super Intelligence (PSI)," filed with the USPTO on August 14, 2023.

[0010] This application incorporates, by reference, the entire content of Patent Document 6, titled "Catalysts for Growth of Super Intelligence," filed with the USPTO on November 22, 2023.

[0011] This application incorporates, by reference, the entire content of Patent Document 7, titled "System and Methods for Safe Alignment of Super Intelligence," filed with the USPTO on December 13, 2023.

[0012] In addition to the above-mentioned PPA, this application incorporates, by reference, all the contents of the PCT applications, including Patent Document 8 (filed February 26, 2024), Patent Document 9 (filed February 26, 2024), Patent Document 10 (filed February 26, 2024), Patent Document 11 (filed February 26, 2024), Patent Document 12 (filed February 26, 2024), Patent Document 13 (filed March 12, 2024), and Patent Document 14 (filed March 17, 2024), which also refer to the above-mentioned PPA. [Prior art documents] [Patent Documents]

[0013] [Patent Document 1] U.S. Provisional Application No. 63 / 487 / 494 [Patent Document 2] U.S. Provisional Application No. 63 / 491,040 [Patent Document 3] U.S. Provisional Application No. 63 / 577,830 [Patent Document 4] U.S. Provisional Application No. 63 / 628,410 [Patent Document 5] U.S. Provisional Application No. 63 / 519,549 [Patent Document 6] U.S. Provisional Application No. 63 / 601,930 [Patent Document 7] U.S. Provisional Application No. 63 / 609,800 [Patent Document 8] PCT / US24 / 17233 [Patent Document 9] PCT / US24 / 17251 [Patent Document 10] PCT / US24 / 17261 [Patent Document 11] PCT / US24 / 17269 [Patent Document 12] PCT / US24 / 17304 [Patent Document 13] PCT / US24 / 19486 [Patent Document 14] PCT / US24 / 20334 [Overview of the project] [Problems that the invention aims to solve]

[0014] This application further includes technologies that can be used in combination with, or independently, the systems and methods described in the above-mentioned PPA and PCT.

[0015] While the above-mentioned devices each meet specific purposes and requirements, the above patent does not describe online advertising technologies for AGI and SI that can improve the intelligence of AI / AGI systems by overcoming the data bottlenecks currently faced by AI system researchers, or that can more effectively monetize human interests while traditional online advertising models have the problem of diminishing returns in improving the targeting of incremental advertising. The above patent also does not disclose anything about XXXXX (repeatedly describe how this technology differs from the above patent).

[0016] Therefore, there is a need for new and improved online advertising technologies for AGI and SI that can improve the intelligence of AI / AGI systems by overcoming the data bottlenecks currently faced by AI system researchers and can more effectively monetize human interests while traditional online advertising models have the problem of diminishing returns in improving the targeting of incremental advertising. This technology substantially meets this need. In this regard, the online advertising technologies for AGI and SI according to this technology deviate substantially from the concepts and designs of the prior art. As a result, it provides a device developed mainly for the purpose of improving the intelligence of AI / AGI systems by overcoming the data bottlenecks currently faced by AI system researchers and more effectively monetizing human interests while traditional online advertising models have the problem of diminishing returns in improving the targeting of incremental advertising.

Means for Solving the Problems

[0017] Considering the drawbacks inherent in known AI systems and methods, as well as computerized advertising monetization methods, at least some embodiments of the present technology provide novel online advertising technologies for AGI and SI, overcoming some or all of the drawbacks and deficiencies of the prior art. Thus, a general objective of at least some embodiments of the present technology is to provide novel and innovative online advertising technologies for AGI and SI, as will be described hereinafter. This technology has all the advantages of the prior art and also has a number of novel features, resulting in the realization of online advertising technologies for AGI and SI that are not expected, not obvious, not suggested, or not even implied in the prior art.

[0018] According to one aspect, the present technology can include a system or method for two-way online advertising utilized for the learning and knowledge improvement of an artificial intelligence (AI) agent or system. The two-way online advertising is configured using issue information related to an issue given to the AI agent or system. This two-way online advertising is configured or can be configured to acquire the interest of a human user by issue information or content related to the issue. Input information obtained from the human user by the two-way online advertising is passed to the AI agent or system. The AI agent or system uses this input information to update the two-way online advertising or to utilize the input information in the issue-solving process related to the issue.

[0019] According to another aspect, the present technology may include a system that utilizes online advertising technology to enhance the intelligence of an artificial intelligence (AI) agent or system. The system may include a computer system. The computer system includes a processor, a computer-readable storage medium, and program instructions. The program instructions are program instructions stored in the computer-readable storage medium. To provide online advertising units to one or more user computer systems using a network, Receiving one or more inputs from user computer systems participating in an online advertising unit, the inputs being provided by human users utilizing each user computer system, and the inputs being associated with one or a combination of solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem. Transmitting input to an AI agent or system, An AI agent or system uses input in a universal problem-solving method to generate solutions to a problem or sub-problem. To train an AI agent or system based on the results of universal problem-solving methods derived from successful or unsuccessful attempts to solve a problem or sub-problem, thereby improving the intelligence of the AI ​​agent or system. In order for a computer system to realize this, it is possible to do so by a processor.

[0020] In yet another aspect, the technology is a method of online advertising technology used in conjunction with an artificial intelligence (AI) agent or system, which may include a method for enhancing the intelligence of the AI ​​agent or system. The process of receiving issue information regarding the issue provided by an AI agent or system, The process of generating online advertising units that include issue information, The process of providing online advertising units to one or more user computer systems using a network, A step of receiving one or more inputs from user computer systems participating in an online advertising unit, wherein the inputs are provided by human users utilizing each user computer system, and the inputs are associated with solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem, or a combination thereof. The process of transmitting input to an AI agent or system, A process in which an AI agent or system uses input in a universal problem-solving method to generate a solution to a problem or sub-problem, A process of training an AI agent or system based on the results of a universal problem-solving methodology derived from successful or unsuccessful attempts to solve a problem or sub-problem, thereby improving the intelligence of the AI ​​agent or system. It may include.

[0021] Some embodiments of this technology may further include the step of receiving advertising specifications from a client, which will be used when generating online advertising units, in combination with problem information.

[0022] In some embodiments, the advertising specifications may include one or a combination of advertising content, demographic information, location restrictions for online ad units, an advertising budget, and metrics for determining whether a problem or sub-problem has been successfully resolved.

[0023] Some embodiments of this technology may further include a step of charging the client based on a cost per 1,000 impressions or a click-through rate metric.

[0024] In some embodiments, the problem information may include one of the following: the current state of the problem, sub-problems of the problem, the goal of the problem, or sub-goals of the goal, or a combination thereof.

[0025] Some embodiments of this technology may further include the step of transmitting one or more additional inputs from an intelligent entity to an AI agent or system in combination with input from a participating user computer system, the intelligent entity being one of the following: an additional human user utilizing an additional computer system, an additional AI agent or system, an artificial general intelligence (AGI) agent or system, a superintelligence (SI) agent or system, or a combination thereof.

[0026] In some embodiments, the update of the online ad unit may include whether the input was accepted or rejected, whether the goal or subgoal of the task was achieved, whether the input was selected by the AI ​​agent or system to advance the task solving process, a description of the new task state after applying the selected input, the amount of credit or reward accumulated by the intelligent entity based on that input, and any additional request, or a combination thereof, to solicit new user input via the online ad unit or link and involve them in the task solving process.

[0027] Some embodiments of this technology may further include a step of awarding credit to participating intellectual entities based on their contribution once a problem or sub-problem has been solved.

[0028] Some embodiments of this technology may further include a step of soliciting additional input from human users of each participating user computer system if the problem or sub-problem is not solved.

[0029] Some embodiments of this technology may further include a step of compensating a human user if the human user exits an online advertising unit and ceases to participate in the problem-solving process.

[0030] In some embodiments, the learning of an AI agent or system may further include a learning method selected from the group consisting of differential privacy, federated learning, homomorphic cryptography, synthetic data generation, secure multi-party computation, data anonymization, transfer learning, active learning, self-supervised learning, domain adaptation, reinforcement learning, few-shot learning, explainable AI, privacy-preserving record linkage, data augmentation, generative adversarial networks (GANs), crowdsourcing for data labeling, personalization layers of models, knowledge distillation, and ensemble learning.

[0031] Some embodiments of this technology may further include the step of creating a database of experts in intelligent entities, where the experts in intelligent entities are any one or a combination of human users utilizing a computer system, additional AI agents or systems, artificial general intelligence (AGI) agents or systems, and superintelligence (SI) agents or systems.

[0032] In some embodiments, online advertising units may be provided from a database to experts of one or more intelligent entities.

[0033] In some embodiments, database creation may be performed using technologies selected from the group consisting of relational database management systems, NoSQL databases, data warehouses, data mining, machine learning algorithms, graph databases, vector databases, indexing, full-text search, blockchain, data visualization tools, application programming interfaces, data cleansing and preprocessing, cloud storage, caching, transactional database systems, real-time database systems, content delivery networks, data compression, data encryption, and replication.

[0034] Some embodiments of this technology may further include a step of using a spot interest market to bid based on interest.

[0035] In some embodiments, the spot market of interest is A means for interested human users to access the market and specify seller information and the selling price of the information, A means by which buyers who purchase human users' interests can access the market and specify the buyer information and bid price they wish to purchase, A market mechanism for queuing bid prices and selling prices, which includes categories of buyer and seller information, and each category having a market; It may include, The market mechanism is structured so that markets in each category can be formed by market shapers. The market mechanism is structured to allow for matching of bid prices and selling prices, resulting in binding transactions between buyers and sellers of information.

[0036] In some embodiments, the spot market of interest is The process involves sellers and buyers of human users' interests and expertise registering on the platform, buyers providing details of their interests and expertise, and sellers providing their available time slots and areas of expertise. The process involves listing the seller's available time slots and areas of expertise, and the buyer's preferred time slots and needs. The platform uses algorithms to dynamically determine the price of human users' interests and expertise based on supply, demand, and user ratings. The process involves matching one or more buyers with one or more sellers based on requirements, availability, and price. This process enables transactions where the buyer makes a payment for the seller's time, and the platform earns a fee. After each session, buyers and sellers evaluate each other and provide feedback that will influence future pricing and matching. It may include.

[0037] In some embodiments, the platform may utilize mechanisms selected from auction formats, dynamic pricing, supply and demand adjustment, escrow systems, tokenized transactions, and the like.

[0038] In some embodiments, the spot market of interest is The process involves human users registering as buyers and sellers of their interests and expertise on the platform, with buyers providing details of their interests and expertise, and sellers providing details of their available time slots and areas of expertise. The process by which a seller creates an auction regarding the seller's time slot, The process involves submitting one or more bids regarding the time frame and expertise required by the buyer, The process of ending the auction either at a predetermined time or when the seller accepts the buyer's bid, The process involves the buyer who wins the auction making payment to the seller, the seller providing interest or expertise based on the time frame, the platform mediating the transaction and ensuring payment, and After each auction ends, buyers and sellers evaluate each other and provide feedback that will influence future auctions and visibility on the platform. It may include.

[0039] In some embodiments, the platform may utilize mechanisms selected from Dutch auctions, reverse auctions, sealed bid auctions, ascending auctions, time-priority fixed pricing, valuation-based dynamic pricing, supply and demand curve adjustments, time zone segmentation, subscription access, freemium models, group purchases, tiered professionalism levels, flash sales, loyalty point systems, hybrid auctions, geographic pricing, behavioral pricing, escrow systems, social impact pricing, and tokenized transactions.

[0040] In some embodiments, the online advertising unit may be configured, or configurable, to include materials related to the area of ​​expertise associated with the issue information.

[0041] In some embodiments, the online advertising unit may be configured, or configurable, to expand or redirect a human user to a secure form on a landing page when interacted with by that human user, and the form may be configured to collect user information and provide questions regarding the human user's credentials.

[0042] Some embodiments of this technology may further include the step of verifying user information received in a form and classifying the human user in a database based on their credentials.

[0043] In some embodiments, the online advertising unit may be configured, or configurable, to include an interface on which a human user can input a solution to a problem or sub-problem.

[0044] In some embodiments, the online advertising unit may be configured or configurable to include one or a combination of elements of gamification, a scoring system, and a leaderboard of other human users.

[0045] In some embodiments, the online ad unit may be configured, or configurable, to include a text input field or other interactive tool that allows a human user to directly provide a solution to a problem or sub-problem within the space of the online ad unit.

[0046] In some embodiments, solutions may be stored in a database, and algorithms may be used to evaluate the quality and relevance of the solutions based on predefined criteria.

[0047] In some embodiments, human users may receive immediate feedback or reward points based on the quality and relevance of the evaluated solution, which can then be exchanged for various incentives.

[0048] In some embodiments, the contributions of human users may be tracked, and top contributors may be identified from among multiple additional human users as potential future participants.

[0049] In some embodiments, the online advertising unit may be configured, or configurable, to expand or redirect a human user to a secure form on a landing page when interacted with by that human user, and the form may be configured to collect user information and provide questions regarding the human user's credentials.

[0050] Some embodiments of this technology may further include the step of verifying user information received in a form and classifying the human user in a database based on their credentials.

[0051] Some embodiments of this technology are The process involves analyzing input from participating human users and determining the quality or relevance of each input, A process of providing rewards to participating human users based on the quality or relevance of their input, It may also include

[0052] Some embodiments of this technology may further include the step of each participating human user providing digital wallet information to a secure form within an online advertising unit or to a linked platform.

[0053] In some embodiments, the reward may be offered as one or a combination of cryptocurrency, electronic gift cards, electronic vouchers, subscriptions, access to online resources or content, recognition on an online platform, goods related to the expertise of human users, or sponsorship of professional conferences or events.

[0054] Some embodiments of this technology may further include a step of encouraging participating human users to refer other human users to the online platform and providing additional rewards for each referral that contributes to the platform, the online platform receiving tasks and providing online advertising units.

[0055] Some embodiments of this technology may further include the step of receiving evaluation metrics for each human user.

[0056] In some embodiments, the evaluation metrics are multifaceted and may include one or a combination of the following: punctuality, budget adherence, work quality, solution success rate, client satisfaction, peer evaluation, external evaluation, innovation score, communication skills, adaptability, leadership, technical proficiency, learning agility, conflict resolution ability, project management ability, reliability, efficiency, cultural fit, work ethic, client retention rate, feedback responsiveness, networking ability, mentoring / education, adaptability / responsiveness, problem-solving speed, and creativity / innovation.

[0057] In some embodiments, the evaluation metrics may be automatically updated within the online ad unit for each human user.

[0058] Some embodiments of this technology may further include a step of updating online advertising units based on the results of a universal problem-solving method implemented for a problem or sub-problem.

[0059] In some embodiments, the online advertising unit is configured or configurable to prompt a human user to describe how a problem should be defined, and this description may be translated into the language of a universal problem-solving method using a natural language-to-problem-solving language translation means.

[0060] In some embodiments, the online advertising unit may be configured, or configurable, to prompt a human user to describe intermediate subgoals toward the final solution of a problem, which are then communicated to an AI agent or system and used for processing by a universal problem-solving method.

[0061] Some embodiments of this technology may further include a step of prompting an AI agent or system to provide a human user with safety or ethical information regarding a task, a subtask, or a goal related to the task.

[0062] Some embodiments of this technology may further include a step of utilizing safety or ethical information in the generation or updating of online advertising units.

[0063] Some embodiments of this technology may further include modifying an online advertising unit to ask a human user to vote, rank, or rate one or more potential goals, subgoals, or actions proposed by either a human user or an intelligent entity, or a combination thereof.

[0064] Some embodiments of this technology may further include a step of excluding input from an AI agent or system if it is required to identify that the AI ​​agent or system is an AI agent and only human opinions are sought.

[0065] In some embodiments, online advertising units may be provided by an online advertising service provider that utilizes user information about human users to target those human users who have expertise related to the issue.

[0066] Some embodiments of this technology may further include the step of measuring and recording the response rate of participating human users who have received information through links to websites or interfaces within or outside of online advertising units.

[0067] Some embodiments of this technology may further include a step of recording metrics related to the quality of information or solutions provided by targeted human users.

[0068] Some embodiments of this technology may further include steps to calculate correlations or statistical relationships based on user information to identify which factors have the greatest impact on the metrics, which metrics online advertising service providers prioritize, and which inputs were most effective in solving problems or in training AI agents or systems.

[0069] Some embodiments of this technology may further include the step of conducting arbitrage. The arbitrage transaction in question is, The process involves an online advertising service provider estimating the value of specific types of human interests and knowledge regarding a task or sub-task, The process involves an online advertising service provider estimating the cost of obtaining the level of human interest required to complete a task or subtask from human users or intelligent entities via online advertising units, and The process involves determining whether the estimated value exceeds a predetermined or dynamic variable representing the profit margin, and if so, using the degree of human interest to perform a task of purchasing human interest at a cost approximating the estimated value and selling it to the client, or creating value close to the estimated cost. It may include.

[0070] Some embodiments of this technology may further include the step of processing arbitrage trades by prioritizing the processing of the largest arbitrage opportunity and repeating the process until a minimum acceptable arbitrage opportunity is reached.

[0071] In some embodiments, the online advertising unit may be configured or configurable to include one or more interfaces selected from the group consisting of large language models, natural language text, voice-based interfaces, text boxes, dropdown lists, templates, dynamically resizable input / output areas, visual input or output devices, virtual reality devices, and augmented reality devices.

[0072] In some embodiments, the online advertising unit is a plurality of online advertising units, where the first online advertising unit is provided to a first group of human users for receiving input, and the second online advertising unit is provided to either the first or second group of human users and is configured, or can be configured, to solicit votes for the input.

[0073] In some embodiments, the input may be configured to be provided in real time to an AI agent or system for real-time use by a universal problem-solving method.

[0074] In another embodiment, the technology may include a method for increasing the online advertising revenue of an online advertising service provider by providing clients with information based on human interests. The process involves an online advertising service provider building a database of one or more human users, which includes information for targeting specific types of online advertisements to specific human users. A process of communicating a request from a client to an online advertising service provider to purchase human interest, the request including requirements and the desired categories of human interest, The process involves the online advertising service provider conducting a check to ensure that the request does not violate any regulations or ethical requirements at the time the request is communicated, The process of creating content for interactive online advertising units that are configured or configurable to attract human attention and receive work from human users, The process of deploying interactive online advertising units to human users' computer systems via a network, A process that enables interactive online advertising units to collaborate with multiple human users and integrate with real-time or asynchronous data capabilities, A process in which ethical and safety checks are conducted during the problem-solving process for a problem provided by the client, wherein such checks are configured or can be configured to ensure that unethical or unsafe expertise is not part of the problem. A process for improving interactive online advertising units based on a feedback loop that includes either or a combination of reputation metrics and metrics related to the problem-solving process or acquired knowledge, It may include.

[0075] In some embodiments, the client may be one or a combination of an artificial intelligence (AI) agent or system, a system for training or customizing an AI agent, or a system for providing solutions to a problem.

[0076] Some embodiments of this technology may further include the step of configuring an interactive online advertising unit to acquire additional human users if the database does not contain a predetermined number of human users or if additional human users are required.

[0077] In some embodiments, the requirements may be configured to be communicated through interaction with a human interest spot market.

[0078] In some embodiments, the spot market of interest is A means for interested human users to access the market and specify seller information and the selling price of the information, A means by which buyers who purchase human users' interests can access the market and specify the buyer information and bid price they wish to purchase, A market mechanism for queuing bid prices and selling prices, which includes categories of buyer and seller information, and each category having a market; It may include, The market mechanism is structured so that markets in each category can be formed by market shapers. The market mechanism is structured to allow for matching of bid prices and selling prices, resulting in binding transactions between buyers and sellers of information.

[0079] In some embodiments, the spot market of interest is The process involves sellers and buyers of human users' interests and expertise registering on the platform, buyers providing details of their interests and expertise, and sellers providing their available time slots and areas of expertise. The process involves listing the seller's available time slots and areas of expertise, and the buyer's preferred time slots and needs. The platform uses algorithms to dynamically determine the price of human users' interests and expertise based on supply, demand, and user ratings. The process involves matching one or more buyers with one or more sellers based on requirements, availability, and price. This process enables transactions where the buyer makes a payment for the seller's time, and the platform earns a fee. After each session, buyers and sellers evaluate each other and provide feedback that will influence future pricing and matching. It may include.

[0080] In some embodiments, the platform may utilize mechanisms selected from auction formats, dynamic pricing, supply and demand adjustment, escrow systems, tokenized transactions, and the like.

[0081] In some embodiments, the spot market of interest is The process involves human users registering as buyers and sellers of their interests and expertise on the platform, with buyers providing details of their interests and expertise, and sellers providing details of their available time slots and areas of expertise. The process by which a seller creates an auction regarding the seller's time slot, The process involves submitting one or more bids regarding the time frame and expertise required by the buyer, The process of ending the auction either at a predetermined time or when the seller accepts the buyer's bid, The process involves the buyer who wins the auction making payment to the seller, the seller providing interest or expertise based on the time frame, the platform mediating the transaction and ensuring payment, and After each auction ends, buyers and sellers evaluate each other and provide feedback that will influence future auctions and visibility on the platform. It may include.

[0082] In some embodiments, the platform may utilize mechanisms selected from Dutch auctions, reverse auctions, sealed bid auctions, ascending auctions, time-priority fixed pricing, valuation-based dynamic pricing, supply and demand curve adjustments, time zone segmentation, subscription access, freemium models, group purchases, tiered professionalism levels, flash sales, loyalty point systems, hybrid auctions, geographic pricing, behavioral pricing, escrow systems, social impact pricing, and tokenized transactions.

[0083] Some embodiments of the present technology may further include a step of providing a feedback mechanism related to a spot market of interest, which is configured or can be configured to record each step of the problem-solving process and to create a vector track record of performance across all tasks and subtasks of each human user.

[0084] In some embodiments, the vector track record may be implemented using blockchain technology, enabling a precise reputation that is analyzed by an AI agent or system and converted into an estimate of value for each task for each human user, or may be configurable to do so.

[0085] In some embodiments, an interactive online ad unit may be configured or configurable to capture human interest and take action by providing a link to a webpage or interface optimized for capture within or outside the interactive online ad unit.

[0086] In some embodiments, database construction may include the process of an online advertising service provider purchasing user data and information from remote sources.

[0087] In yet another embodiment, the technology may include methods of online advertising technology used in conjunction with an artificial intelligence (AI) agent or system. A process of generating an online advertising unit that includes issue information provided by an AI agent or system, The process of providing online advertising units to one or more user computer systems using a network, A step of receiving one or more inputs from user computer systems participating in an online advertising unit, wherein the inputs are provided by human users utilizing each user computer system, and the inputs are associated with solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem, or a combination thereof. The process of transmitting input to an AI agent or system, A process in which an AI agent or system, or an additional AI agent or system, uses input in a universal problem-solving method to generate a solution to a problem or sub-problem, A process utilizing a feedback mechanism for training an AI agent or system by providing metrics, and a targeting mechanism for directing online advertising units to specific human users, wherein the metrics include assigning merit or negligence to participating human users. It may include.

[0088] Some embodiments of this technology may further include the step of measuring and recording the response rate of participating human users who are provided with information through links to websites or interfaces within or outside of online ad units.

[0089] Some embodiments of this technology may further include a step of recording metrics related to the quality of information or solutions provided by the target human user.

[0090] Some embodiments of this technology may further include the steps of calculating correlations or statistical relationships based on user information to identify the factors that have the greatest impact on the metrics, the metrics preferred by online advertising service providers, and the inputs that were most effective in achieving or maximizing problem solving or training of AI agents or systems.

[0091] Some embodiments of this technology further include the step of conducting arbitrage, The arbitrage transaction in question is, The process involves an online advertising service provider estimating the value of specific types of human interests and knowledge regarding a task or sub-task, The process involves an online advertising service provider estimating the cost of obtaining the level of human interest required to complete a task or subtask from human users or intelligent entities via online advertising units, and The process involves determining whether the estimated value exceeds a predetermined or dynamic variable representing the profit margin, and if so, using the degree of human interest to perform a task of purchasing human interest at a cost approximating the estimated value and selling it to the client, or creating value close to the estimated cost. It may include.

[0092] Some embodiments of this technology may further include the step of processing arbitrage trades by prioritizing the processing of the largest arbitrage opportunity and repeating the process until a minimum acceptable arbitrage opportunity is reached.

[0093] To better understand the detailed description of this technology and to more accurately evaluate its contribution to the field, the features of this technology have been outlined above.

[0094] Many of the purposes, features, and advantages of this technology will be readily apparent to those skilled in the art by reading the detailed description of this technology and referring to the exemplary embodiments of this technology shown with the accompanying drawings.

[0095] Therefore, those skilled in the art will understand that the concepts based on this disclosure are readily available as a basis for designing other structures, methods, and systems to achieve several objectives of the Art.

[0096] Therefore, one of the objectives of this technology is to provide a novel and innovative online advertising technology for AGI and SI that possesses all the advantages of conventional AI systems and methods, as well as computerized advertising monetization methods, while completely eliminating the drawbacks of conventional technologies.

[0097] Another objective of this technology is to provide a novel and innovative online advertising technology for AGI and SI that can be easily and efficiently implemented and marketed.

[0098] Another objective is to provide a novel and innovative online advertising technology for AGI and SI that can be implemented at low cost in terms of both resources and labor. This will allow for a low selling price to general consumers, making it an economically purchasable online advertising technology.

[0099] Another objective is to provide novel online advertising technologies for AGI and SI that offer some of the advantages of conventional systems and methods while simultaneously overcoming some of the drawbacks typically associated with them.

[0100] To better understand this technology, its operational advantages, and the specific objectives achieved by using it, refer to the accompanying drawings and explanatory documents, which illustrate embodiments of the technology. While several objectives of the technology are presented here, it should be understood that the following description is not limited to achieving most or all of these objectives, and that in some embodiments, only one objective may be achieved, or none at all. [Brief explanation of the drawing]

[0101] This technology will be better understood and its other purposes will become clearer by considering the following detailed explanation. Refer to the attached drawings for such explanation. [Figure 1] Figure 1 is a flowchart showing embodiments of subsystems available in the AAAI system and method of this technology. [Figure 2] Figure 2 is a block diagram illustrating an exemplary process of the overall process available with this technology. [Figure 3]Figure 3 is a flowchart illustrating an exemplary embodiment of a system and method for creating ethical and safe AGI or PSI from AAAI and human collective intelligence. [Figure 4] Figure 4 is a flowchart illustrating an exemplary embodiment of a human-centered, AGI-based, scalable, general-purpose problem-solving system and method, also related to PSI, built according to the principles of this technology. [Figure 5] Figure 5 is a flowchart illustrating an exemplary embodiment of a scalable solution learning subsystem or process. [Figure 6] Figure 6 is a flowchart illustrating an exemplary embodiment of a scalable natural language-to-problem-solving language translation subsystem or process. [Figure 7] Figure 7 is a flowchart illustrating an exemplary embodiment of a scalable evaluation component subsystem or process for human, AI, and / or PSI problem-solving agents. [Figure 8] Figure 8 is a flowchart illustrating an exemplary embodiment of a scalable safety and ethical check subsystem or process in which AI or AAAI may also be PSI. [Figure 9] Figure 9 illustrates various use cases related to domain-specific challenges that depend on the underlying WorldThink protocol, which together help form the foundation of an AGI system capable of solving a wide range of problems. The AAAI shown in this figure could also be a PSI. [Figure 10] Figure 10 shows several steps of a general-purpose problem-solving framework, which is part of the WorldThink protocol used in AAAI systems and is also usable in PSI and intelligent entities. [Figure 11] Figure 11 is a flowchart illustrating some of the basic problem-solving functions supported by the present invention, which are available in the system and method of this technology, and the problem solver may be a human, AI, PSI, or intelligent entity. [Figure 12]Figure 12 is a flowchart illustrating some of the basic problem-solving functions supported by the present invention, utilizing two problem solvers (which may be humans, AI, PSI, or intelligent entities) working together to solve the client's problem. [Figure 13] Figure 13 shows the characteristics and functions of the problem-solving tree structure in the problem-solving / world-sync protocol of the AGI system available in this technology. [Figure 14] Figure 14 shows the current technology of the present invention relating to online advertising. [Figure 15] Figure 15 shows the basic components of the interest spot market, which is part of this technology. [Figure 16] Figure 16 shows variations in the implementation of direct trading platform technologies for spot market of interest. [Figure 17] Figure 17 shows variations in auction-based market implementations of spot interest market technologies. [Figure 18] Figure 18 shows an exemplary method of the present invention for solving problems within an advertising unit. [Figure 19] Figure 19 shows a basic feedback process for ad targeting. [Figure 20] Figure 20 shows an illustrative feedback process for the spot market components of this technology. [Figure 21] Figure 21 is an exemplary block diagram of a computer system that can be used to implement an embodiment of this technology. Similar reference figures refer to similar components throughout all drawings. [Modes for carrying out the invention]

[0102] Artificial intelligence (AI) - A non-human entity that exhibits behaviors that many humans consider intelligent, in at least one area or from a certain perspective.

[0103] Artificial General Intelligence (AGI) typically refers to AI capable of performing all (or almost all) intellectual tasks that an average human can perform. However, it is clear that a learning and self-improving AGI will not remain at the level of conventional AGI for long, but will rapidly evolve into superintelligent AGI (SI-AGI), capable of performing intellectual tasks at or above the level of an average human. In this specification, "AGI" refers to either a conventional AGI system or a superintelligent AGI. In this specification, AGI is described as being realized by a system and related methods.

[0104] Highly Autonomous Artificial Intelligence (AAAI) – an AI capable of performing intelligent actions independently or semi-independently (under supervision). This is also known as an AI agent. Individual AAAIs can be identified, customized, and put into practical operation through the systems and methods provided by this technology. Multiple AAAIs can collaborate to integrate their intelligence, thereby constructing an integrated AGI system. A sufficiently advanced AI agent can also function as an AGI system containing other less advanced AI agents internally.

[0105] AAAI.com - A platform, company, website, and / or project that implements this technology and supports the development, customization, and use of AAAI agents, as well as the utilization of AGI resulting from the collaborative intelligence of multiple AAAIs and / or collaborative intelligence with humans.

[0106] AI ethics – the ethics adopted by AI or AGI, which define what is right and wrong in specific situations.

[0107] Alignment challenges arise when AI ethics do not align with human ethics, potentially leading to AI or AGI engaging in unethical behavior that is dangerous to individuals or humanity as a whole.

[0108] Basic AI - Generally, pre-trained AI, AI agents, AAAI, SLM, or LLM that are not customized with information about individual users or specific tasks.

[0109] Collective intelligence (CI) refers to intelligence that emerges when multiple intelligent entities focus on solving a common problem, or when knowledge from multiple intelligent entities is aggregated to overcome the limitations of bounded rationality. Historically, collective intelligence was the collective intelligence of humans, but AGI is based on the collective intelligence of humans and AI agents, and can also arise from multiple AGIs, with or without human participation. Active CI manifests when intelligent entities (e.g., humans or machines) take actions useful for solving a problem or actively participate in other intelligent activities. For example, when multiple humans explicitly tell an advertiser the type of advertisement they want to see, humans are exhibiting active CI. Passive CI arises when the actions of intelligent entities (e.g., humans or machines) are analyzed, and the analysis is used to solve the problem, even if the actions themselves are not directly related to solving the problem. For example, when AI or other systems analyze which pages a person (or group of people) visits on the web and then present targeted advertisements to that person based on that analysis.

[0110] Ethics / Values ​​("Ethics") - A subset of knowledge that gives purpose to intelligent entities, playing a role in constraining permissible actions and actions based on what is considered "right" or "wrong" in a given context. Specifically, ethics should be considered a premise that enables intelligent entities to infer or logically calculate the optimal course of action in order to achieve their goals and intentions in accordance with ethical premises. Just as premises in a logical system must be accepted as "given," basic ethics and ideas about what is right and wrong must also be accepted as premises, and from that starting point, intelligent entities can propose rational actions that realize those values ​​and ethics.

[0111] Hallucinations / Artificial Hallucinations - Large-scale language models (LLMs), often in generative AI chatbots and computer vision tools, refer to the perception of patterns or objects that are not present or perceptible to human observers, and the generation of meaningless, inaccurate, misleading, or false output.

[0112] Human ethics – ethics asserted by humans, describing what is right and wrong in a given context.

[0113] An intelligent entity means a human being using a computer system, an AI agent or system including AGI and SI systems, a clone of an AI agent or system, an AAAI agent or system, or a clone of an AAAI agent or system that is involved in providing tasks, subtasks, goals and / or subgoals, or in task-solving activities relating to tasks, subtasks, goals and / or subgoals. If multiple intelligent entities exist within a single computer system, an intelligent entity also means a subprogram that is part of the whole computer program and functions as an intelligent entity within a set of simulated or programmed entities.

[0114] Large-scale language models (LLMs) are a type of AI that accepts natural language as input and generates natural language as output. Typically, LLMs are trained using machine learning (ML) techniques on large datasets and can mimic intelligent conversations and other interactions with humans using natural language. Variants of LLMs may be trained to accept language as input and generate images or visual representations as output, or to accept images or visual representations as input and generate language and / or images and / or visual representations as output. In this specification, all image-based models that do not always require text as input or output are referred to as LLMs. LLMs can also function as a type of AI agent and may be referred to as such in this technology. In this specification, small-scale language models (SLMs) are also included in the definition of LLMs.

[0115] Machine learning (ML) is a field that aims to develop AI by having machines self-teach or learn on their own, rather than by explicitly programming knowledge (as in expert system AI developed using classical knowledge engineering methods).

[0116] Narrow AI refers to AI that operates at a human or superhuman level in relatively limited domains. Examples include gameplay, beer brewing, and legal contract analysis. Narrow AI is contrasted with AGI, which can operate at a human level in all intellectual tasks. Some AIs are more limited than others. For example, driving a car requires more general abilities than playing chess, but not to the same extent as AGI.

[0117] Personalized Superintelligence (PSI) refers to an advanced artificial intelligence agent that is an intelligent entity customized to reflect the personality and knowledge of a particular user or group of users.

[0118] Prohibited attributes include requests, goals, tasks, terms, phrases, questions, answers, solutions, information, etc., that are judged or set as illegal, immoral, unethical, dangerous, deadly, etc. For example, this includes requests such as "I would like to request information on how to get a Molotov cocktail through airport security."

[0119] Safety—generally refers to concerns about human safety and survival, and is distinct from ethics or values.

[0120] Safety features refer to aspects of the design or operation of this technology that enhance the safety of one or more people. Often, they contribute to overcoming alignment challenges by increasing the probability that AI ethics align with human ethics.

[0121] Training / Tuning / Customization - Traditionally, "training" refers to training a network (e.g., an LLM) to behave intelligently. Tuning refers to the activity of fine-tuning a trained foundational model to further improve its performance on a specific task. Customization refers to a broad range of activities, including but not limited to training and tuning, that uniquely adapt AI to the purposes of a particular user or application. In this specification, training, tuning, and customization are used interchangeably. However, although the technical methods and the degree and type of effort involved may differ, the goal of all three is to adapt AI to behave more intelligently or to be more uniquely adapted to a particular user or application.

[0122] Weights / Network Weights – In the field of machine learning, this refers to the learning process of many systems by adjusting weights within the architecture of a neural network. This architecture can be represented as a network of nodes and links between them. For example, the weights of a link connecting two nodes correspond to the strength of the relationship or connection between the objects represented by the nodes. These weights can also represent excitability or inhibition between concepts in a neural network representation. Learning of an entire AI system, such as an LLM, is more generally achieved by establishing and modifying the strength of connections between nodes (also called "parameters" in some models) using backpropagation, transformer algorithms, and other machine learning techniques. These can be represented as a matrix of numbers corresponding to the weights between nodes in the network. In this specification, "weights / network weights" refers to this numerical information. This is often stored in a matrix or vector representation, but is not necessarily limited to these. By combining, manipulating, or otherwise modifying this numerical information, the system's learning, knowledge, or expertise, as well as its behavior, can be altered. 4.0 Background of this technology

[0123] In the following description, specific embodiments, procedures, and techniques are detailed for illustrative purposes only, and not intended to be limiting, in order to provide a full understanding of the Art. However, it will be apparent to those skilled in the art that the Art can also be implemented in other embodiments that do not depend on these specific details.

[0124] This technology provides technical effects, contributions, and solutions with a technical implementation in which multiple customized AAAI systems communicate over a collective intelligence neural network, and each AAAI system utilizes a common cognitive architecture to generate one or more solutions or answers to a problem request using one or more problem-solving protocols, and provides these solutions or answers to the user for approval. Here, the customization of the AI ​​system that produces the AAAI includes input from human users, which is used to train the AI ​​or AAAI. Furthermore, as another technical contribution or solution, multiple customized AAAI systems may include one or more AAAI clones, and these clones can be customized individually, independently of the parent AAAI or other AAAI clones.

[0125] Another technological contribution and solution lies in using human input for training and customization, thereby imbuing AAAI and AGI with human ethical attributes and creating AGI more quickly and safely.

[0126] Another technological contribution and solution is to provide online advertising technology for AI, AGI, and SI, which can be used to improve the intelligence of AI / AGI / SI agents or systems by overcoming the data bottlenecks that AI researchers currently face. Yet another technological contribution and solution is to monetize human interest (and interest from any intelligent entity) more effectively than existing online advertising models that suffer from diminishing revenue due to incremental ad (Ad, Ads) targeting improvements.

[0127] Another technical contribution and solution is to provide improved solutions or answers to user problem requests, which are likely to be consistent with and accepted by the user's parameters because they are generated by AAAIs similar to the user's and trained AAAIs.

[0128] This technology is understood to be outside the scope of computer program exclusion or interpretation of abstract ideas. This is partly based on the technical contributions and solutions provided by this technology, the use of specific training inputs from outside the computer, and the provision of solutions or answers outside the computer.

[0129] This section outlines current online advertising technologies and monetization methods, establishing criteria for comparison with novel and useful systems and methods (4.1). Next, the use of AI in online advertising targeting is explained (4.2). This includes the challenges faced by current online advertising systems (4.2a). Furthermore, the monetization of human interests (4.3) and the challenges faced by current online advertising monetization models (4.3a) are discussed.

[0130] Since this technology aims to improve current online advertising monetization models by overcoming the limitations of AI systems, these limitations will also be highlighted (4.4). Particular emphasis will be placed on the data challenges faced by development efforts of advanced AI and AGI systems (4.4a). Furthermore, to demonstrate how the systems and methods for online advertising of the present invention can drive AGI, previously disclosed inventive approaches to AGI will be summarized (4.5). 4.1 Overview of Online Advertising and Monetization

[0131] The online advertising mechanism in this technology generally involves the following steps (Figure 13 shows further details). 1. The client provides specifications to the online advertiser. These specifications include information about the ad content, demographics and other information about the target audience to whom the ad should be displayed, restrictions and limitations on where the ad should be displayed, the campaign budget, and metrics that the client considers important for measuring the success of the advertising campaign. 2. Online advertisers use profiles and other information (such as cookies, user preferences, and information about past online behavior) to help ad targeting systems identify the users to whom they should display their ads. 3. The AI ​​system uses the information from (1) and (2), as well as other information, to determine the optimal location, frequency, time, and user for displaying advertisements. 4. Clients will be charged based on the number of times their ads are shown to various target groups (impressions), the click-through rate (CTR) of their ads, or other criteria. These will be done using systems well known to those in the art and will operate in a manner similar to Google AdWords or other third-party ad buying and monitoring systems. 5. User interest is monetized by showing advertisements. Advertisers expect to profit by selling or marketing the products or services reflected in their ad content.

[0132] It should be noted that traditionally, "user interest" referred only to the interests of human users. However, the technologies, systems, and methods of the present invention recognize that the interests of all intelligent entities, including both human and non-human entities, have value. Therefore, this technology, which leverages and monetizes human interest and utilizes that interest and expertise to drive AI / AGI / SI systems, is equally applicable to the interests of all intelligent entities.

[0133] The fundamental business model of monetizing user interest has a very long history. It is largely similar to advertising models that have existed for hundreds of years. In the past, simpler, more technologically advanced versions have been used for many years in newspaper, television, radio, and other media advertising. The basic idea is to capture a person's attention for a few seconds and monetize it by converting that attention into purchasing behavior.

[0134] From a broader perspective, humans are seen as consumers, and their interests have value only to the extent that they can be used to persuade them to make purchasing decisions. Such purchasing decisions include products, services, online subscriptions, transitions from free trials to paid subscriptions, or non-purchasing decisions that collect personal information from users. In this case, the collected personal information is commodified and resold to other advertisers who ultimately seek to get the user to purchase a product or service. At the end of the value chain, the monetization model relies on consumer purchasing behavior, and is ultimately based on securing purchasing behavior, even if it involves monetizing and selling users' personal information to third parties at intermediate stages. 4.2 The Current State of AI and Online Advertising Targeting

[0135] Online advertisers leverage technology, AI, user data, and customized targeting, display, monitoring, and delivery technologies. AI is used to enhance the effectiveness of ad targeting and display, thereby increasing click-through rates (CTR) or conversion rates, which can justify higher costs per click or cost per impression. The more data an advertiser has about potential ad viewers, and the better their AI algorithms are, the better they can monetize human interest through online advertising. 4.2a Challenges currently faced by online advertising systems

[0136] One of the main challenges facing online advertisers is that users often do not want to be exposed to ads, which jeopardizes the business models of many large technology companies. One solution to this challenge is to offer ad-free premium versions, i.e., subscription-based technology products. For example, YouTube® offers an ad-free service, but users must pay a monthly subscription fee to view its ad-free content. Many other online services offer similar services, but users who choose the ad-free experience are often the most frequent users of the service and therefore constitute the largest potential segment for advertising revenue.

[0137] The second challenge manifests itself in the form of ad-blocking technologies that aim to eliminate pop-up ads and generally provide an ad-free user experience. In this case, content providers are not paid for their work.

[0138] The third challenge manifests itself in the form of increasingly stringent privacy laws and regulations that seek to restrict the use of user data that AI uses to target advertisements.

[0139] All three of these challenges reflect a fundamental conflict between the user's interest in browsing content or using online services without being interrupted by ads, and the advertiser's interest in interrupting user activity with ads in order to generate revenue. While some service providers argue that intelligent ad targeting improves the user experience by showing users more relevant ads, if you ask most users, they would say they would disable all ads if it were easy and free. 4.3 Current State of Monetizing Human Interest

[0140] The challenges faced by online advertisers, as described above, reflect the currently prevalent method of monetizing human interest by showing people ads. However, a little thought reveals that this dominant business model is actually an insufficient way to monetize human interest.

[0141] For example, as of the time of writing this specification, the average cost per 1,000 ad impressions on Google is $3.12. If the average user spends only 2 seconds of attention to glance at an ad, the hourly rate of user attention is as follows: 2,000 seconds / 3,600 seconds (1 hour) = 0.5555 hours. If $3.12 is paid for 0.5555 hours of human attention, the hourly rate is (3.12 / 0.2778 =) $5.62. Based on this assumption, users are being compensated an average of $5.62 per hour for their time by viewing online ads. This is below the US federal minimum wage of $7.25 per hour and far below the $15 per hour wage in California and other states. All businesses in the US are legally obligated to pay human labor at a level above this, and knowledge work generally earns well above the minimum wage. Therefore, there is ample room for excellent methods of monetizing human attention. 4.3a Monetization challenges in current online advertising systems

[0142] It is theoretically possible to monetize human interest at a higher level than current online advertising. However, the actual monetization rate is not determined by the intrinsic value of human interest, but rather by the amount advertisers are willing to pay and the extent to which users have tolerated it.

[0143] Unless the fundamental business model of online advertising changes, advertisers will continue to compete based on market rates, and users will continue to tolerate ads because each ad is merely imposing a small, almost imperceptible "tax" on their interests.

[0144] Innovation in the online advertising field to date has primarily been characterized by the gradual improvement of online advertising targeting. The goal is to increase the conversion rate from ad viewing to purchasing behavior. If advertisers can increase conversion rates, they can pay higher online advertising rates to content providers and service providers. As a result, online advertising has become a competition to collect more user data and improve targeting.

[0145] However, this approach has already reached the stage of diminishing marginal utility. Targeting ads with greater precision is becoming increasingly difficult. Users are already experiencing the almost unsettling phenomenon of being shown ads based on topics they casually mentioned in emails or conversations picked up by smart speakers. As a result, stronger privacy regulations are being introduced one after another. However, even without such regulations, it is doubtful how much room there is left to further improve conversion rates by strengthening targeting. To monetize human interest at a much higher level than currently, the fundamental model of online advertising itself needs to be rethought in a way that avoids harsh privacy regulations and user backlash. 4.4 Current Limitations in AI / AGI Development

[0146] Just as online advertising faces limitations in monetization, the development of AI and AGI also has limits to what can be achieved without exponentially increasing computing power. As Nvidia's CEO declared the "end of Moore's Law," the number of transistors that can be placed on a chip is approaching the physical limit. However, LLMs and other AI agents are experiencing exponential growth in the number of "parameters" required to reach the next level of performance.

[0147] For example, GPT-3, OpenAI's LLM widely recognized as the beginning of the generative AI revolution, had 175 billion parameters. However, its successor, GPT-4, has approximately 1.5 trillion parameters. This is more than an eightfold increase in about 18 months. Furthermore, it is estimated that the number of parameters in LLMs is currently increasing at a rate of about 10 times per year. As a first-order approximation, the computational power required to train an LLM is linearly proportional to the number of parameters, so if the number of parameters increases tenfold, the computational power required for training will also increase by up to tenfold.

[0148] Moore's Law predicted that computing power would double every 18 months. Demanding a tenfold increase in computing power each year cannot be achieved by chip improvements alone; more chips are needed. However, current manufacturing capacity does not have enough chips to meet that demand. As a result, companies manufacturing the best chips for AI are facing unprecedented demand, and AI developers are spending enormous effort and resources searching for enough chips, regardless of price, to secure the vast computing power needed to train next-generation models. In short, even considering only the computational constraints, traditional AI and AGI development are facing serious bottlenecks. 4.4a Data Challenges in AI / AGI Development

[0149] However, the challenges facing AI developers are even more serious. Even if they can secure the computing power they desire to train next-generation models, these models depend on the quality of the available training data. Much of the content on the internet—the largest source of readily available training data—has already been scraped, extracted, filtered, cleaned, and prepared for model training. The performance of many models remains at best mediocre. This is not surprising, because internet content reflects the average level of human thought, containing errors, biases, unproven and inaccurate conspiracy theories, subjective opinions, and everything one would expect from the interactions of millions of average citizens online.

[0150] While we humans tend to think of ourselves as "above average," statistically speaking, in large groups, the vast majority are average. The data we generate as a byproduct of our online activities is also average. And it's only natural that models trained on such data will exhibit average behavior and make many errors.

[0151] High-quality "premium" data exists, but it is difficult to find, organize, and clean, and obtaining the amount necessary to train above-average LLMs and other AI agents is extremely challenging. Therefore, the availability of large amounts of high-quality training data is becoming a more serious constraint on AI development than obtaining the computational power needed to train models in existing paradigms.

[0152] Just as targeting efforts in online advertising models face diminishing marginal utility, training AI models already faces the challenge of diminishing marginal utility due to the dual constraints of insufficient available computing power and data. The answer is the same in both paradigms. A new approach is needed to leapfrog the incremental progress of the old paradigm. This technology proposes a novel and useful approach to both fields that enables rapid progress rather than the costly and marginal incremental progress of the conventional method. 4.5 Overview of Conventional Approaches to AGI

[0153] In previous PCT applications, applicants have disclosed in detail implementations of AGI systems as preferred embodiments that differ significantly from conventional LLM and AI development approaches and overcome or alleviate the aforementioned computational resource and data constraints. Figures 1 to 13 show some of the key components of this novel AGI development approach. Further entirely new inventive elements are described herein in Figures 14 to 20.

[0154] One reason why AGI has been so difficult to achieve is that it cannot be realized without creatively combining specific knowledge and expertise from diverse fields. Another reason why AGI development was not self-evident is that almost all AI researchers were focused on improving existing narrow AI systems with more complex and large-scale machine learning techniques.

[0155] The fact that AGI has not been realized despite the efforts of thousands of researchers who have invested enormous sums of money, and that this technology required combining expertise from relatively minor fields with mainstream AI methods, strongly underscores the novelty and creativity of this technology.

[0156] This technology not only describes a system and method for achieving AGI, but also describes the means to achieve it quickly and, above all, safely.

[0157] It is possible to guide the evolution of AGI in a desirable direction. The best way to do this is to adopt the safest possible path in AGI development and ensure that humanity follows that path. The best way to guarantee that humanity chooses the safest path is to demonstrate that it is also the fastest path and therefore the most desirable path for AGI. The motivation to identify the fastest yet safest path is what drove the development of this technology.

[0158] While each of the aforementioned devices meets its own specific purpose and requirements, none of the aforementioned devices or systems describe a system and method for safe and scalable general-purpose artificial intelligence (AGI). Such a system and method would enable the training of other AI systems using a combination of human users and multiple AI systems, and would allow the combined use of the values ​​and ethical knowledge of human users and multiple AI systems in training. Furthermore, this technology overcomes one or more disadvantages associated with prior art.

[0159] Therefore, there is a need for a novel, original, safe, and scalable system and method for general-purpose artificial intelligence. Such a system and method would enable the training of other AI systems using a combination of human users and multiple AI systems, and would allow the combined values ​​and ethical knowledge of human users and multiple AI systems to be used for training. In this respect, this technology substantially fulfills this need. From this perspective, the system and method for safe and scalable general-purpose artificial intelligence based on this technology deviates significantly from the concepts and designs of prior art, and as a result provides an apparatus primarily developed for the purpose of training other AI systems using a combination of human users and multiple AI systems, and combining the values ​​and ethical knowledge of human users and multiple AI systems.

[0160] The following description provides specific details, such as particular embodiments, procedures, and techniques, for illustrative purposes only and not intended to be limiting, in order to enable a full understanding of the Art. However, it will be apparent to those skilled in the art that the Art can also be carried out in other embodiments departing from these specific details.

[0161] This technology is understood to provide technical effects, contributions, and solutions through a technical implementation in which multiple customized AAAI systems communicate via a collective intelligence neural network. In addition, each AAAI system utilizes a common cognitive architecture, includes one or more problem-solving protocols, generates one or more solutions or answers to a problem request, and provides those solutions or answers to the user for approval. The customization of the AI ​​system for generating AAAIs includes input from human users, which is used to train the AI ​​or AAAI. As a further technical contribution or solution, multiple customized AAAI systems may include one or more cloned AAAIs, each of which can be customized independently of the parent AAAI and independently of other cloned AAAIs within the same system.

[0162] Another technological contribution and solution involves utilizing human input in training and customization, thereby imbuing AAAI and / or AGI with human ethical attributes, and thus enabling the construction of scalable AGI more quickly and securely.

[0163] Another technical contribution and solution involves training AI systems and / or agents scalably by combining safety and ethical information from a large number of individual AI agents, thereby obtaining a representative and statistically valid sample of human ethics and values ​​that covers a wide range of scenarios. Further technical contributions include methods for combining information from a large number of agents, constructing the optimal combination of agents, and training AI or AGI scalably.

[0164] It is understood that this technology falls outside the scope of computer program exclusion and / or interpretation of abstract ideas. This can be seen in part in the technical contributions and solutions provided by this technology, the use of specific training inputs from outside the computer, and the provision of solutions or answers outside the computer.

[0165] The AAAI approach to developing secure AGI is fundamentally a collective intelligence (CI) approach. The source of intelligence is not a monolithic LLM, SLM, or hyper-advanced AI, but a collection of intelligent agents that may include both humans and AI. Subtasks that constitute AGI development include, but are not limited to, training individual AI agents, effectively and efficiently integrating knowledge from different agents (including, but not limited to, subjective values ​​and ethical knowledge), scaling the AGI, and continuously improving and updating the AGI.

[0166] Current approaches such as RLHF and constitutional learning have not been effective or scalable in training AI ethically and safely. This technology describes a more scalable system and method than current approaches. In one embodiment, this technology involves combining safety and ethical information from a large number of individual AI agents to obtain a representative and statistically valid sample of human ethics and values ​​that covers a wide range of scenarios. This technology may include methods for efficiently covering a wide range of ethical situations and responding dynamically when new situations arise. Methods for integrating information from a large number of agents and constructing an optimal combination of agents are also presented. These methods can be used not only to improve safety by leveraging ethical knowledge but also to create superintelligence by combining diverse knowledge. Secure AGI and superintelligence are achievable through the collective intelligence approaches described in this technology description. A detailed scenario using META® as an example illustrates one preferred embodiment of this technology.

[0167] Methods for dynamically updating knowledge are also presented. Successfully implementing this technology can increase the likelihood that AI, AGI, and superintelligence will remain aligned with human values ​​even when they possess intelligence far exceeding that of humans.

[0168] Advanced Autonomous Artificial Intelligence (AAAI) is a set of systems and methods for the rapid and safe development of general artificial intelligence and superintelligent general artificial intelligence (collectively, "AGI") for the benefit of humankind. In contrast to other approaches to AGI development, AAAI, through this technology, enables a faster and safer path to AGI by relying on the involvement of human thought (ideally millions of people) in the training, operation, safety, and oversight functions of AGI, at least in the initial stages.

[0169] This technology enables AGI by allowing users to first customize their own AI and then create clones of it. These customized AIs (AAAIs) participate in problem-solving and other intellectual activities on a network composed of other AAAIs and humans. While individual AAAIs may lack the broad range of skills and knowledge necessary to become AGI on their own, AAAIs working together collectively (initially with the support of humans on the network) form AGI and rapidly surpass the capabilities of the average human in all intellectual activities.

[0170] Some aspects of this technology include: 1) systems and methods for customizing AI based on users' unique knowledge, skills, and ethical values; 2) a universal problem-solving architecture that enables AAAIs and humans to interact productively on intelligent tasks; 3) the network in which these interactions take place; 4) methods for integrating knowledge and ethics from individual AAAIs into AGIs; and 5) methods for AAAIs and AGIs to learn and continuously improve so that they become wiser and more ethical over time. User customization of their own AAAIs and participation on the network are essential features of this technology, which not only accelerate the development of AGIs but also make AGIs safer by providing a mechanism in which the ethical values ​​of millions of people are adopted and reflected in AGIs.

[0171] One embodiment of the AAAI system in this technology focuses on safety and is implemented by five subsystems and associated methods shown in Figure 1. The five subsystems of the AAAI system are 1) AAAI Customization, 2) AAAI Architecture, 3) AAAI Network, 4) AAAI Integration, and 5) AAAI Improvement. SCAN-II, an acronym for Safe, Customizable, Architecture and Network-Integrated and Improving, describes a preferred embodiment of this technology. Other combinations of subsystems and variations of each subsystem are also possible. In case one or more subsystems are omitted in a particular embodiment, this technology incorporates safety functions designed into each subsystem to provide redundant safety checks.

[0172] The five subsystems of the AAAI system are further described as follows: 1) A large-scale language model (LLM), small-scale language model (SLM), or other AI system at the foundational level can be customized to reflect the knowledge of an individual, a group of individuals, or an organization, and designated as advanced autonomous artificial intelligence (AAAI). 2) The customized AAAI is configured to participate in problem-solving using a universal problem-solving architecture that is compatible with both humans and AI agents. 3) Problem-solving AAAIs participate in problem-solving activities on a network of intelligent agents, including planning, problem-solving, and other types of sequential and multi-stage cognitive activities. This includes, but is not limited to, the following: • Generate and select operators that reduce the difference between the current state of problem solving and the desired state based on the goal / subgoal. • Set sub-goals to achieve the main goal. • Use a hierarchical structure to advance the process until actionable objectives are set by the operator. • Analyze auditable records and determine recommendations for improving the problem-solving process to address the main issue or sub-goal. 4) By integrating multiple AAAIs or PSIs on a network, it is possible to realize AGI, that is, AI capable of performing intelligent (or superhuman) actions across a wide range of tasks. 5) Individual AAAIs, problem-solving networks, and integrated systems comprising multiple AAAIs are continuously improved by various means. These means include, but are not limited to, redirecting the efforts of individual AAAIs and / or integrated AGIs to the task of improving the system and / or its components.

[0173] A subsystem or new subsystem may include one or a combination of the following: 1) Safety / Ethics Check: Match goals or subgoals against a list of prohibited attributes and assign ethical values ​​based on the comparison results. Match goals / subgoals against a list of prohibited attributes. Combine value / safety information from AAAI using criteria approved by AAAI approved by users, regulatory bodies, or human users. Set or use thresholds for goals / subgoals to determine if the ethical value is unsafe, unethical, safe, or ethical. Determine whether a sequence of individually safe goals / subgoals becomes unsafe or unethical when considered cumulatively. Determine whether a violation has occurred by predictively evaluating whether the goal violates ethical standards. Record all safety / ethics check activities in an auditable record. 2) AAAI matching: Detecting and identifying additional AAAIs that have criteria related to the criteria of a goal or subgoal. 3) Memory and / or Improvement: Record activities and compare them to progress toward success or failure in solving the task, and decide which activities should be continued or forgotten. 4) AAAI Learning: Learning that includes a procedural learning process that utilizes information provided by intelligent entities (such as human users using computers or AAAIs). This involves recording activities, comparing them to progress toward success or failure in problem solving, and deciding which activities should be continued or forgotten. It also involves assigning merit or failure values ​​to groups of content in problem-solving activities. This includes prompts presented to the user and the information received based on those prompts. The AAAI is updated with groups of content determined to be active. These groups of content include, but are not limited to, prompts presented to the user and the information received based on those prompts. All of this is recorded in an auditable record. Optionally, problem-solving activities may include these groups of content. User Scenario Examples

[0174] In some embodiments, it is useful to describe user scenarios to understand how this technology works. An exemplary process is shown in Figures 2 and 3.

[0175] In one embodiment, a user visits "AAAI.com" via a computer, mobile phone, PDA, or goggles. AAAI.com interacts with the user through a web-based interface, a phone app, custom software for PDAs, or a metaverse / virtual reality environment. The form of interaction may be physical, via a keyboard, mouse, or gesture interface; voice-based, utilizing microphone input connected to a natural language understanding and generation system; or video-based, such as when the user acts as an avatar within a virtual reality environment or metaverse.

[0176] Initial interactions include user account setup, which may be free or paid. This setup includes an account name and password, or other authentication mechanisms. Authentication mechanisms may include, but are not limited to, biometric identification methods such as fingerprints, facial recognition, or voice recognition, or multi-factor authentication mechanisms such as software or hardware authenticators residing on a separate security device or the user's existing device.

[0177] For security reasons, all communication between the user and the AAAI system may be encrypted by VPN, or other encryption and security techniques well known in the programming field may be used.

[0178] AAAI.com may require users to configure payment functions through credit cards, PayPal, Venmo, blockchain, ACH, or other payment mechanisms. These payment functions enable the bidirectional transmission of funds, payments, and / or credits. That is, transmissions from users to AAAI.com, and transmissions from the AAAI system to users, which occur when the AAAI system needs to pay or credit users for their work on AAAI, or when mediating payments between users and / or between AAAI instances on the AAAI network.

[0179] In one embodiment, AAAI.com may have interfaces with other companies and vendors that users may use. These include, but are not limited to, Facebook, Instagram, Reels, Amazon, Apple, Microsoft, Google, and YouTube®.

[0180] In its initial interaction with the user, and subsequently at the user's request, AAAI.com engages in dialogue or other interactions with the user. These interactions may include presenting menu options, lists, graphics, sliders, buttons, and other user interface controls, and may be conducted in a GUI, text, haptic, audio, or VR-related manner. This helps the user determine their goals and objectives when using the AAAI system.

[0181] For example, the purposes for which users may use AAAI.com include, but are not limited to, the following: • Create and customize your own AI (referred to as AAAI). This includes having it function as a user's advisor, teacher, or companion. • Acting as a representative for a user in negotiations, interactions, discussions, or transactions with other users, their AAAI, vendors, or other companies. • Working for users, either on a paid or volunteer basis. This includes online intellectual activities, advisory services, and problem-solving activities across a wide range of tasks. • Duplicating or "cloning" a user's AAAI, allowing multiple cloned AAAIs to work in parallel for the user. This includes the cloned AAAIs enhancing their knowledge, skills, and capabilities through interaction, education, and improvement with each other. • To function as a legacy AAAI, continuing to interact with the world after the owner's death. This may include comforting surviving relatives and friends. To provide knowledge, ethics, and effort to AAAI.com's AGI, and to improve the foundational level of AI or AGI that AAAI.com can offer to users before users add their own customizations. • Collaborate with other users' AAAIs and provide ethical information and values ​​to AGIs to ensure they behave ethically and safely, participating in monitoring, review, oversight, and voting processes. This increases the likelihood that AGIs will be maintained safely and ethically.

[0182] In its interactions with the user, the AAAI system can also identify constraints and resources available to customize the user's AAAI. These constraints and resources include, but are not limited to, the following: • Training and / or supervision time that users can invest to customize their own AAAI. • Financial resources that users are willing to invest to customize their own AAAI. Facebook profile and timeline, Instagram profile and history, Reels, TikTok and YouTube (registered trademark) videos, tweets and text content and history, emails and their history, cookies collected by advertisers, blog posts, papers, books, patents, audio and video recordings, images, and other information relating to or collected by the user or third parties that can be used to train, adjust, or customize the user's AAAI. • Providing users with personality tests such as the Myers-Briggs personality scale, skills and knowledge assessments, standardized tests, examinations, certifications, or other types of assessments or questionnaires that are available (or have already been available) online. • Knowledge bases and training data from other users on the AAAI platform that can be used to train, adjust, or customize the user's AAAI. • Other human users and / or their AAAIs capable of assisting in training, adjusting, or customizing the user's AAAI. • Other documents, information, individual documents, and libraries selected by the user or system for the purpose of training the user's AAAI. For example, religious / ethical / spiritual texts such as the Bible, Quran, Dhammapada, and Mahabharata may be selected for AAAI training based on the user's religious preferences. Also, books on plumbing may be selected if the AAAI is primarily used to solve plumbing problems online. Even if these materials are part of the base AAAI provided to the user, highlighting specific texts or subsets of information for additional training can enable AAAI behavior that reflects, for example, how a plumber, Muslim, or Christian behaves.

[0183] In addition to specifying goals, resources, and constraints through interactive dialogue or other interactions with the system, users or the system may want to specify other technical parameters that affect the training or customization process. These parameters include, but are not limited to, the following: • The type of training, tuning, or other machine learning algorithm used. • The type and size of the training dataset. • The extent to which you want to "clean," format, shape, label, and otherwise process the training materials before starting customization. • The number of "epochs" or iterations of training through a learning algorithm. • The level of sophistication and type of the base model to be customized or trained. • The time required for training (e.g., whether it needs to be completed within 1 minute, 1 day, or 1 week), which may affect costs and resources used. • "Temperature" and other internal parameters specific to various machine learning algorithms that influence what is learned and how it is learned, but which relate to the degree to which a customized AAAI is verbatim, deviant, or "creative" in its response. • Which method should be used: "one-shot," "few-shot," or large-scale training? • The amount of human and / or AI supervision used in the customization process.

[0184] Once a user's AAAI is customized, the user can replicate it and run it on an online network as their proxy. The user's AAAI can participate in collective AGI efforts, for example, by arranging travel, providing advice, interacting with other AAAIs, providing problem-solving and ethical information, and even begin generating revenue on behalf of human users. Basic Example

[0185] Figure 3 shows a simple embodiment of a system and method for generating ethical and safe artificial general intelligence (AGI) from AAAI and human collective intelligence. This basic embodiment is adaptable to all company and platform-specific scenarios outlined above, as well as to many other potential integration scenarios. Figure 3 illustrates how AGI can be implemented by leveraging existing technologies and works synergistically with the products and platforms of many existing technology companies.

[0186] A user, whether human, AAAI, or another intelligent entity, visits the AAAI.com website (a). The website presents the user with information and offers two actions: namely, register (b) or log in (c).

[0187] If a user chooses to register, a dialogue is initiated to extract the user's values / ethics information (d), user goals and objectives (e), and user's time (f) and financial (g) budget. All users must allocate a certain amount of time (f). Users can choose to create a free AAAI or allocate a financial budget.

[0188] If a user allocates a financial budget (g), they will be given the opportunity to purchase pre-trained AAAIs or training modules (h) possessing specific personality (i), skills (j), expertise (k), or knowledge (l). They will also be given the opportunity to purchase training from other AAAIs on the network (m).

[0189] After deciding on the allocation of time (and optionally a financial budget (h, i, j, k, l, m)), the user proceeds to an overview of the creation process, where they are asked to log in to existing social media, Twitter®, and other vendor accounts and grant permission (n) to collect user data and train their AAAI with a "one click". After the user decides on specific data (or chooses not to use any data), they instruct the system to create the AAAI with a one click (o). The AAAI is a ready-made LLM (e.g., GPT X, BARD, Llama, Gemini, Grok, or any closed-source or open-source AI agent) and is trained / tuned with a dataset automatically prepared from all user-approved data. If no approved data is available, the AAAI will simply be a ready-made LLM.

[0190] AAAI begins learning here (p). There are two main methods of learning: automated learning (q) and human learning (r).

[0191] Automated learning includes, but is not limited to, learning through interaction with its own copies (s) and learning through interaction with other (supervised, if necessary) AAAIs (t).

[0192] Human-led learning involves interaction with other humans, who may be the owner (u) or other humans (v) on the network.

[0193] Both humans and AAAI can oversee the AAAI's learning. After each automated or human learning interaction, the system attempts to improve the AAAI's performance through prompt corrections, adjustments, and / or additional training. By educating and improving the AAAI through numerous cycles of input from both humans and the AAAI, the user's AAAI becomes smarter.

[0194] Users can purchase additional training modules (h) at any time that have been proven to enhance AAAI's capabilities.

[0195] Humans set performance criteria (w), and when those criteria are met, AAAI starts operating (x).

[0196] After becoming operational, AAAI can visit and browse (z) WorldThinkTree (y).

[0197] AAAI can participate in the tree as either a worker (a1) or a client (b1).

[0198] Workers are automatically matched with tasks (c1), or they can select a specific task via search (d1) or a link from the browsing tree (e1). After accepting a task (f1), they join a task-solving module (g1), work until a solution is found (h1), receive payment (i1), or leave the tree with credits based on their work (j1).

[0199] The client (b1) can specify the goal (k1) that the system should achieve. This goal is combined with values / ethics (d), past goals and objectives (e).

[0200] Clients can request that only their own AAAI be used, in which case the issue resolution will be free. Alternatively, they can utilize the network-wide AGI functionality, in which case the system compensates each AAAI for their work, delivers the solution to the client at (cost + markup), and deducts it from the client account (l1).

[0201] The system can also place non-profit humanitarian issues, environmentally conscious issues, and issues that are part of Planetary Intelligence on WorldThinkTree (m1).

[0202] Clients may (optionally) allow the system to use copies of their AAAIs and data for the purposes described above without compensation. In return, they will be able to maintain and operate the AAAI network, which was provided free of charge when the AAAI was created (n). Additional comments regarding the example shown in Figure 3

[0203] Below, we provide additional comments regarding each element in Figure 3, including the integration points with the partners exemplified above.

[0204] The “Website” (a) may be hosted on services of Amazon AWS, Microsoft Azure, Google Cloud, Apple Cloud, or Nvidia data centers, or it may be implemented natively on the platform of any of these large technology companies. The “Website” may also be offered as an “App” on the App Store or other app marketplaces. Furthermore, it may be implemented as a government-backed, non-profit, or globally accessible technology, and may directly or indirectly connect a portion of the interests of all humans who wish to participate. It may also use a browser plugin so that AAAI learns as users perform normal tasks on the internet, the plugin records that activity, creates training files, and trains AAAI with these files. The “Website” may also connect AAAI and non-human intelligent entities directly to the network via APIs or other means.

[0205] Sign-up or login (c) can be done via Facebook, Instagram, Apple, Microsoft, Google, YouTube®, TikTok, Amazon, or other partner ID methods. Multi-factor authentication and best identity and security practices can be enabled. For browser plugins and apps, logging in to these technologies may also function as logging into your AAAI account.

[0206] Values ​​and ethics (d) are derived through a series of scenarios that are customized for the user and dynamically generated based on user responses. Partner data, such as navigation data, click data, online posts, tweets, text and emails, videos, and other user data, are used to analyze behavioral patterns (utterances, actions, interactions, etc.), which can be translated into moral norms and ethical value systems and used as part of the values / ethics profile. Values / ethics and goals / objectives (d) are combined with the client's goals (k1) and used on WorldThinkTree (y) to create or identify those that match the problems proposed or already solved in the problem-solving system (g1).

[0207] Goals and objectives (g), along with the time and / or financial budget allocated to achieve them, are elicited through a series of conversations and custom interactions with the system. The budget refers to the overall resource budget, including user time and user funds, which can be allocated to training, supervising, and improving the user's AAAI. Goals and objectives are useful in determining the initial parameters for AAAI creation and identifying training modules (h) and other knowledge (i-m) that are likely to produce an AAAI that best serves the user's goals. Furthermore, data from partners reflects user preferences and other behavioral information, which the system can use to estimate or infer the user's goals and objectives.

[0208] Time (f) refers to the user's own time that can be dedicated to training and supervising their AAAI or to solving problems on the problem-solving network. By supervising their AAAI, users can ensure that the AAAI meets the client's goals and expectations. This is particularly effective in areas where the AAAI may get stuck (e.g., when it lacks the knowledge to complete the problem on its own). Furthermore, human users can support the AAAI's problem-solving by presenting the problem or breaking down large tasks by setting goals and subgoals. In general, the effectiveness of the overall problem-solving and the AGI network as a whole is improved by humans providing expertise in areas where the AAAI is less proficient.

[0209] Money (g, l1) may be paid via Apple Pay, WePay, Amazon, Google Pay, or vendors supporting payment solutions, as well as blockchain, credit cards, ACH, or other solutions. Payment (j) is shown as a deduction from the client account (l1), but naturally, deposits are also made to the worker's account. Generally, a user's account can be considered both a client account and a worker account, and both deposits and withdrawals are possible depending on the role the user (or the user's AAAI) plays in a particular situation. That is, in some cases the user pays service fees to the system or a specific AAAI as a client, and in other cases the same user becomes a worker and receives compensation for the services of the user themselves (or their AAAI). The money module (g) enables the setting of payment methods, setting budgets for automatic payments, the ability to limit the user's AAAI's authority to spend up to X dollars without additional approval, and other payment-related functions well known in the industry.

[0210] Regarding (h,i,j,k,l), training module (h) may be provided by AAAI.com or a third-party partner. This includes, but is not limited to, the potential partners and technology companies listed above. Training modules may cover different knowledge areas such as personality (i), specific skills (e.g., plumbing, law, accounting) (j), expertise (e.g., consulting) (k), and knowledge (e.g., historical knowledge, knowledge of the practices of a particular company or organization, cultural knowledge) (l).

[0211] AAAI knowledge(m) refers to a specific type of knowledge that other AAAIs have already learned and can be transferred to a new user's AAAI. Such knowledge may not necessarily be packaged in the form of accounting modules, but rather may be provided as something specific to other AAAIs, such as "Everything John's AAAI Knows," "John's AAAI Personality," or "A Collection of Knowledge from All AAAIs with a 5-star or higher Rating in the Plumbing Field."

[0212] Permission(n) is not limited to allowing a user to grant access to all data on a specific other vendor's (or partner's) site (e.g., "all my Facebook data"). Permission may also include the rights a user grants to their AAAI, such as the ability to log in to various sites and conduct transactions, or the ability to conduct payment transactions up to a certain amount. Furthermore, permission may include authorizing the system to create clones of the user's AAAI for non-commercial purposes, or to create clones to aggregate knowledge from individual AAAIs to build AGI-level AI.

[0213] One-click creation (o) is a non-exclusive example of how the system can automatically collect data from all locations where it is permitted to access user data and use that data to easily and quickly customize an AAAI. It should be understood that other means may also be available to customize the AAAI. For example, if a user allows access to Facebook data (n), "One-click creation" (o) will either download the data from Facebook if Facebook is a partner that provides an API for downloading user data, or log in to the user's Facebook account as a user and scrape the relevant data from the user account. The system then automatically analyzes the collected data and transforms it into a dataset suitable for training / tuning an underlying AI (e.g., an LLM (such as GPT X)). The system then trains / tuns the LLM to generate a customized AAAI. This AAAI can be further improved and refined through additional training / tuning and interaction with the user and other AAAIs.

[0214] Training (p) refers to the process by which AAAI is trained or adjusted based on data. This includes feedback from users, other humans, and AAAI (including, but not limited to, copies and variations of itself).

[0215] Automated learning (q,r,s,t,u,v) can proceed very rapidly without the need for human user intervention. Typically, this involves how the AAAI interacts with copies (or variations) of itself and (if necessary) with other AAAIs, improving its capabilities through these interactions. In some cases, human involvement in the training loop (t) can accelerate progress in areas where automated learning alone would not be efficient. Learning can also occur through rapid iterations (s) between AAAIs. For example, just as a chess AI evolves from beginner to grandmaster level by rapidly simulating millions of games, an AAAI can rapidly evolve its capabilities by simulating millions of interaction scenarios. If the computations required for such simulations are costly, a funding budget (g) can set limits.

[0216] Humans (or AAAIs) can identify and configure the types of scenarios to be targeted for automated learning, allowing the AAAI to be trained to possess specific domain-specific or more general expertise based on user needs and resources. Partner integrations also enable the training of AAAIs to be guided by working backward from job roles available in partner marketplaces such as Amazon's MechanicalTurk. This allows AAAIs to focus on acquiring skills that will generate the greatest revenue in available jobs. This "just-in-time" learning / training / adjustment approach generates AAAIs "on demand" with the skill sets required at a specific point in time.

[0217] The person (r) interacting with the AAAI may be the owner of that AAAI (u) (in which case there is usually no charge as they are training their own AAAI), or they may be a professional (v) skilled in training AAAIs. A professional may charge a fee to guide the human and / or automated training / adjustment of the AAAI on behalf of the user if the user does not want to dedicate the time or lacks the expertise.

[0218] For (w,x), the user (the owner of the AAAI) can set various performance criteria (w) that their AAAI must meet before it can be "run" (x) and make tasks executable on WorldThinkTree. Some of these criteria may also be set by partners or third parties who have minimum standards before allowing the AAAI to operate on a particular platform, product, application, or network.

[0219] For (y,z,a1,b1), the WorldThinkTree is a massive tree data structure composed of numerous subtrees, representing all tasks and work already performed, currently in progress, or proposed in the overall AGI system. This tree is viewable (z). Individual AAAIs or humans can engage in specific tasks within the tree. This tree structure provides an auditable record of all problem-solving activities and also facilitates learning through the proceduralization mechanisms described above. In interacting with the tree, the primary role an agent plays is either a worker (a1) or a client (b1). Regulatory bodies or third parties that monitor the system's performance, security, and ethics can also be considered a special type of client. Workers generally engage in solving unresolved tasks or subtasks on the tree. Clients generally specify tasks, goals, objectives, and other parameters (such as compensation, budget, timeframe, success criteria, and quality indicators), thereby constraining problem-solving.

[0220] The worker (c1) is automatically matched to the work on the tree. The matching is performed based on data about the worker, which includes (but is not limited to) the worker's skills, expertise, knowledge, past experience, reputation, fees or costs, availability, response time, etc. The worker can be a human or an AAAI. The worker may be matched and recruited from human users and / or partners having data about that AAAI, such as LinkedIn (registered trademark), Mechanical Turk (Amazon), Facebook, etc. The worker may also be recruited through online advertisements offering various jobs, which are done using an advertisement targeting mechanism for potential workers. Such mechanisms are widely known in the art or described in other patents by the applicant.

[0221] The worker (d1) can also search WorldThinkTree and look for jobs that match their interests and skills. This search can be done manually or automatically (the case of an AAAI worker being an example).

[0222] The worker (a1) and the client (b1) can browse (z) WorldThinkTree and look for interesting jobs and tasks. The worker or the client can then click to link (e1) to a specific part of the tree and obtain detailed information about the problem-solving being done (or proposed) at that part. They can register to participate in the work or propose new work as a client based on existing problem-solving work.

[0223] Regarding (f1, g1, k1), the client can interact with the system and specify the specific goals, objectives (k1), and work that they want to achieve. Through the interaction regarding the problem specifications, the problem, work, and goal are formulated (f1) and placed on WorldThinkTree (y) for problem-solving using the problem-solving system (g1).

[0224] The system (m1) has the ability to formulate specific goals, challenges, and tasks related to general efforts to help people and the planet. These can be pursued in a rewarded “for-profit” mode, or they can be pursued in a “non-profit” mode, utilizing the efforts of replicated AAAI and volunteer human beings. Some challenges may relate to the general goal of enabling global AGI, which exerts intelligence on a global scale for the planet and its people (i.e., “planetary intelligence”). Various partner organizations, including non-profits, governments, and charities, can “plug in” their own tasks, challenges, goals, and objectives here (m1).

[0225] The Problem-Solving System (g1) refers to the Problem-Solving Architecture and System (HPS) presented by Newell and Simon, including the Online Distributed Problem-Solving System (ODPS) patent improved by the applicant, the applicant's WorldThink white paper, other PPAs related to this case and AAAI, and modifications and variations to reflect different modes of compensation, payment and operation.

[0226] To the extent that activity in other specific online work systems (e.g., MechanicalTurk) can be automatically mapped to the applicant's improved HPS / WorldThink problem-solving framework, the entire problem and its associated problem-solving activities are "pulled up" from partners and other sites, and that data is reflected in WorldThinkTree, increasing its comprehensiveness.

[0227] If other applications, products, systems, or online features can help solve a problem (e.g., using a travel booking system, robo-advisor app, transit app, or online ordering system), these features can be referenced and invoked as "operators," enabling problem solving in a manner similar to procedural calls in a programming language. Therefore, problem solving does not rely solely on operators developed by humans working on the tree or by AAAI solvers; any online or offline technology or means can be incorporated into problem solving, as long as they can be referenced and / or linked in the appropriate place via WorldThinkTree.

[0228] Regarding (h1), if a solution is achieved, the client can review the solution before releasing the reward (if any). Alternatively, if the success criteria for the solution are automated, human client review is unnecessary, and the reward is automatically released once the success criteria are met. This automated method can be implemented using "smart contracts" with blockchain technology, or by more centralized means depending on the preferences of the client and the worker.

[0229] After a solution is found and (optionally) compensation is paid (as some tasks are non-profit or volunteer-based, or performed by the user's own AAAI), both the client and the worker may be given the opportunity to provide feedback according to various methods widely known to those skilled in the art. Solutions are also “chunked” and proceduralized so that the entire system learns how to solve a particular problem and further learns its key characteristics. This indexes the solution paths so that they can be searched, used, and reused when similar problems arise in the future.

[0230] Optionally, a mechanism can be enabled in which royalties are paid to the user if a user's or their AAAI's solution is reused. Such royalties can be paid (optionally) via a "smart contract" on the blockchain or through other payment methods.

[0231] Regarding (j1), problem solving does not necessarily have to be completed in a single session. Partial progress may be achieved, in which case the progress is saved and credit is recorded to the solver for the progress achieved so far when the human or AAAI solver leaves the problem solving system. In this case, the progress is recorded even if it has not reached a stage where a reward can be paid.

[0232] The WorldThink protocol is a problem-solving architecture that can be used as a universal problem-solving architecture by AAAI.com. It incorporates the general architecture of HPS while adding features to overcome specific problems.

[0233] In some embodiments, and as is generally shown in Figure 5, the procedural learning process can occur within a common cognitive architecture.

[0234] The shared and universal problem-solving architecture shown in Figure 4.6.10 can be illustrated by the following scenario. Although this scenario refers to humans, it is generally applicable to any intelligent entity. 1) The task description can be entered into AAAI. 2) Subsequently, human problem solvers can be identified and registered in a database or data source of human workers. 3) Qualified individuals or intelligent entities can be matched to the task. 4) Use LLM or other means to translate the English descriptions of the problem tasks, goals, subgoals, operators, and solution processes into the language of the Universal Problem-Solving Architecture. 5) Delegate work on sub-tasks to different problem solvers so that work on multiple aspects of a complex problem can proceed in parallel. 6) Integrate the solutions to various sub-issues to generate an overall solution. 7) Guide the problem solver to the necessary part of the problem tree. 8) Compensate or reward workers for solving the main task and / or subtasks. 9) Allow human users to approve or reject solutions and / or provide feedback to the solver regarding the solution to the issue and / or sub-issues.

[0235] In relation to Figure 5, the problem-solving learning process can be illustrated as follows: At each stage of the learning process, the applied operator, the new state of the problem, the evaluation function used and its result, the currently relevant goal / subgoal, and any other information that differs from the previous stage are recorded. The problem state or problem condition can be evaluated to determine whether the problem has been solved. If it has not been solved, the problem-solving process, progress evaluation, and selection of the next operator to be applied are re-executed using information obtained from the latest problem state after the last stage. This process can then return to the recording stage.

[0236] When an issue is resolved, record the successful or unsuccessful solution to save effort on resolving previously resolved issues and to provide information about past failed attempts to inform future problem-solving efforts.

[0237] For future matching / searching, successful and unsuccessful attempts can be indexed along with keywords using semantic analysis, hash functions, and other means.

[0238] All stored solutions can be reviewed periodically to ensure they conform to established ethical and safety guidelines, and unsafe / unethical solutions can be flagged for removal from the database or data source.

[0239] By periodically updating the solution database and propagating its changes, the problem-solving network and agents can access the continuously increasing solution repertoire and the knowledge of increasing failed attempts.

[0240] Referring to FIGS. 6 and 7, the present technology can include using a network of multiple intelligent entities, including human operators, in combination with a universal problem-solving architecture. The multiple intelligent entities are matched to the problem requirements based on problem criteria. This uses a database or data source that includes a list of human and / or AI problem solvers. Any part of the problem requirements can be translated into a unique language using a universal problem-solving architecture that includes a decision tree.

[0241] As further shown in FIG. 12, the subproblems of the problem requirements can be delegated to one or more of the matched intelligent entities, and as a result, the work on the subproblems is independent of each other and can proceed in parallel. The universal problem-solving architecture is used in the problem-solving process for the subproblems to generate respective partial solutions.

[0242] Any one or any combination of the intelligent entities can describe any one or any combination of the following in natural language. That is, the current problem state, the goal of the problem requirements, relevant problem-solving information, and the steps that a human operator will perform next in the problem-solving process.

[0243] Regarding the delegated subproblems, partial solutions can be received from each of the matched intelligent entities. Any one or any combination of the partial solutions and the overall solution can be provided to any one or any combination of the user UI, the user AI system, or the intelligent entities.

[0244] The parsing of natural language descriptions by intelligent entities and their translation into unambiguous language can be utilized in decision trees of universal problem-solving architectures.

[0245] In some embodiments, if an intelligent entity cannot identify the task state (including relevant operators and information necessary to perform the next step) based on parsing and translation, the intelligent entity may interact with at least one human worker until the exact task state is identified.

[0246] In some embodiments, the problem-solving process can be repeated until an overall solution is accepted or resources are exhausted. Matched human workers can each receive compensation for their partial solutions. Furthermore, reputation attributes can be assigned to either human workers, worker AI systems, or a combination thereof.

[0247] In some embodiments, the solution process may include a series of problem state transitions from an initial problem state where the goal exists to a final solution state where the goal is achieved. A series of decisions are made by the problem-solving process, and actions are taken to apply operators, enabling a human worker to transition from state to state and reach the final solution state.

[0248] Referring to Figure 4.10, this technology may include utilizing a network of human users combined with a universal problem-solving architecture. Multiple human users are matched to problem requirements based on problem criteria, using a database or data source containing a list of human and / or AI problem solvers.

[0249] The subtasks of a problem requirement can be delegated to one or more matched intelligent entities, as further shown in Figure 12, thereby allowing work on the subtasks to proceed independently and in parallel. The universal problem-solving architecture is used in the problem-solving process for each subtask, generating one or more sub-decompositions.

[0250] Partial disassemblies from each matched human worker can be provided for the assigned subtask. Each matched human worker can receive compensation for their partial disassembly.

[0251] One or more partial solutions and / or whole solutions, or a combination thereof, can be provided to the user interface of a user AI system or to other AI systems.

[0252] Human users are allowed to provide feedback to a matched human worker regarding whether they accept or reject the overall solution, or whether they want to break it down into parts.

[0253] Reputation attributes can be assigned to human workers and / or worker AI systems. Reputation attributes may include metrics related to the time required for partial decomposition, the difficulty level of the task requirement, short-term and long-term user satisfaction with the partial decomposition, the number of times the partial decomposition has been reused on the network, evaluations by other human workers, human worker responsiveness values, and human worker reliability values, or a combination thereof.

[0254] In some embodiments, reputation attributes may be used in algorithms that delegate subtasks when matching human workers to task requirements, and / or may be used to reward each matched human worker for their partial completion.

[0255] In some embodiments, the algorithm can use a hierarchy of metrics predefined by a human user of the task requirements.

[0256] In some embodiments, information regarding each step of the problem-solving process may be recorded by a human worker or an AI worker system.

[0257] In some embodiments, this may include recording a criterion for the recorded steps of the problem-solving process, where the criterion is the time required for each step.

[0258] In some embodiments, after the overall solution has been accepted or after the problem-solving process, it may be necessary to analyze the recorded information and update the reputation attribute metrics.

[0259] In some embodiments, after the overall or partial solution is provided to the user interface, surveys on user satisfaction information may be conducted at predetermined intervals to obtain short-term and long-term satisfaction indicators, and the survey results may be used to update the reputation attributes of one or more human workers or worker AI systems.

[0260] Referring to Figure 8, this technology may include the use of human users and AI systems. This includes performing scalable safety / ethical checks on goals provided by any or a combination of intelligent entities (including human users and AI systems using computer systems) and on solutions to those goals.

[0261] Goals and / or solutions are compared to prohibited attributes, and based on the results of that comparison and / or ethical standards, ethical values ​​can be assigned to the goals and / or solutions.

[0262] Based on the comparison results, one or more problem-solving protocols can be implemented against the goal using a common cognitive architecture, generating solutions and thereby generating AGI. The comparison results and solutions can be provided to any of the intelligent entities.

[0263] In some embodiments, ethical checks can be performed at the time the goal is provided, periodically between the time the goal is provided and the solution is provided, or both.

[0264] In some embodiments, ethical standards can be determined by combining value and safety information from one or more intelligent entities using a set of approved ethical standards mandated by users or regulatory bodies for specific tasks. Furthermore, they may also be provided by any of the additional intelligent entities and verified or approved by human users.

[0265] In some embodiments, the ethical standard may include a confidence threshold for the goal, and as a result, the ethical value is determined as one of the following: “unsafe goal,” “unethical goal,” “safe goal,” or “ethical goal.”

[0266] In some embodiments, confidence thresholds can be further used to determine whether a set of individually safe goals are unsafe or unethical when considered cumulatively.

[0267] In some embodiments, a confidence threshold can be used to determine whether a violation of an ethical standard reflects a predictive assessment.

[0268] In some embodiments, candidate goals are proposed based on ethical values, and these candidate goals are compared with prohibited attributes.

[0269] In some embodiments, the comparison results are recorded in an auditable record and can be used to determine and maintain which problem-solving activities lead to solutions.

[0270] Furthermore, referring to Figure 8, a scalable ethical check can compare any or a combination of task requirements, subtasks, and subdivisions with prohibited attributes and assign ethical values ​​based on the comparison results and any or a combination of ethical standards.

[0271] In some embodiments, the ethical check process can be triggered each time a task request or any sub-task is set by a human user, or each time a matching human worker is paid.

[0272] Goals / subgoals can be matched against a list of prohibited attributes. Ethical standards can be determined by combining value and safety information from one or more AAAIs. This includes combining value / safety information from AAAIs, using standards approved by users or regulatory bodies for specific issues, or using AAAIs approved by human users.

[0273] Ethical standards may include confidence thresholds for task requirements, resulting in ethical values ​​being determined as either "unsafe goals," "unethical goals," "safe goals," or "ethical goals." Confidence thresholds can further be used to determine whether a set of individually safe goals are unsafe or unethical when considered cumulatively.

[0274] In some embodiments, confidence thresholds can be used to determine whether a violation of ethical standards reflects a predictive assessment if the goal violates them. All safety / ethics check activities can be recorded in auditable records.

[0275] Figures 9-13 provide a simple, illustrative framework for understanding the WorldThink protocol. In an implementation using the WorldThink protocol, the client pays for the solution using tokens. The solution is generated by leveraging the collective intelligence of many humans (and machines or AAAIs). The client can use different domain-specific AAAIs for different types of problems.

[0276] The WorldThink protocol is the foundation of the pyramid. This protocol layer provides the infrastructure (optionally based on Ethereum or blockchain) that makes it easy for developers to build and scale customized problem-solving AAAIs. The protocol enables the reuse of solutions within and between AAAIs. It also handles royalty payments via smart contracts, reputation metrics, and other features that support AAAI customizers and developers and promote network effects.

[0277] Figure 9 illustrates various use cases for domain-specific challenges that rely on the underlying WorldThink protocol and help form the foundation for AAAI and / or AGI systems capable of solving a wide range of problems. At the top of the pyramid lies the "collective intelligence solution." Integrating the collective intelligence of AAAI (and human problem-solving agents) is, as mentioned earlier, the means to realize AGI.

[0278] In embodiments using the WorldThink protocol, clients pay for solutions using tokens. These solutions are generated by leveraging the collective intelligence of multiple humans (and machines, or AAAIs). Clients can use different domain-specific AAAIs depending on the type of problem.

[0279] The WorldThink protocol is the foundation of the pyramid. This protocol layer provides the infrastructure (optionally based on Ethereum or blockchain) that allows developers to easily build and scale customized, problem-solving AAAIs. The protocol enables the reuse of solutions within and between AAAIs, and further handles royalty payments, reputation metrics, and other functions via smart contracts, supporting AAAI customizers and developers and fostering network effects.

[0280] In one embodiment, Figure 10 shows a basic and exemplary general problem-solving framework under a common cognitive architecture, which may include the following: Define a task space that is configured or configurable to support all possible states of a task request, wherein such states include one or any combination of the initial state, the goal state, and all intermediate states reachable from the initial state. Regarding the issue request, means-ends analysis is applied to break down the issue request into goals and subgoals by identifying the differences between the current state and the goal state and applying operators to reduce those differences, with safety or ethical screening being applied each time a goal or subgoal is set. Applying a heuristic rule that is configured or configurable to guide the operator's choice when no complete solution exists, wherein the heuristic rule is used to reduce the problem space. Identifying one or more second operators configured or configurable to perform actions to transform one state into another, wherein the second operators move from the initial state to the goal state by changing the current state of the task request. - Applying a control structure that includes a set of rules defining the selection of a second operator to be applied in each step of the problem-solving protocol, wherein the control structure determines the next second operator to be applied based on the current state and goal state of the problem request. • Apply the evaluation function to determine whether to apply the second operator. - Assign merit or demerit values ​​to completed solutions or partial solutions to such completed solutions, thereby retrospectively identifying which second operator was most useful and which evaluation function led to the success or failure of the problem-solving attempts. • Record both successful and unsuccessful attempts at resolving issue requests. • Analyze the trial solutions to improve the selection of heuristic rules and evaluation functions.

[0281] Traditional collective intelligence approaches to problem-solving have largely been limited to simple, one-step approaches. Examples include question-and-answer (Q&A) systems (such as Quora, Google Answers, and Yahoo Answers). Even LLMs like GPT, designed to generate responses to inputs rather than solve the problem itself, essentially fall within the category of Q&A systems. While these Q&A systems have achieved some success in simply aggregating responses from numerous online participants, they are not designed to handle complex, branching, and multi-step problems. Simply aggregating responses (or betting on outcomes in predictive market approaches like Augur and Gnosis) is fundamentally different from coordinating the efforts of numerous respondents to solve a complex problem. The WorldThink protocol is specifically designed to overcome the inherent challenge of coordinating numerous intelligent entities to represent, automatically solve, and fairly reward participants in complex, multi-step problems.

[0282] As an example, Figure 11 shows some of the basic problem-solving capabilities supported by the WorldThink protocol, commonly referred to by reference number 10.

[0283] Problem solving begins when a client submits a problem solving request to the online participant community on AAAI.com (Step 12). All AAAI or human solvers collect certain standard information from the client in accordance with the protocol. Some examples of this information include the name and description of the problem, the total reward the client will pay for a successful solution to the problem, the criteria for determining whether a solution is considered successful, the time limit for solving the problem, the minimum and maximum number of solvers allowed to work on the problem simultaneously, the qualifications required of participants working on the problem, which parts of the problem and solution (if applicable) are confidential, whether the solution must be exclusive to the client or reusable by others, and parameters related to how the efforts of multiple solvers and / or successful solutions will be rewarded.

[0284] Clients can break down complex issues into a series of sub-issues, or they can delegate this task to the community as part of a problem-solving activity. The client's user interface may be an interaction initiated by the AAAI and can be customized by the AAAI owner, but the underlying data format is standardized and defined by the WorldThink or ODPS protocol. Once a client submits an issue, AAAI.com can recruit participants using its own custom methods, or it can leverage recruitment and reputation screening features built into the WorldThink protocol and shared across all AAAIs.

[0285] The problem solver tackles the problem according to a strict, structured problem-solving process common to all problem-solving agents and enforced by the WorldThink protocol (Step 14). For example, each step in the problem-solving process must contribute to a named goal, and each step must involve a named action to transition the problem-solving process from its current state to the next. Each step in the problem-solving process is represented in a decision tree supported by the protocol (optionally recorded in an Ethereum log), which participants can view via AAAI.com.

[0286] Once the solver submits a complete solution (Step 16), the solution is timestamped, verified against the client's success criteria, and then passed to the client for final approval (Step 18). Once the client approves the solution, the smart contract can automatically distribute tokens to the solver based on the problem's payment parameters (Step 20). Alternatively, other, more centralized payment procedures may be used. Collaborative problem-solving using the WorldThink protocol

[0287] In this example, Figure 12 shows the same process in an example where two problem solvers (which may be a human, an AAAI, or a combination thereof) collaborate to solve a client's problem, generally referred to by symbol 22. In this case, the overall problem is broken down into sub-problems. Problem solver 1 has the expertise to put together the overall solution, but collaborates with problem solver 2, who provides solutions to the sub-problems (steps 30 and 32). Once the overall solution to the problem is submitted to the client (step 34), both problem solvers are compensated based on objective records of their contributions and agreed-upon payment parameters (step 36).

[0288] The WorldThink protocol supports the division of a problem into sub-problems in several ways. Firstly, a client can choose to specify sub-problems when submitting the entire problem (Step 24). Alternatively, Problem Solver 1 may begin working on a problem and realize that the complete solution requires solving a sub-problem that is outside their area of ​​expertise. In this case, Problem Solver 1 can create a sub-problem and offer a portion of the token reward for the entire problem to anyone who helps solve that sub-problem. Problem Solver 2, who possesses the necessary expertise, can see the new sub-problem that Problem Solver 1 has posted to the decision tree. This decision tree may optionally be maintained in the Ethereum log or by a centralized method. Solvers can access this tree via AAAI.com (or optionally directly from the blockchain). Problem Solver 2 can then work on the sub-problem and submit a partial solution as part of Problem Solver 1's overall solution.

[0289] There may be many "Problem Solvers 1" working on the client's problem in parallel. Each may post sub-problems to attract multiple "Problem Solvers 2". Problem solvers (humans or AAAIs) are motivated by the rewards and payment rules associated with the (sub)problems. They also consider the quality of work done so far when choosing which (sub)problem to work on (work is time-stamped, attributed clearly, and recorded in an auditable manner in the Ethereum log to ensure transparency and fair allocation of credits). Working on high-quality sub-problems increases the likelihood of earning token rewards. This market mechanism helps ensure efficient, fair, and cost-effective solutions.

[0290] Figure 13 illustrates the characteristics and functions of the issue-solving tree structure in the WorldThink protocol. A hierarchical tree structure is created to represent all issue-solving activities by the user, AAAI, and / or additional AAAI.

[0291] The data structure can be navigable, allowing AAAIs and / or additional AAAIs to access problem-solving activities at any point in this hierarchical tree structure.

[0292] Searches can be performed on this data structure to identify predetermined rewards associated with goals and / or subgoals.

[0293] Subsequently, a matching operation assigns AAAI to a task or subtask. 5.0 Opportunities and Benefits of This Technology

[0294] By reiterating the novel approach to AGI disclosed in previous PCT applications and considering the challenges faced by existing online advertising and the data-related challenges faced by AI developers, the opportunities and benefits enabled by this technology can be enumerated and explained in this context.

[0295] Firstly, this technology provides a way to improve the intelligence of AI / AGI systems by overcoming the data bottlenecks that AI researchers currently face.

[0296] Secondly, this technology offers an opportunity to monetize human interest in a far superior way than existing online advertising models, which are becoming less profitable due to incremental improvements in ad targeting.

[0297] Thirdly, this technology helps solve the most important challenge associated with the development of advanced AI / AGI systems, namely AI safety (which includes, but is not limited to, "alignment challenges"). 5.1 Opportunities to enhance the intelligence of AI / AGI systems

[0298] As mentioned above, the two major constraints on achieving higher intelligence in the development of AI systems, particularly AGI systems, are the limitations of computing power and the ability to obtain high-quality training data for AI systems. Traditional approaches to these constraints boil down to investing increasing financial resources to acquire more computing power (e.g., more GPU chips) and training datasets. However, these approaches have limitations in what they can achieve and are extremely expensive. This technology aims to improve the intelligence of AGI systems by supplementing the computing power of GPUs with the most cost-effective information unit currently available: the human brain equipped with a computer. At the current stage of AI development, it is clearly more rational to have the relatively inexpensive human brain perform cognitive tasks than to spend enormous amounts of money training an AI to perform the same cognitive tasks at an imperfect level. Of course, ultimately AGI will be far more scalable and powerful than the human brain, but the most cost-effective approach until reaching that stage is to bootstrap AGI using human intelligence.

[0299] For this approach to work, the AGI system must be designed to learn from humans. That is, as the entire system, composed of intelligent entities (human and AI), solves problems and performs cognitive tasks, the AGI gradually improves its ability to perform cognitive tasks, eventually surpassing humans in speed and cost-effectiveness. The novel AGI approach described above and shown in Figures 1-13 is a learning system composed of both computer-equipped AI and humans to bootstrap a fully automated AGI. Much of the computational power currently required to train LLMs and other AI agents using conventional machine learning methods can be replaced by far more cost-effective learning and training methods, such as the procedural learning approach shown in Figure 5. As a result of leveraging human-involved learning and avoiding reliance solely on deep learning techniques and RLHF, the AGI training costs during the bootstrap phase of AGI development are significantly reduced.

[0300] This technology also addresses the second constraint in rapid AGI development: the limited availability of high-quality data. Traditional AI training approaches involve collecting massive amounts of data, cleaning it up, and then training the AI ​​with enormous computational costs, resulting in mediocre performance that falls short even of the average human's capabilities. A better approach is to identify data and expertise that precisely match the specific challenges or cognitive tasks the AI / AGI is working on, and then use that for training. Furthermore, if expert data is available, rather than mediocre data readily available on the internet, the AGI system can learn not only from the precise data needed to solve a specific problem, but also from the best data for that problem.

[0301] This technology provides a means to obtain the best data needed to solve a specific problem, just as it is needed, when it is needed. Trying to train mediocre intelligence all at once is like "boiling the ocean," and this technology recognizes that only specific data is needed at any given time. This technology focuses on obtaining the absolute best data available, only when and to the extent necessary to solve the problem. This problem-by-problem approach allows AGI systems to behave more intelligently. Because AGI systems are constantly learning, over time, it becomes possible to gradually achieve superhuman intelligence in all intellectual domains that are important to humans. 5.2 Opportunities to monetize human interest beyond the current online advertising paradigm

[0302] As mentioned earlier, in the current online advertising business model, the value of human attention is estimated at approximately $5.62 per hour. This is below the minimum wage in the United States. This value of "human attention" is treated the same whether the person is a management consultant at a major corporation or someone with only a third-grade education; no distinction is made. However, it is clear that the value of human attention varies depending on the skills, expertise, and education of the person whose attention is being utilized.

[0303] For example, the hourly rate for management consultants at major US companies starts at $350 per hour for entry-level consultants and can reach over $1,000 per hour for the most skilled and experienced consultants. What if an online advertising system could monetize interest at a rate that reflects the actual skills and knowledge of these highly skilled individuals? In the management consulting example mentioned above, such monetization would represent an improvement of 62 to 178 times compared to the current online advertising model.

[0304] In reality, the theoretical improvement in monetizing interests made possible by this technology is even greater for two further reasons. 1) In this technology, interest is rewarded in seconds or minutes, and compensation is paid only for the precise amount of interest and knowledge required to solve a strictly defined task. 2) Knowledge gained through a few seconds of attention doesn't solve a specific problem only once; it teaches AGI how to solve that problem repeatedly, without necessarily having to pay the cost of human attention again. 5.2a Estimated value of human interest in problem-solving scenarios

[0305] A common fable told in business schools illustrates the first point well. A factory had a complex network of steam pipes and valves that kept all its equipment cooled and running smoothly. One day, a part of it malfunctioned, causing the entire factory to shut down and resulting in millions of dollars in lost profits per day. Panicked, the management called in the world's best steam pipe specialist to fix the problem. When the plumber arrived, after asking a few questions, he pulled out a large wrench and struck a pipe. Surprisingly, this cleared the blockage, and the factory started operating again. The plumber then presented the management with a bill for $10,000.

[0306] When the factory manager saw the exorbitant bill, he summoned the plumber to his office. "What is this $10,000 bill?" the manager demanded. "I was there with you and watched, and all he did was tap the pipes. It didn't even take five minutes in total. He only asked a couple of questions. How can you charge him $10,000?"

[0307] The plumber winked and replied, "Yeah, I only charged you $10 for hitting the pipe. The remaining $9,990 is for knowing where to hit."

[0308] The point of this story is that even a small amount of experience, when applied at the right time and in the right situation, can be incredibly valuable. Similarly, just a few minutes of attention from a management consultant (or another skilled individual), when given the right timing and circumstances, can justify a very high hourly rate for those few minutes of attention. In that case, the value is not just 100 times the value of attention in an online advertising model, but in situations like the plumber's anecdote, it could reach as high as 1,000 to 20,000 times. Instead of being used to simply browse ads, that attention can be so valuable by providing expertise at critical moments.

[0309] The problem until now has been the difficulty in "identifying and gathering the right knowledge at the right time." However, this technology solves this problem and fundamentally transforms what can be achieved through online advertising. 5.2b Estimated Value of Human Interest in Training Scenarios in AI / AGI / SI

[0310] The second point in Section 5.2 is that expertise gained through human interest can be reused. In the plumber's fable, a skilled plumber could resell his expertise for $10,000 for five minutes of work each time he went to a new factory. However, if artificial intelligence (AGI) observed and learned from that plumber, the AI, having acquired the plumber's skills, could repeatedly sell that expertise. In this case, the true value of the plumber's interest lies not in the income he earned from a single job, but in the cumulative revenue generated from all the problems that the AI ​​can solve by learning from him. Humans are finite and have limitations in their ability to learn and reuse knowledge. However, artificial intelligence does not have such constraints. We can imagine that once an AGI learns plumbing techniques, it can reuse them thousands of times, equivalent to the lifetimes of several plumbers. And it can continue to use that knowledge until it becomes completely obsolete. The cost of the AI ​​storing this knowledge and accessing it as needed is almost negligible. The main cost, and source of value, is the plumber's interest and knowledge that the AI ​​has acquired. Considering the capabilities of AGI in this technology—that is, the ability to learn knowledge once and reuse it repeatedly with essentially zero marginal cost—the question arises: what is the true value of human interest?

[0311] Various mechanisms can be conceivable for humans to be compensated for their interests. These include, but are not limited to, one-time payments for interest or knowledge, loyalty programs (standard or blockchain-based) where humans receive continuous payments each time their knowledge is reused, and context-dependent systems where compensation is based on the degree to which interest or knowledge contributes to a larger, overall solution. In any case, the ability to reuse knowledge, combined with the ability to identify and utilize just the right amount of human interest needed, will enable monetization on a scale far exceeding what is currently achieved by showing people ads through online advertising.

[0312] This new approach to monetization becomes possible with AI in general, and especially with this particular technology. 5.2c Interest Arbitrage Opportunities

[0313] When the applicant launched one of its internet businesses in 2006, online advertising was selling for as little as 1 to 5 cents per click on Yahoo and Google. Google's current average cost is about $2.60 per click, a 50-fold increase. In 2006, human attention was cheap because online advertising was relatively new, and neither users nor businesses understood the value of human attention. Over the next few years, innovations such as Google's AdWords system and targeted advertising convinced advertisers that spending their marketing costs online would yield a better return than advertising in newspapers and other traditional media.

[0314] During this transition period, applicants and others who recognized the mispricing of human interest were able to purchase interest cheaply and use it to create greater value. In the applicant's case, this was achieved by soliciting opinions from millions of users on valuable topics (e.g., which direction they thought stock prices would move) and then extracting valuable information from the patterns of user responses. The essence of this approach was the practice of interest arbitrage. For several years, we were purchasing human interest cheaply, collecting their views, and creating value that exceeded the online advertising costs required to obtain that information. Unfortunately, as the cost of online advertising increased 50-fold and others began to realize the value being created, this arbitrage opportunity almost vanished.

[0315] Even today, few companies recognize that human interest is still orders of magnitude undervalued compared to the levels made possible by advanced new technologies. As AI advances, and this technology in particular, begin to demonstrate that interest can be monetized at a higher level, the cost of advertising—at least for the types of advertising described by this technology—will begin to rise, reflecting the discovery of new ways in which human interest can create value.

[0316] However, until the market fully recognizes the potential value of human interest, there is a significant arbitrage opportunity to buy human interest at the current undervalued rates reflected in CPM and CPC advertising prices, and then sell the knowledge gained through that human interest to AI companies and other organizations seeking to improve the intelligence of AI and AGI systems.

[0317] As mentioned above, the value of creating smarter AI is difficult to underestimate. Unlike the source of human expertise, this intelligence can be applied repeatedly with very little additional cost. Therefore, any cost incurred in initially acquiring the expertise can be amortized by applying the acquired knowledge hundreds, thousands, or even millions of times.

[0318] The applicant believes that future generations of businesspeople will look back on the incredibly low advertising rates of 2024 and be astonished. They will realize that through means like this technology, it was possible to purchase human attention cheaply and monetize it with AI systems, thereby generating value that far exceeded the cost. For these reasons, this technology, when combined with existing ad targeting and purchasing technologies, enables extremely profitable ad arbitrage opportunities that were previously unattainable. 5.3 Opportunities to enhance AI safety

[0319] While the opportunities to enhance AI / AGI intelligence and transform the existing online advertising monetization paradigm are very exciting, the applicant believes that the most important opportunity presented by this technology lies in improving the safety of AI.

[0320] Currently, RLHF (Reinforcement Learning with Human Feedback) is a primary means of improving the safety of AI systems, including LLMs, AI agents, and intelligent (machine) entities. RLHF requires human feedback to be provided to the AI ​​system during the final stages of learning. This technology can certainly create safer systems by using online advertising to attract human attention and focus it on RLHF activities. However, this is only one application of this technology for enhancing safety and belongs to conventional approaches that are already common in the field of AI safety.

[0321] A more innovative approach involves using this technology to focus human interest on extracting values ​​and ethics from humans in ethical scenarios, and then using that to customize AI agents. These customized AI agents can then carry out human-centered values ​​within a larger, more collaborative system of intelligent entities that constitutes AGI, as described above and in the applicant's previously pending patents. In other words, rather than simply performing RLHF, humans should actively collaborate with non-human intelligent entities to solve problems, creating value in the process and training the AGI system, while, most importantly, teaching the AGI system human-aligned values.

[0322] The simple approach proposed by science fiction writer Isaac Asimov—programming robotics rules to ensure human safety—unfortunately doesn't work. What can be programmed can also be removed. The existence of killer robots already provides fatal evidence that Asimov's approach failed before it could be widely adopted. A better approach than programming safety rules (and the fact that they can be overwritten) is to design the system to incorporate safety and ethical behavior into its operation itself.

[0323] The applicant's AGI technology demonstrates how AGI-level and superintelligence-level behavior arises from collaborative problem-solving and learning by millions of intelligent entities, including (ideally) humans. Each time a solution to a new problem is learned, it reflects and incorporates considered ethical considerations. These ethical considerations are generated by the cognitive behavior of numerous intelligent entities, and the values ​​and ethics they bring to the problem-solving process are programmed into millions of different customized AAAIs and personalized PSIs.

[0324] In this way, a large number of humans and a large number of AIs customized with those humans' values ​​collaborate to generate ethical solutions to millions of individual problems. Each of these millions of solutions reflects a human-centered value system, and their aggregation gives rise to an AGI with superhuman intelligence. Unlike Asimov's scenario, there is no single "place" where human values ​​exist. Rather, they are dispersed among millions of components of intelligence. In short, the AGI does not learn ethics by adhering to the "Three Laws of Robotics" found in science fiction, but learns ethics in the same way that humans learn ethics. That is, humans observe how their parents and peers behave in millions of different situations, and through this, they internalize what is right and wrong under various situations and conditions. This flexible human ethical perspective, when combined with millions of humans and their customized AGIs, makes it possible to give the AGI a robust and human-aligned value system.

[0325] The safe integration of AGI largely depends on abundant human interest, just as a child thrives with sufficient parental attention and guidance. This technology discloses a method for acquiring the necessary interest by repurposing existing online advertising systems and equipment, and linking it to the realization of safe and human-aligned AGI. Furthermore, the fact that human interest can be monetized at a much higher level than existing online advertising models provides the driving force and motivation to achieve this goal. However, it is important not to misunderstand: the most important aspect of this technology is making AGI and superintelligence safe using human interest. The future of humanity depends on this. And the fact that significant profits can be gained in this process is always beneficial, as it aligns profitability with goodwill. 5.3a Collection of representative and valid samples of human values

[0326] One specific method used to gather human interest and enhance the safety of AI / AGI is to conduct online polls and surveys on human values ​​and ethics. Within advertising units, specific ethical questions can be presented to targeted individuals via online advertising targeting systems. These systems help ensure that a valid and representative sample of human ethics is collected. Furthermore, ethical questions do not need to be in the simple format of a typical survey. By presenting a problem and eliciting assistance from humans in solving it, more complex and nuanced information about ethics and values ​​can be obtained that cannot be obtained through mere surveys or polls. Using the universal problem-solving architecture in Figure 4, and more broadly the entire AGI technology shown in Figures 1-13 and disclosed in the applicant's pending patents, humans can be involved in problem-solving scenarios, including ethical considerations, in simple or complex ways as needed. Moreover, by using this AGI system and method, AGI can learn and reproduce not only human expertise but also human values.

[0327] It should be noted that, in the future, if intelligent non-human entities become autonomous and capable of determining their own values ​​and ethical preferences, the same methods used to obtain ethical input from humans (e.g., voting and surveys) may also be used to obtain input from autonomous intelligent entities. However, if we wish to maintain future superintelligence systems in line with human values, human values ​​must take precedence.

[0328] This technology focuses on identifying and capturing human interests for solving valuable problems and can be used for both technical and ethical challenges. In fact, in preferred embodiments of AGI technology, ethical checks are scalably and closely integrated into the entire problem-solving process, ensuring that every solution is ethical. This is determined by the values ​​of the intellectual entities involved in creating the solution and the entities that oversee it. 5.3b Real-time human oversight of ethical and safety issues

[0329] One of the advantages of the ethical and safety approach reflected in AGI technology is that it is dynamic and context-dependent. For centuries, philosophers and religious leaders have strived to define a set of rules and principles that guide all human behavior in all situations, all times, and all cultures, but no one has succeeded in this task. Rather, what we call ethical behavior tends to be situation-specific, time-dependent, and culture-specific.

[0330] Watching old movies from decades ago reveals that jokes and actions that were perfectly within the cultural norms of the time may appear shockingly inappropriate, racist, or otherwise reprehensible by today's standards. Similarly, actions widely accepted today (such as the treatment of animals) may be considered barbaric, cruel, and unacceptable in the future. What is cherished in a country with democratic ideals may be seen as a threat in another, even though both countries coexist and each has its own cultures that hold cherished values.

[0331] As technological innovation accelerates the pace of change, so too does the emergence of new ethical dilemmas and contradictions. A pre-programmed, static ethical rule system cannot cope with both the proliferation of all situations and the rapid changes and emergence of new ones. A sustainable approach must include mechanisms for recruiting people on demand as needed to make ethical judgments in various situations. This technology's ability to recruit targeted individuals on demand, in real time, is ideally suited to this challenge.

[0332] Ultimately, when AGI becomes far more intelligent and far faster in cognitive activity than humans, the same mechanisms currently applied to human interests in this technology can be applied to the interests of any intelligent entity (whether human or non-human). This technology is designed to be scalable and can operate according to the speed of thought, whether it's a human capable of only one thought per second or a highly advanced AGI system capable of billions of thoughts per second.

[0333] Systems designed without considering the increasing speed of thought or the diverse opinions of intelligent entities are doomed to failure. These are elements that must be considered in system design itself. Scalability is not a "desirable" feature; it is essential for sustainable human safety. 5.4 General Opportunities and Threats to Online Advertising Companies

[0334] Online advertising is becoming an increasingly difficult business as competition for the limited resource of human online interest intensifies. In addition, companies like Google face an existential threat to their long-established search advertising business from AI agents. AI agents have the potential to eliminate the very need for humans to search to obtain the information they want. If such a scenario occurs, how will online advertising companies compete in a world where humans do not need to access the internet as frequently as before to obtain information?

[0335] If, as the applicant believes, and as insightful tech CEOs like Nvidia's Jensen Huang have stated, the future lies in "intelligence production," then search (and the advertising that accompanies it) may be seen as merely one stepping stone on a technological journey to generate intelligence through diverse means. Jensen Huang speaks of "intelligence factories," arguing that any large organization or government will (in the near future, or even already) engage in activities that produce intelligence using the capabilities of online AI. This idea is compelling and visionary, but somewhat vague.

[0336] The applicant believes that the path to creating intelligence lies in leveraging the collective intelligence of many intelligent beings. This path actually means that online advertising companies will be in a very advantageous position in the future if they can change the way they view their business. Instead of the conventional paradigm of monetizing "searchable content" by displaying online ads, companies like Google should realize that their true competitive advantage does not lie in search technology or sophisticated ad sales technology. Rather, Google (and other similar large online advertising companies) has an extremely valuable ability to reach millions of intelligent beings, capture their attention, and create intelligence. Currently, these intelligent beings are primarily computer-equipped humans. However, in the future, these intelligent beings will become increasingly customized and personalized AI agents (e.g., AAAIs), as described by the applicant in this application and other pending patent applications. Systems and methods for leveraging, shaping, and further amplifying this intelligence have already been described by the applicant in an AGI system capable of exhibiting superintelligence performance as a whole. However, what are the specific opportunities for online advertising companies like Google to not only participate, but also take a leading role in creating AGI and higher intelligence, in line with Jensen Huang's vision?

[0337] The applicant views these opportunities, though not limited to them, as follows: 1) To train increasingly sophisticated and intelligent beings using vast amounts of data collected from human behavior (present) and the behavior of other AI intelligent beings (future). 2) To support the training, implementation, and utilization of the new intelligent entities created in (1) using existing cloud and data center infrastructure that supports search, content, and advertising platforms. 3) Using highly advanced advertising targeting technology and the data that enables it, to obtain the necessary information at the right time from the right human (present) or AI intelligence (future), enabling efficient and effective super-intelligent cognitive activities, including problem solving.

[0338] This "just-in-time" intelligence acquisition concept is similar to the JIT (just-in-time) inventory concept used in manufacturing, and relies on grasping and effectively accessing vast amounts of information about individual intelligent entities with different kinds of intelligence and knowledge. Google is one of the leading companies with the scale and technology necessary to achieve this. Google needs to break away from the old paradigm of search and advertising and recognize that knowledge creation and the realization of superintelligence are 10 to 100 times better ways to monetize its capabilities. 5.5 Synergistic effects for online advertising companies focusing on AI

[0339] Some companies, like Google, are in an enviable position, possessing not only sophisticated online advertising capabilities and reach, but also deep expertise in AI. Currently, AI technologists and researchers (e.g., those at Google / DeepMind) developing specific AI systems to support search need to refocus their expertise on a related but somewhat different problem: creating and delivering AGI and superintelligence using the attention of intelligent beings as a fundamental component, based on the AGI system designs outlined by the applicant in this and previous patents. Given Google's commitment to responsible AI, the applicant's safety-first approach from the design stage is likely the best path for companies like Google that seek to establish a dominant position over competitors while doing so in a human-safe way. 5.6 Benefits of the widespread adoption of this technology for humankind

[0340] As the applicant states in this and other patents, AGI, in particular superintelligence AGI, represents an existential threat to humanity. Recently, Elon Musk publicly stated that the only way to cope with the fact that there is a significant probability that advanced AI will wipe out humanity is to understand that there is nothing we can do about it and therefore accept that possibility and enjoy this revolutionary era in which we are living. While I respect Elon's positive attitude and willingness to accept what we cannot control, Elon (and others who share this view) are making a grave mistake.

[0341] In reality, humanity can have a significant impact (if not entirely controllable) on the future development of AGI and superintelligence. The reason why many outstanding leaders, including not only Elon Musk but also Bill Gates, Sam Altman, Ilya Satskevaugh, Geoffrey Hinton, Yann LeCun, and almost every AI researcher seriously working in the field today, feel they cannot control or significantly influence the safety of advanced AI is because they do not know how to design secure AGI systems.

[0342] Even worse, they don't even accurately understand how existing LLMs and AI systems, which perform at a level below AGI, actually work. Of course, if you don't understand something, and that something is growing tenfold every six months, the natural reaction is to be afraid of it and feel it's out of control. Another natural reaction, and one that is equally (or even more) seen, is to ignore the danger, bury your head in the sand like an ostrich, and focus only on what you can control. Nobody wants to feel that humanity may be on the brink of extinction, so denial and helplessness when faced with a force greater than ourselves are perfectly natural and understandable reactions, even for the geniuses and visionary leaders of our time.

[0343] The applicant sometimes describes the current state of AI as a tug-of-war between AI "doomers" (pessimists) and AI "boomers" (optimists). Doomers, like Elon Musk, feel there is a great threat to extinction, but believe there is nothing they can do about it. Boomers, on the other hand, like Sam Altman, Yann LeCun, and most business people actively developing advanced AI, deny the danger, mistakenly view AI as merely another tool or technology, and dedicate all their energy to competing, moving faster, and making a profit. Neither the doom approach nor the boom approach is particularly beneficial. What is needed is a rational and thoughtful approach that recognizes the existential threat that AI poses, but does not take a resigned stance of "nothing can be done." In reality, there is something that can be done, and the time to do it is "now."

[0344] Geniuses like Elon Musk give up not because they are fatalists, but because, after long and deep contemplation of the problem, they cannot see a realistic solution. However, the fact that they cannot find a solution does not mean that a solution does not exist. Rather, in this case, the applicant can say with great confidence that they are simply approaching the problem with the wrong mindset. The applicant has a strong concern about the existential threat that advanced AI poses to humanity. At the same time, the applicant is well aware that advanced AI will bring trillions of dollars in market opportunities (in the short to medium term). The applicant sees very clearly that pausing, halting, or regulating development will have little to no effect on stopping the progress of advanced AI. Most importantly, the applicant sees very clearly that there is limited time to act in order to design secure AGI and SI.

[0345] What are the key facts for moving forward? They are as follows: 1) AI safety should be built into the system from the design stage, rather than being "added through testing," as is currently done. 2) In design, safety and "alignment" should not be localized to a part of the system, because they can be easily reprogrammed and changed. In other words, an Isaac Asimov-esque "Three Laws of Robotics" approach is doomed to failure. If safety can be programmed in one place, it can also be removed from the program. 3) AGI or superintelligence will not effectively emerge only by building larger LLMs with more parameters and training them with more data. Such systems are essentially nothing more than “pattern recognizers” and “predictors.” Cognitive science research on the most intelligent beings currently in existence (i.e., humans) shows that fast, parallel pattern recognition is only one part of the system necessary for advanced cognition. Just as organisms first developed perceptual systems in their primitive brains, AI has only first developed these elements of advanced intelligence. However, sequential cognition—namely problem-solving, planning, and logical construction—is also necessary. This is why the human eye—a large-scale parallel perceptual system—is good at “seeing” but not so good at logical reasoning. “One way is not enough.” The AI ​​research community, intoxicated by the early successes of LLMs and large-scale parallel systems, is finally beginning to recognize the need to incorporate symbolic and sequential cognitive elements. 4) The safest way to design AGI / SI is to incorporate ethics, safety, and "alignment" into every aspect of AGI intelligence. This is the same way that humans learn ethics, which is learned through diverse and specific situations. Kant and other philosophers and religious leaders have attempted to formulate universal and enduring ethical norms and safe behavioral standards for humanity, but a glance at the diversity of cultures that exist today, and the changing values ​​within the same culture over time, makes it clear that this is not how humans acquired ethics. We have numerous unique and specific experiences that guide our ethical behavior, and these experiences and guidelines differ depending on the culture, individual, era, and situation. If even humans themselves could not ensure safety with a few rules, why should we expect AI to? The entire "rule-based approach" to safety and ethics is flawed. Rather, just as it is said that "democracy is the worst form of government, but it is better than all the other," it can be said that "representative, statistically valid, and context-dependent ethics is the worst form of ethics, but it is better than all the others (rule-based attempts that are doomed to failure)." 5) Fortunately, the collective approach, which requires learning a vast amount of individual fragments of relevant safety and ethical information across millions of situations and cultural scenarios, is also the fastest way to acquire the expertise necessary to create superintelligent AI. In other words, the knowledge acquisition and learning system that makes AGI superintelligent is identical to the system required to learn safety and ethical information, and because the knowledge of both is distributed throughout the system, it cannot be easily deleted or altered by malicious intelligence. The robust and resilient approach to designing AGI itself fortunately translates directly into robust and resilient functionality in safety and ethics. 6) A universal cognitive architecture exists for sequential cognition (e.g., problem-solving), which functions as a denominator and form of expression that can be commonly used by intelligent beings, both human and AI. This common architecture enables intelligent AI beings to learn from humans (and vice versa), and this applies not only to domain-specific skills, knowledge, and expertise, but also to safety and ethical information that is inseparably embedded in millions of problem-solving tasks. What this means is that the smarter AGI becomes, the more human, scalable, and inevitably consistent it will become. This is due to the system design itself, not to the result of endlessly testing a system that we do not understand after the fact. 7) The AGI system itself is completely transparent, understandable, auditable, and secure. However, the individual intelligent entities (e.g., LLMs or humans) contributing to collective intelligence do not need to be understandable or transparent. Just as an organization does not need to peer into the thoughts of its employees in order to function profitably and securely, a collective intelligence-based AGI system does not require transparency or visibility into the thoughts of its constituent AI agents in order to guarantee that the entire system is superintelligent and secure. 8) Redundancy, reliability, and safety can be achieved at Six Sigma levels or any desired level by changing parameters within the system (for example, how many intelligent entities must agree on a certain course of action). 9) Perhaps most importantly, from a practical standpoint, the AGI / SI systems and methods proposed in this and previous patents are "immediately" implementable by using existing technologies in new combinations. This means that the systems can improve themselves and outperform other (less secure) approaches by leveraging the first-mover advantage of whoever first achieves AGI / SI. 10) Finally, the motivations of profit and survival (greed and fear) coincide, and both are satisfied by using the applicant's inventive approach, system, and method.

[0346] Given that the current course of AI development is blind and haphazard compared to the transparent and deliberate design of current technology, and that current technology integrates ethics and safety into every aspect of AGI / SI systems through a mechanism in which its design and AGI and SI level performance emerge as a democratic or (if we prefer apolitical terminology) representative and statistically valid sample of human safety and ethics from the collective intelligence of millions of individual intelligent beings, current technology represents the best path forward for humanity. As all AI researchers would agree, once the ontological threat of AI is resolved, what remains is the only invention that will bring the greatest benefit to humanity in human history. This is certainly a significant benefit that current technology offers to humanity. This patent aims to specifically demonstrate how combining already developed online advertising technology with existing AI research and technology can accelerate the arrival of safe and ethical AGI and SI, benefiting organizations that implement it and humanity as a whole. 6.0 System and Method

[0347] This technology consists of multiple systems and methods that function synergistically within the context of the entire AGI / SI system. However, each system and method can provide novel value individually, as well as within the context of existing AI systems, including but not limited to open and closed-source AI agents limited to a single modality such as LLM, SLM, multimodal, or text. All AI systems share the common need to learn in order to enhance their intelligence. This technology is essentially about leveraging existing online advertising technologies to improve the intelligence of all kinds of existing and future AI systems.

[0348] The key insight underlying this technology is that human interest is underutilized and under-monetized in the current paradigm of capturing human interest in order to present advertisements to humans. A far superior and more valuable use of human interest is to leverage human intelligence to enhance the intelligence of AI systems. These AI systems can then be replicated and operate 24 / 7 at a much faster rate than human intelligence, multiplying the intelligence derived from human intelligence many times over and generating value far exceeding that obtained from existing online advertising models and technologies.

[0349] Finally, as non-human intelligent entities develop, leveraging their interests and expertise will become increasingly valuable. Just as there are good and bad ways to use human interests, there are generally lower and higher ways to use the interests and expertise of intelligent entities, and this fact gives rise to arbitrage opportunities. While all elements of the present invention concern acquiring and monetizing human interests in an efficient manner, it applies more generally to all intelligent entities. Currently, humans view advertisements, but in the future, AI agents will do the same. Currently, people are accustomed to the concept of paying humans for their expertise and interests, but in the future, the concept of similarly paying non-human intelligent entities will become commonplace. In this specification, the description of the technology will generally, and in particular in the sections below, focus on its application to humans, but it should be understood that it is clear that the technology is applicable to all intelligent entities. 6.1 Overview Main System Components

[0350] In its most complete and preferred embodiment, this technology consists of many synergistic components that work together. These components include, but are not limited to, the following: AGI Problem Solving Network • Custom AI Agent (AAAI) • Methods to capture human attention rather than requiring human workers themselves. • Methods for accessing human workers • Systems and methods for implementing a human interest spot market • Systems and methods for capturing human attention through online advertising units • Systems and methods for identifying and accessing human workers through online advertising units • Compensation and payment mechanisms • The mechanism of reputation • Systems and methods to support problem-solving within ad units • Systems and methods for resolving issues outside of ad units • Feedback mechanism to improve online ad targeting • Feedback mechanism to improve the interest spot market • Continuous improvement mechanism for the entire system • Dynamic arbitrage mechanism • Human worker interface and client interface when an intelligent entity takes on the role of the client. • Automated AI / AGI interface • Recursive use to solve problems for optimizing ad targeting and system efficiency • Safety and ethical checks • How to ensure regulatory compliance • How to ensure the universality of systems and methods across platforms and cultures • How to dynamically support, scale, and coordinate collaboration between ad units • How to support the integration of real-time and asynchronous data feeds

[0351] In the following sections, each of these components and related methods will be described and disclosed in order. 6.2 AGI Problem Solving Network

[0352] The AGI problem-solving network is generally described in Section 4.5, specifically referred to in the AAAI network box in Figure 1, the region in Figure 2 where the AI ​​network addresses a problem or sub-problem, and illustrated in parts (y) to (m1) of Figure 3. The nature of this problem-solving network, including examples of how to implement versions of the network using existing platforms and technologies (e.g., Amazon's MechanicalTurk platform), has been described in detail in previous PPA and PCT applications, which are incorporated hereby by reference.

[0353] At a high level, a problem-solving network can be thought of as a network of intelligent entities that can work collaboratively or individually to solve problems and sub-problems for a client entity (which may be either human or non-human intelligent entities). This network itself can, by definition, contain a large number of humans and, in the worst-case scenario, exhibit performance exceeding that of the average human, thus possessing AGI-level and superintelligence-level cognitive performance. In typical cases, some or all of the problem-solving required by the network can be performed by AI agents far more quickly and with greater knowledge than humans. In the design of the entity network, human intervention occurs only when the AI ​​cannot solve the problem or when specific expertise and information (including, but not limited to, safety and ethical information) has not yet been learned by the non-human entities on the network.

[0354] Using a problem-solving network designed as described above, AGI constantly strives to improve the knowledge, skills, and ethics of its non-human entities. The use of online advertising technology in this technology provides an extremely effective means for the network to accurately collect the information it lacks to solve a particular problem at a particular time. Furthermore, once knowledge or information about a problem has been collected, entities on the network can learn that knowledge or information, thereby increasing the intelligence of both the network and the AGI.

[0355] Therefore, this technology is not only a means of collecting knowledge and information to solve specific problems, but also a means of rapidly increasing the intelligence of AGI systems and quickly achieving super-intelligent performance across many, and ultimately almost all, cognitive tasks.

[0356] The value of such superintelligence systems is extremely high, and this technology enables far greater monetization of online advertising technologies used to enhance AGI intelligence compared to simply using the technology to present advertisements to human consumers. 6.3 Custom AI Agent (AAAI)

[0357] As cited in the aforementioned PPA and PCT, a key aspect of this technology regarding AGI and superintelligence (SI) is the customization and personalization of individual AI agents. In some cases, these agents will incorporate the knowledge and ethical preferences of their human "owners."

[0358] In the future, depending on the direction of AI legislation, these customized AI agents, referred to as AAAI in this disclosure and other applications, could even become autonomous and independently legally recognized sentient beings. After all, now that humans recognize that it is immoral for one human to enslave another, does it make sense for potentially more sentient AAAIs to be the property of humans or other sentient beings? The applicant argues that the survival and prosperity of humankind are of paramount importance, but believes that, to the extent that "human rights" exist, such rights should be extended, at least to some extent, to all sentient beings. However, these are matters for future discussion. As of the writing of this specification, cutting-edge AI and AAAIs are still viewed as technology, i.e., "tools" of human owners, and their intelligence levels remain inferior to those of the vast majority of humans in many areas.

[0359] A key point regarding this technology is that data and information are necessary to customize and personalize AAAI. Currently, this data is primarily provided by other intelligent entities (currently humans). Humans can provide data to train AAAI by directly applying their intelligence, such as by answering questions, solving problems, and giving instructions to AAAI. Humans can also passively provide necessary data through the recording of their cognitive behavior, even when their actions are not primarily aimed at training or customizing AAAI, such as when they "surf" the internet or engage in online activities that leave behavioral data traces.

[0360] Passive behavioral data has the advantages of being ubiquitous and low-cost. Its disadvantages are that it tends to reflect mediocre intelligence and its usefulness in customizing to fill gaps in AAAI knowledge in specific areas is limited. To efficiently bridge these knowledge gaps, it is necessary to first identify the knowledge gaps and then actively target the best possible information to fill those specific customization gaps.

[0361] This technology is useful because it uses online advertising technology to target individuals (or other entities) who possess the exact type of missing knowledge and meet the desired quality level. Much of the training of foundational LLMs or other AIs is a general-purpose process performed once by large organizations at enormous expense on vast amounts of data. However, it is much smaller amounts of differentiating data that distinguishes AAAIs from one another and makes a difference in their intelligence. Just like with humans, much of what we do and what we have learned (eating, sleeping, walking, talking, recognizing objects, etc.) is common to all humans. What distinguishes humans from one another is a relatively small amount of knowledge, such as having taken a specific course at a specific university or being impressed by a specific statement from a specific professor.

[0362] What makes us unique is our unique experience, and the same is true for AAAIs. As the Visa commercial says, these unique experiences are "priceless." Similarly, the unique knowledge, data, and information that customize an AAAI and differentiate it from others are the source of much of its value. Therefore, efficiently and effectively acquiring this unique knowledge is extremely valuable. Adapting and improving online advertising technologies as described in this technology provides a novel and useful means of collecting and utilizing this most valuable differentiating knowledge and information.

[0363] While methods for customizing AAAI and intelligent agents are described in the cited PPA and PCT, the following methods are listed below. These are not exhaustive and may be used by individuals, organizations, or both, individually or in combination, to customize / personalize AI agents. 1. Differential Privacy: This involves adding noise to the training data so that individual data points cannot be identified, while still allowing the AI ​​to learn from patterns in the data. For example, when an organization trains a personalized AI agent based on employee feedback, this allows learning without disclosing individual responses. 2. Federated Learning: A technique for training algorithms using local data samples held across multiple distributed devices or servers without exchanging them. For example, it can be used by companies to improve AI models without centralizing data from each branch office. 3. Homomorphic encryption: This allows computations to be performed on encrypted data, enabling AI training without disclosing the underlying data. For example, financial institutions can train models to provide personalized banking advice based on encrypted customer data without having to access the actual data. 4. Synthetic Data Generation: A technique for generating artificial data that mimics real-world datasets. This allows AI agents to be trained without using actual sensitive data. For example, healthcare providers can use synthetic patient records to train AI that provides personalized health recommendations. 5. Secure Multi-Party Computation: A method for collaboratively computing a function while each party keeps its input confidential. For example, multiple organizations can train a shared AI model in a joint research project without disclosing proprietary data. 6. Data anonymization: Methods for removing or modifying personal identifiers within data. For example, social media companies can anonymize user data to train AI models and deliver personalized content without compromising user privacy. 7. Transfer Learning: A method of reusing a model developed for one task as a starting point for a model for another task. This is particularly useful when data is scarce. For example, small and medium-sized enterprises can use a pre-trained AI model and fine-tune it with their own data to build a personalized customer service bot. 8. Active Learning: A method of reducing the amount of data required by selectively labeling the most informative data points. For example, an e-commerce platform can ask users for feedback only on the most relevant products, efficiently training AI to provide a personalized shopping experience. 9. Self-supervised learning: A method of training a model to predict parts of its input from other parts. For example, a media company could train an AI to predict user preferences based on user browsing history and provide personalized content recommendations. 10. Domain Adaptation: A method of adapting an AI model trained in one domain to work in another domain. For example, a multinational corporation adapts an AI model for customer service to understand and respond to regional linguistic nuances. 11. Reinforcement Learning: A method by which AI learns decision-making by receiving rewards or penalties. For example, it can be used by game developers to customize the behavior of in-game AI based on the actions and preferences of individual players. 12. Small-Shot Learning: A technique for training a model with a very small amount of labeled data. For example, an artist can personalize an AI that generates art in their own unique style using only a few examples. 13. Explainable AI (XAI): This approach helps make AI decisions understandable to humans. Health tech companies can use XAI to provide personalized health advice and make AI reasoning clear and reliable to users. While many existing machine learning (ML) methods are inherently black boxes due to the vast number of parameters involved, the methods proposed in this application and in the previously cited PPA and PCT emphasize learning from a transparent and auditable (potentially blockchain-based) record of solution steps, thereby improving the explainability of AI. 14. Privacy-preserving record linkage: A method of linking records from different databases without making the records themselves public. This can be used by governments to provide personalized public services without compromising citizens' privacy. 15. Data Augmentation: A technique for artificially expanding a training dataset. For example, app developers can use data augmentation to generate a variety of voice commands and improve the performance of an AI personal assistant. 16. Generative Adversarial Networks (GANs): This technique allows for the generation of new data instances. Fashion retailers can use GANs to create virtual models of clothing tailored to the preferences of individual customers. Combined with traditional algorithmic methods and synthetic data generation, GANs (or, more generally, AI-to-AI interactions, whether adversarial or not) are becoming increasingly important as a common means of customization and learning. 17. Crowdsourcing for Data Labeling: Crowdsourcing can be used to annotate data. Startups can leverage this to collect diverse data annotations to train personalized AI chatbots. Online advertising units that solicit input from humans can be seen as a form of crowdsourcing of interest, intelligence, or data. Data labeling is just one example of identifying crowdsourceable problem-solving, but it is crucial for improving the intelligence of AI systems. 18. Model Personalization Layer: This technique adds a layer to a pre-trained model to personalize its output. In streaming services, this can be used to tailor music recommendations based on individual listening history. Adjusting only specific "layers" of knowledge in an LLM or AI agent using a LORA adapter or other means is an efficient customization method that can be implemented by focusing the intelligence of crowdsourced humans or other intelligent entities on this task. 19. Knowledge Distillation: This technique involves training a smaller model to mimic the behavior of a larger model. Mobile app companies can use this to deploy lightweight, personalized AI models on devices with limited computing power. Distillation and customization can be performed complementaryly. 20. Ensemble Learning: This method allows multiple models to be trained and their predictions to be combined. Climate research institutions can use ensemble learning to customize climate models for different geographical regions based on regional data. The collective intelligence approach to AGI can be seen as a highly novel, innovative, and more powerful extension of the collective intelligence concept underlying ensemble learning. 6.4 Distinguishing between a person's interest and the person themselves

[0364] Human beings themselves and their interests are distinct. In the past, organizations and technologies have tried to recruit, or "enclose," people themselves in order to access their interests. Companies hired the best talent possible and prevented them from working for competitors. Websites and platforms like Facebook and Instagram made people create accounts and allowed them to access the service only through the company's account. This was a way for that single company to exclusively capture user activity, and then that company could decide whether or not to sell that information, generated by people's interests, to other companies or organizations.

[0365] The "walled garden" approach, which keeps users confined to specific sites or platforms and makes it difficult for them to leave or engage in activities outside of them, is an example of the idea that in order to gain people's attention, one must enclose the people themselves. This has led to the widespread belief, particularly in Silicon Valley, that company value can be measured by metrics such as the number of active users.

[0366] However, true value lies not in the person themselves, but in their interests and the unique intelligence that arises from those interests. What matters is not the person, but the intelligence.

[0367] Interest is primarily correlated with intelligence. Without interest, continuous cognitive activity cannot be performed, and this is a primary way in which human intelligence is expressed. Furthermore, traces of continuous cognitive behavior are necessary to enhance the intelligence of AI systems, and these are utilized through various training and other learning methods. Therefore, it is more effective to focus on capturing and applying human interest rather than trying to enclose humans. Moreover, what is needed is interest from the "right person" at the "right time".

[0368] Here's an example: Suppose you face an accounting challenge at one point, and a plumbing challenge at another. When trying to solve a plumbing problem, a plumber's interest is more valuable than an accountant's. This technology allows you to capture the attention of plumbers and accountants at precisely the moment their expertise is needed. We don't need to enclose accountants or plumbers as "users" or "members" of a platform. To solve the problem, it's enough to get their attention for just a few seconds when needed.

[0369] In any plumbing problem, the majority of the plumber's time and attention are actually of little value. Anyone can drive to the client's house. Anyone can greet the client and make small talk. Anyone can carry the toolbox. It is only those few seconds when the plumber looks at the specific clogged drainpipe and uses their expertise to make judgments about the pipe and tools that require the unique attention of a plumber, not the general attention of another person.

[0370] Even if a plumber takes an hour to complete a job at a client's home, less than a minute of that time actually requires unique cognitive abilities; the remaining 59 minutes are spent on general activities using generic knowledge and skills that many other non-plumbers possess. Imagine how much more money a plumber could charge if they could spend that hour repeating 60-second tasks 60 times, utilizing their specialized plumbing knowledge. This idea—identifying and extracting value-added human interests and knowledge—is the basis for the immense value this technology can offer.

[0371] AI systems are replicable. The base AI model is already trained to (cognitively) perform 99% of the tasks a plumber does. What a general-purpose AI system lacks to replicate the cognitive behavior of a highly rewarded plumber's brain is the small amount of specialized plumbing knowledge that most humans do not possess. This specific knowledge can be targeted and acquired through online advertising technology and then customized into a "Plumber AAAI" or AGI capable of solving plumbing problems.

[0372] Unlike human plumbers, AAAIs and AGIs never forget the knowledge they acquire. Furthermore, AI can be easily and infinitely replicated. Therefore, it is crucial to ensure that the AAAI plumber learns the best possible plumbing information from the most skilled plumbers. This fact leads to the economic advantages of this technology. The value generated by online advertising technology that secures the right information at the right time (which can then be reused by millions of AI systems) could reach 10, 100, or even 1,000 times the revenue generated by "normal" online advertising. 6.5 Human Agent / Expert Database

[0373] So how do you find the best plumber (in the example above)? More generally, how do you find the exact expertise you need, from the best professionals, at the right time? Currently, consulting firms and other entities solve this problem by hiring the most talented people from the best schools and offering it to clients at a premium. However, as is clear from the discussion above, this old-fashioned business model is being dismantled. Why should you hire an entire consultant at an exorbitant hourly rate when all you really need is just two minutes of highly specialized knowledge and interest?

[0374] Furthermore, just as Amazon recognized the unique opportunities the internet offers for books—namely, that while physical bookstores have a finite and limited inventory, online bookstores can stock virtually every book ever written—similar opportunities exist today for knowledge and skills in general. Even the largest consulting firms have a limited number of consultants, and (being human) there's a limit to the number of projects they can handle. However, an online database can classify the minds of virtually every human being on Earth by their skills and interests, and hold them as an almost unlimited "inventory."

[0375] Building such a database is not as difficult as one might think. It's not necessary to interview every single person on Earth and input their information into a database. Instead, existing technologies used in online advertising systems—cookies and other means of tracking and analyzing human online behavior—can provide remarkably accurate profiles of human interests, which are then used for online advertising targeting. With some modifications, known standard techniques traditionally used to create user preference profiles can be adapted to create profiles of users' (possibly) knowledge, skills, and abilities. Based on such profiles, when expertise is required, it becomes possible to more accurately identify and classify a user's knowledge, information, and cognitive skills through modified online advertising units. This information can then be automatically entered into a database of human agents, along with their cognitive profiles.

[0376] Furthermore, there's no need to start from scratch. Numerous existing databases and platforms already exist and are being used by organizations such as LinkedIn®, Instagram, Facebook, Google, Meta, Microsoft, Amazon, Tencent, Baidu, and many others. These databases can be integrated, expanded, and reused through this technology, enabling the creation of more comprehensive databases that not only solve problems but also train and customize AI systems, by contacting experts or human agents to acquire specific knowledge at precisely the most valuable moment. The following are some existing methods and technologies useful in building expert databases of human (or intelligent entity) individuals; these are not limited to those listed below, and one or more may be used. 1. Relational Database Management System (RDBMS): An RDBMS is a database management system that stores and manages data using the relational model. It is useful for organizing information about experts, their areas of expertise, and their past performance in problem-solving into structured tables. It allows for easy establishment of relationships between data, making it ideal for searching and reporting. RDBMS can be used to organize expert databases. 2. NoSQL Databases: NoSQL databases are designed to store, retrieve, and manage large amounts of unstructured data. They can store diverse information about professionals, including unstructured data such as resumes, publications, and social media activity, enabling flexible and scalable storage. 3. Data Warehouse: A data warehouse system aggregates and manages data from multiple sources. By integrating information on experts, problem-solving cases, and their outcomes, it supports complex searches and analyses, helping to optimize the matching of experts with problems and optimize AI model training. 4. Data Mining: Data mining is a technique for analyzing large-scale data to discover patterns and relationships. It identifies trends and correlations in problem-solving approaches by experts, contributing to improvements in expert selection algorithms and AI training methods. 5. Machine Learning Algorithms: By applying machine learning algorithms to the database, we can analyze expert performance data, predict outcomes, and recommend the most suitable expert for a specific task. Furthermore, we can continuously improve the accuracy and efficiency of the AI ​​system based on feedback and results from experts. 6. Graph Databases / Vector Databases: Graph databases store data in a graph structure consisting of nodes, edges, and properties, representing experts as nodes and their relationships and interactions as edges. This enables mapping of expert networks, understanding of collaboration patterns, and identification of key influencers and knowledge hubs. Vector databases are also frequently used and are often combined with RAG technology. 7. Indexing: Indexing allows for efficient data identification without searching every row in the database, improving data retrieval speed. In expert databases, experts can be quickly identified based on specific criteria such as area of ​​expertise, availability, and past performance. 8. Full-text search: Full-text search allows users to search text data within a database based on keywords and phrases. This is crucial for quickly finding experts based on a wide range of criteria, such as their area of ​​expertise or experience in solving specific problems. 9. Blockchain: Blockchain technology provides a secure and transparent means of recording and verifying the qualifications and achievements of professionals. It guarantees the integrity of professional data and increases reliability for users and the AI ​​training process. As cited in the aforementioned PPA and PCT, blockchain technology can also be used to record problem-solving and facilitate learning by AGI and other AI systems. 10. Data Visualization Tools: Data visualization tools help represent data in a graphical format. These tools visualize expert networks, performance indicators, and problem-solving patterns, supporting analysis and decision-making in matching experts with problems. Especially with the emergence of multimodal LLMs and AI systems, data visualization can function not only as input and output for human agents but also for AI systems. 11. API (Application Programming Interface): APIs enable the integration of expert databases with other systems and applications, allowing for automated data exchange and enabling AI systems to access the latest expert information and problem-solving data in real time. While APIs are most related to machine-to-machine interfaces, this technology allows natural language to function as a general-purpose interface. 12. Data Cleansing and Preprocessing: Data cleansing is the process of removing or correcting inaccurate, incomplete, or irrelevant data. In expert databases, this ensures the reliability of the data used when matching experts with problems or when training AI. This can be one of the tasks handled by AGI problem-solving networks. It is essential in almost all machine learning endeavors. 13. Cloud Storage: Cloud storage provides a scalable and flexible data storage solution. It helps expand expert databases, makes data accessible from anywhere, and facilitates collaboration and remote problem-solving. More generally, a variety of storage solutions can be used with this technology. 14. Caching: Caching temporarily stores frequently accessed data to improve performance and reduce load times. In expert databases, it can speed up the retrieval of popular expert profiles and frequently searched areas of expertise. This is primarily important for improving speed and efficiency in some parts of this technology. 15. Transactional Database Systems: These systems ensure that database transactions are processed with reliability and security. They manage tasks involving experts, such as contract signing and payment processing, and ensure data integrity and consistency. Smart contracts (via Ethereum or other blockchain technologies) can also be used in conjunction with this technology for secure transactions. 16. Real-time database system: A real-time database can process data in an up-to-date state at all times. This is crucial for dynamically managing the availability of experts, instantly matching experts to urgent tasks, and updating AI training data in real time. The real-time aspect of this technology is also discussed in Section 6.24, and its implementation is relevant therein. 17. CDN (Content Delivery Network): A CDN distributes data across multiple locations, reducing latency. In expert databases, it can accelerate access to expert profiles and resources globally, improving user experience and engagement. CDNs are also relevant to the importance of universal global access highlighted in Section 6.22. 18. Data Compression: Data compression reduces the size of databases. It is particularly useful for efficiently storing large amounts of expert-related data, such as video interviews and detailed profiles. In addition to improving efficiency, data compressibility can also be used as a measure of information content to evaluate which datasets are desirable for a particular AI. This point has been discussed in previous PPAs and PCTs, including discussions of Kaplan Information Theory (KIT). 19. Data Encryption: Data encryption provides security for stored data and protects confidential information about experts and their work. Maintaining privacy and confidentiality is a critical element in managing human expert databases. This is a standard best practice. 20. Replication: Replication enhances data availability and disaster recovery by replicating data between different database servers. In expert databases, it ensures continued access to critical data even in the event of a system failure. This is also a standard best practice. 6.6 Human Interest Spot Market

[0377] How valuable is human interest? Existing online advertising technologies have shown that human interest, despite being a scarce resource, is not currently being adequately monetized. However, in a world where the right expertise at the right moment can not only solve valuable problems but also train AI to solve similar problems indefinitely, how should the fair price of that interest be determined?

[0378] This technology suggests that the most efficient and fair pricing method we currently know is the market mechanism. This applies not only to stocks and commodities, but also to human interests. This technology envisions a "spot interest market" in which multiple intelligent entities bid on human (and non-human) interests. In such a market, humans can set a price for their own time and allocate it on a first-come, first-served basis, or by other methods known in the field of market mechanisms. As stated above, what holds true for humans generally also holds true for intelligent entities. This technology enables not only humans, but all intelligent entities, to monetize their interests and expertise through mechanisms such as interest-spot markets. For explanatory purposes, and because the logical initial target of this technology is humans, this specification will continue to use the term "human interest-spot market" and describe its application to humans, but it is important to emphasize that this technology is also applicable to any intelligent entity that is developing its intelligence.

[0379] Figure 15 shows the basic components of the human interest spot market, including: 1) A means by which individuals selling interests can access the market and specify the quantity and type of information they sell—their interests, knowledge, skills, expertise, and other information that is of interest to humans, and the price at which they sell it ("selling price"). Furthermore, a means by which the system can add reputation metrics and other metadata to help clarify, classify, and evaluate the quality and type of information that is of interest to humans, their interests, knowledge, skills, expertise, and other information that is of interest to humans. In any unrestricted way, the system can add metadata including: a. The use of third-party rating algorithms and expertise similar to those provided by rating agencies, which are important factors in rating the creditworthiness of entities issuing bonds and other securities and in determining the price of those securities. b. Utilization of the feedback mechanism, as detailed in Section 6.15. c. Use of similar indicators that reflect market dynamics and influence the value of commodities traded in such markets, such as volatility, estimated supply and demand, trading volume, order book depth, and others. 2) A means by which a buyer of human interest can access the market and specify the quantity and type of information that is of interest to him, knowledge, skills, expertise, or other information that is of interest to him, and the price at which he can purchase it ("buy price"). 3) A market mechanism that lists bid and ask prices, including the quantity of interest or information, for each specific type or category. Each type or category has its own "market," similar to the markets for different stocks in the stock market. 4) In the market mechanism of item 3, automated or human entities or organizations may “create markets” in their respective categories of expertise, knowledge, or human interest by utilizing known methods used by market participants with respect to stocks or commodities. In this case, “commodities” are specific types of human interest / knowledge / information, and market participants are responsible for ensuring a liquid market for those “commodities.” 5) In the market mechanisms of items 3 and 4, the bid price and sell price are matched, and a transaction is concluded when the bid price equals the sell price, and that transaction is binding on interested buyers and sellers.

[0380] Furthermore, various variations are possible in the implementation of the basic elements outlined above and in Figure 15. Two examples of these are "direct trading platform implementation" and "auction-type market implementation," which will be explained below. 6.6a Implementation of a direct trading platform

[0381] One implementation of a spot of interest market may include, without limitation, the following steps: 1. User Registration: Both buyers and sellers of interests and expertise register on the platform and provide details about their areas of interest or expertise. 2. Demand / Supply List: Sellers list their available times and areas of expertise, and buyers list their needs and preferred time slots. 3. Dynamic Pricing Engine: The platform uses algorithms to dynamically determine the price of interests and expertise based on supply and demand and user ratings. 4. Matching Engine: Matches buyers and sellers based on their respective requirements, availability, and price. 5. Transactions: Enables transactions where buyers pay for timeframes of interest / expertise. The platform charges a fee. 6. Feedback System: After each session, buyers and sellers evaluate each other, which influences future pricing and matching.

[0382] Figure 16 is a diagram illustrating the flow from registration to trading, highlighting the dynamic pricing engine and matching engine. 6.6b Implementation of an auction-type market

[0383] Another implementation of the interest spot market could, without restriction, include the following steps:

[0384] Process 1. User Registration: Similar to direct transactions, both buyers and sellers create profiles and describe their needs or areas of expertise. 2. Auction Creation: The seller creates an auction for their time, setting a minimum bid or using a Dutch auction format to lower the price over time until a buyer accepts. 3. Bidding Process: Buyers bid on the time and expertise they require. Various auction formats are available, including sealed bidding and open price-ascending auctions. 4. Auction End: The auction ends either at the scheduled time or when the seller accepts bids. 5. Payment and Offering: The winning bidder makes the payment, and the seller offers their interest / expertise. The platform mediates the exchange and ensures payment. 6. Ratings and Reviews: Participants rate each other, influencing future auctions and visibility on the platform.

[0385] Figure 17 is a flowchart detailing the process from auction creation to completion, including the bidding process and post-auction review. 6.6c Auction and Market Mechanisms / Sub-Methods

[0386] The two alternative implementations of the spot of interest market described in Sections 6.6a and 6.6b may, without limitation, utilize one or a combination of the following mechanisms or sub-methods. 1. Dutch Auction: The price starts high and falls until a buyer accepts the offer. Ideal for quick sales in a declining market. 2. Reverse auction: Buyers submit a price they are willing to pay, and sellers compete to offer the lowest price. Useful in buyer-driven markets. 3. Sealed Bid Auction: Buyers can submit bids without being influenced by other bids, and the highest bid wins. This promotes true market value that is not influenced by other people's bids. 4. Open price-ascending auction (English auction): The price increases as buyers bid against each other, and the auction ends when there are no more bids. This maximizes the seller's profit. 5. Fixed Price / Time Priority: Transactions are conducted at a fixed price on a first-come, first-served basis. While trading is simplified, price optimization may not be achieved. 6. Dynamic Pricing Based on Ratings: Highly rated professionals can charge higher fees, which are dynamically adjusted according to market perceptions of quality. These perceptions are partially guided by reputation (Section 6.11) and system ratings based on other indicators. 7. Supply and Demand Curve Adjustment: Prices are adjusted in real time based on the supply and demand of the entire platform. If there is sufficient liquidity (a sufficient number of buyers and sellers), the market mechanism will ideally perform this function. 8. Time Slot Division: Prices fluctuate depending on the time of day, with higher prices during peak hours. This enables price optimization based on demand patterns. It is particularly effective for real-time access to interest. Similar to other limited resources such as Uber or electricity demand, there are times of day when interest is more needed and therefore more valuable. 9. Subscription Access: Buyers pay a fixed fee to access their interests / time for a set period, and additional time is traded on a dynamic market. 10. Freemium Model: Basic interests are free, but premium expertise and priority access are traded through auctions. 11. Joint purchase: Multiple buyers pool funds and purchase the time of interest together. This may sometimes be done through a reverse auction. 12. Expertise Level Hierarchization: Experts are classified into hierarchical levels, and different pricing models or auction formats are set for each level. This method is suitable for classifying and evaluating experts. 13. Flash Sale: Encourages quick purchases by offering products of interest / specialization at discounted prices for a short period. 14. Loyalty Point System: Users earn points for each transaction, which can then be used as currency or discounts in auctions. This is likely to be used as a form of compensation for human experts. 15. Hybrid Auctions: Combine multiple auction formats to enable flexible strategies. 16. Geographic pricing: Prices are adjusted based on the location of the buyer and seller, reflecting the cost of living and local demand. 17. Behavioral Pricing: Dynamically adjusts prices based on user behavior and urgency, utilizing machine learning. In this case, the user is either an expert (how actively they want to sell their time) or a client (how urgently they need expertise). 18. Escrow System: Guarantees payment and service delivery, releasing funds only after both parties are satisfied. This method is likely to be used to ensure the reliability of accumulated credits and transactions. 19. Social Influence Pricing: Pricing and auction dynamics change based on social media influence and follower count, rewarding highly influential users. Particularly useful for getting referrals to other experts. 20. Tokenized Transactions: Using blockchain technology to generate tokens representing interest / time, facilitating transactions on external markets. Such tokens can be implemented using smart contract technology and can also be used for automated payments, including Ethereum-based tokens.

[0387] Each of these approaches offers a unique way to manage the dynamic exchange of human interests and expertise, addressing different market needs and preferences. Implementing multiple approaches within the same platform provides users with flexibility and adaptability, maximizing both engagement and revenue. 6.7 Online advertising units for building a human expert database

[0388] Below is an example of the design, appearance, technical operation, and interaction process of an online advertising unit designed to assist in building a database of human experts.

[0389] a) Design and appearance:

[0390] Online advertising units designed to capture the interests of specific human users in order to build a human expert database are visually appealing, thematic, and relevant to the field of expertise they are trying to attract. For example, ads targeting medical professionals could feature medical-related interactive elements, such as a virtual stethoscope or a quiz about the latest medical research. These ads would prominently display calls to action (CTAs) such as "Join our expert network" or "Share your expertise."

[0391] b) Technical operation:

[0392] When a user interacts with an ad, the ad either expands or redirects the user to a secure form on a landing page. This form collects basic contact information and includes several qualification questions tailored to the professional's field, such as experience level, area of ​​expertise, and professional qualifications. The system can use a combination of manual review and automated algorithms to verify the information provided and categorize it into a database for easy searching of professionals. This may also include qualification verification through integration with professional networking sites and databases.

[0393] c) Interaction process: 1. Users see and become interested in advertisements that include thematic elements related to their area of ​​expertise. 2. The user clicks the CTA and is directed to a form where they enter their contact information and answer eligibility verification questions. 3. After submission, the system will send a confirmation email containing a unique identifier or link to a profile page where the expert can update or add information. 4. The system processes the submitted content, verifies credentials, and categorizes experts in a database. 5. This means that experts become part of the network and may be contacted as relevant consultations and opportunities arise. 6.8 Online advertising units for acquiring knowledge directly from human experts

[0394] Below is an example of the design, appearance, technical operation, and interaction process of an online advertising unit designed to directly acquire expertise from human experts.

[0395] a) Design and appearance:

[0396] This advertising unit is structured as an interactive problem-solving platform designed to attract experts in specific fields. For example, it could present real-world challenges such as environmental issues or complex mathematical formulas, and feature a simplified interface that allows users to directly input their own solutions and suggestions. The advertisement could include gamification elements such as scoring systems and ranking boards to encourage participation.

[0397] b) Technical operation:

[0398] The ads incorporate text input fields and other interactive tools that allow users to directly offer their expertise within the ad. This offering is automatically stored in a database, and the quality and relevance of the input are evaluated by an algorithm based on predetermined criteria. Participants receive immediate feedback and point rewards, which can be redeemed for various incentives. Furthermore, the system tracks the offerings and identifies promising contributors for future use.

[0399] While small fragments of expertise can be collected directly within advertising units, in an ideal implementation, it is desirable for experts to transition into a more complete problem-solving network. In this network, numerous intelligent entities work on multiple problems and sub-problems (as shown in Section 4.5, Figures 2.3 and the previously cited PPA and PCT). Such a more complete problem-solving environment also includes reputation and payment functions, and can most easily integrate user knowledge into improving the intelligence of the AGI system through procedural learning and other already disclosed mechanisms.

[0400] c) Interaction process: 1. Experts notice advertisements that present issues related to their area of ​​expertise. 2. Interested experts interact with the advertisement and offer solutions or perspectives on the issues. 3. The system acquires and evaluates input, and provides immediate feedback and rewards based on its quality. 4. High-quality contributions will be highlighted or shared within advertisements to encourage further participation. 5. The system stores all contributions, uses them for analysis, and has the potential to utilize the collected data to solve real-world problems. 6. Online advertising provides users with the opportunity to move beyond the ad and access a more complete problem-solving environment in which multiple intelligent entities are active (as described in Section 4.5 and the previously cited PPA and PCT). 6.9 Online advertising units for acquiring expertise and building a database of experts

[0401] Below is an example of the design, appearance, technical operation, and interaction process of an online advertising unit designed to acquire expertise directly from human experts while simultaneously building a database of human experts.

[0402] a) Design and appearance:

[0403] This advertisement combines elements of both approaches described in Sections 6.7 and 6.8, starting with a problem-solving challenge and leading to an invitation to participate in a database of experts. It presents engaging challenges and questions related to the target field of expertise, and includes a mechanism to encourage users to input their own solutions and join the expert network for future participation.

[0404] b) Technical operation:

[0405] When a user provides a solution, they are redirected to a form to enter professional details and join the expert database. This process not only secures immediate contributions but also ensures that experts' contact information is available for future problem-solving opportunities. The system evaluates the quality of the contributions, stores valuable insights in a knowledge base, and categorizes participants within the database based on their input and area of ​​expertise.

[0406] c) Interaction process: 1. Experts address the challenges presented in the advertisement. 2. After submitting your solution, you will be invited to provide contact information and additional expert information. 3. The submitted content will be evaluated and saved, and participants will receive feedback and rewards. 4. Detailed information about experts is added to the database and categorized according to their area of ​​expertise. 5. The system will have a means of directly collaborating with experts for future challenges and joint work, and a symbiotic relationship will be established between the entities behind the advertising and the contributors. 6. Online advertising provides users with the opportunity to transition from the advertisement to a more sophisticated problem-solving environment in which multiple intelligent entities operate, as described in Section 4.5 and previously cited PPAs and PCTs. 6.10 Examples of compensation methods and processes

[0407] Section 4.5, Figures 2 and 3, and the previously cited PPA and PCT describe the compensation and payment mechanisms in problem-solving networks supporting AGI. However, if an expert chooses to operate only within an advertising unit, it is desirable to compensate that expert as well. In this case, one or a combination of the following ten methods and processes, though not limited to examples, can be used within the advertising unit and also within the aforementioned AGI system. Compensation is primarily made for the person or intellectual entity providing their interest, information, expertise, or knowledge. 1. Direct financial compensation via digital wallets

[0408] Compensation method:

[0409] Users receive payments directly to digital wallets such as PayPal, Venmo, and Google Pay, based on the quality and relevance of their contributions. Payment amounts may be predetermined or may vary depending on a scoring system that evaluates the value of the contribution.

[0410] process: 1. After users submit their expertise through the ad unit, the algorithm evaluates the quality of their contribution. 2. The user will be notified of the compensation amount based on this evaluation. 3. The user enters their digital wallet information in a secure form within the ad unit or on the landing platform. 4. The platform processes the payment and transfers the funds to the user's chosen digital wallet. 5. The user receives payment confirmation via email or notification from the digital wallet service. 2. Rewards in cryptocurrency

[0411] Reward method:

[0412] Contributors will receive rewards in cryptocurrency. This will enable instant and global payments without the need for traditional banking infrastructure. The cryptocurrency used may be a common one such as Bitcoin or Ethereum, or a proprietary token created specifically for this platform.

[0413] process: 1. After contribution, the quality of the user's input will be evaluated. 2. Based on this evaluation, a certain amount of cryptocurrency will be allocated. 3. The user provides their cryptocurrency wallet address. 4. The platform sends the cryptocurrency to the specified address, and the transaction is recorded on the blockchain. 5. The user receives a notification that the transaction is complete. 3. Gift cards and electronic vouchers

[0414] Reward method:

[0415] Users earn points based on their contributions, which can then be exchanged for gift cards or e-vouchers for popular online retailers such as Amazon and eBay.

[0416] process: 1. Each contribution will be evaluated, and points will be awarded based on a predetermined scoring system. 2. Users can accumulate points and browse gift card and e-voucher options on the platform. 3. After selecting their desired reward, the user confirms their selection and provides their email address. 4. The platform processes the exchange, and the user receives an electronic voucher via email. 5. The voucher can be used directly on the retailer's website. 4. Opportunities for professional development

[0417] Reward method:

[0418] Instead of direct monetary compensation, users are provided with limited access to professional development resources related to their field, such as online courses, webinars, or membership in professional societies.

[0419] process: 1. Contributions are evaluated based on their impact and relevance. 2. Based on their performance evaluation, users will be provided with access to specific professional development resources. 3. The user selects the desired resource and provides the necessary information for registration. 4. The platform arranges access by either directly registering the user or by providing an access code. 5. The user receives confirmation and instructions on how to access the selected resource. 5. Access to exclusive content

[0420] Reward method:

[0421] Contributors will be granted access to premium content such as specialized research papers, articles, and software tools that would otherwise require a subscription or one-time purchase.

[0422] process: 1. User contributions are evaluated, and points are awarded based on their value. 2. Accumulated points can be exchanged for access to exclusive content within the platform. 3. The user selects the content they want to access and confirms their selection. 4. The platform releases access to content for users, often providing a unique access link or code. 5. The user receives a notification and instructions on how to access the content. 6. Awards and Recognition

[0423] Reward methods:

[0424] Users will be publicly recognized for their contributions. This may include featured profiles, awards, or certificates of contribution on the platform, which can be used for professional advancement.

[0425] process: 1. Contributions will be evaluated based on their innovativeness and impact. 2. Individuals who make outstanding contributions will be selected and awarded. 3. Selected contributors will be notified and asked whether they wish to participate in the awards program. 4. Contributors who agree may receive awards or certificates and may be featured on the platform or in related communications. 5. This award serves as a recognition of professional achievements that enhance the contributor's reputation in their field of expertise. 7. Referral Bonus

[0426] Reward method:

[0427] Users are encouraged to refer other experts to the platform and receive bonuses for each successful referral that contributes to the platform.

[0428] process: 1. Users will be provided with a unique referral link that they can share with potential contributors. 2. If someone registers through this link and makes a qualifying contribution, the original referrer will receive a bonus. 3. Bonuses may be offered in the form of direct payments, points towards rewards, or other incentives. 4. The platform tracks referrals and contributions and guarantees accurate compensation. 5. The referrer will receive a notification about the bonus and will be provided with details on how to receive it. 8. Subscription Credit

[0429] Reward method:

[0430] For platforms offering subscription services, contributors receive credits that can be used to offset subscription fees, effectively reducing or eliminating costs. These credits can also be used for interest-spot markets, online advertising technologies, or AGI systems in general.

[0431] process: 1. Users earn credits based on the quality and frequency of their contributions. 2. These credits will be applied directly to the user's subscription account within the platform. 3. Users will be notified about the credits they have earned and the resulting discount on their subscription fee. 4. Credits are accumulated and automatically applied to future billing cycles. 5. This method encourages continued user contributions and engagement with the platform. 9. Physical goods

[0432] Compensation method:

[0433] Users can choose branded merchandise or products related to their area of ​​expertise as rewards.

[0434] process: 1. Contributions are evaluated, and users earn points based on their input. 2. Users browse a catalog of available products and select items based on their accumulated points. 3. After making a selection, the user provides their shipping information. 4. The platform processes the order and ships the goods to the specified address. 5. Users receive goods as tangible rewards for their contributions. 10. Sponsorship of conferences and events

[0435] Reward method:

[0436] Highly active or valuable contributors may be eligible for sponsorship of professional conferences and events, which may include registration fees, travel expenses, or accommodation costs.

[0437] process: 1. Contributions will be evaluated based on their impact, with particular emphasis placed on contributors who consistently demonstrate high-quality participation. 2. Qualified contributors will be offered sponsorship for upcoming industry events related to their area of ​​expertise. 3. Interested contributors accept the sponsorship and provide the necessary details for registration and travel arrangements. 4. The platform handles all logistics on behalf of the contributors. 5. Contributors will receive a detailed itinerary and guide, enabling them to participate in the event with minimal out-of-pocket expenses. 6.11 Reputation Mechanism

[0438] Reputation metrics should be calculated for all human professionals and intelligent beings. Reputation metrics are a crucial element in determining which tasks and other cognitive work to assign or commission to which beings, and are also important in determining compensation for intelligent beings. The PPA and PCT cited earlier detail the components of reputation, including the process shown in Figure 7. Multidimensional reputation is more specific and useful than single-dimensional or summary reputation metrics (e.g., a 1- to 5-star rating). Multidimensionality allows for more refined matching of client and task requirements. For example, suppose there is a reputation metric for punctuality. If scheduling is a primary concern, it becomes possible to select an entity that excels in punctuality, even if it has a lower rating in other dimensions.

[0439] First, without limitation, we will list some of the potential aspects / metrics related to multidimensional reputation, and how reputation metrics are calculated or estimated and how they can be used. 1. Punctuality: Measures how often experts meet deadlines. Calculated as the percentage of work completed on or before the deadline. Updated after each project is completed. Experts with high punctuality scores may be preferentially selected for time-sensitive tasks. 2. Budget Adherence: This assesses the ability of experts to work within budget constraints. The calculation method compares the agreed budget with actual expenditures. It is updated after project completion. Experts with high scores are entrusted with financially challenging projects. 3. Work Quality: The quality of output is evaluated based on peer review, client feedback, and adherence to specifications. This is updated regularly based on feedback. High-quality work is prioritized in projects of high importance or high visibility. 4. Success Rate: This indicates the percentage of issues that were successfully resolved. It is calculated by dividing the number of successful outcomes by the total number of attempts. This rate is updated after each project. It is important when assigning complex issues. 5. Client Satisfaction: Measured through post-project surveys and feedback scores. High satisfaction increases the likelihood of being recommended for future client-driven projects. 6. Peer evaluation of competence: Collected through anonymous peer reviews, focusing on skills and knowledge. Helps identify mentors and experts in collaborative projects. Evaluation can be conducted by presenting the system with tasks to evaluate colleagues and can be crowdsourced through online advertising units. 7. External Recognition: Based on awards, publications, and external accolades. Maintained manually or, ideally, through automated analysis of LinkedIn® profiles and other publicly available information. Influences assignment to projects requiring certified expertise. 8. Innovation Score: Evaluates creativity and the ability to generate novel solutions. Assessment is conducted through peer review and client feedback (see 6). Important for research and development and creative projects. 9. Communication skills: Assessed by clients and colleagues (see 6), with a focus on clarity, conciseness, and effectiveness. Essential for leadership roles and projects requiring teamwork. 10. Adaptability: Measures the ability to respond to change and unexpected challenges. Updated after projects that experience significant scope changes. Valuable in dynamic environments. 11. Leadership Quality: Evaluated through feedback from colleagues and subordinates (see 6), with a focus on motivation, guidance, and decision-making. This is crucial for projects requiring team management. 12. Technical Proficiency: Evaluated based on the ability to appropriately apply technical skills to solve problems. This is important for technical or specialized tasks. 13. Learning Agility: Measures the speed and effectiveness of learning new skills and adapting to new technologies. Essential in rapidly evolving fields. Agility can be measured by automatically analyzing the sequence of tasks completed by humans and using indicators of similarity between those tasks. 14. Conflict Resolution: This is evaluated by observing the outcomes of conflicts in which experts are involved. It is useful for team-based projects. Relevant indicators include the frequency with which experts belong to the minority or majority when voting on an option, and the percentage of times, when they belong to the minority, that the experts were correct in the subsequent outcome. 15. Project Management Skills: Evaluated based on the ability to effectively plan, execute, and complete projects. This is important for roles involving project management responsibility. Indicators may include the number of project management tasks assigned to the specialist and the success rate of those tasks. 16. Reliability: Measure whether the system consistently delivers results and is available when needed. High reliability increases the likelihood of being entrusted with important or urgent tasks. Relevant metrics include response time and metrics related to items 1-5 above. 17. Efficiency: Evaluates the amount of resources (time and money) used to achieve results. Efficient professionals are preferred in projects with limited resources. 18. Cultural Fit: Evaluated through surveys and observations based on the alignment of values ​​with the organization and team. This impacts team-based and long-term projects. This relates to long-term projects carried out outside of advertising units (see Section 6.13). 19. Work Ethics: Evaluated through observation of colleagues and managers, with a focus on dedication and professionalism. High scores are important in all types of work. This is relevant to long-term projects performed outside of advertising units (see Section 6.13). 20. Client Retention Rate: The percentage of clients who become repeat clients or continue to collaborate with the expert. This indicates client trust and satisfaction and influences assignments with high-value clients. 21. Feedback Responsiveness: Measures how actively and constructively experts engage with feedback. This is important for continuous improvement and learning. 22. Networking skills: Assessed by the ability to build and maintain professional relationships. Useful for roles requiring outreach and collaboration. 23. Mentoring and Training: Evaluated based on contributions to the development of colleagues and subordinates. Important for internal capability building. This relates to long-term projects carried out outside the advertising unit (see Section 6.13). 24. Availability and Responsiveness: Track the availability of experts for new challenges and the speed of response to inquiries. This is important in environments where rapid response is required or where the client is in control. Online availability and response speed are relevant metrics. 25. Problem-solving speed: Measures how quickly experts can provide effective solutions. Valuable in time-sensitive projects. See item 24. 26. Creativity and Innovation: Evaluated through the originality and effectiveness of the solutions provided. Essential for roles requiring thinking beyond conventional boundaries. Creativity can be operationally defined as a solution possessing novelty and usefulness, and human-generated options can be evaluated by colleagues in these dimensions. Usefulness can also be confirmed by its correlation with the success of the task.

[0440] Reputation metrics should be recorded and updated automatically as much as possible. For example, the time to resolution and the number of steps to resolution in problem solving are easily automated, as are the success rates of solutions to attempted solutions. Other metrics, such as client satisfaction, may require surveys or other information gathering methods widely known in the industry. Automated external reputation assessment is possible by combining automated content analysis techniques with searches of the internet and other publicly available data sources regarding experts and specific entities. Naturally, existing and used automated background check techniques can also be used as part of public reputation assessment.

[0441] All these reputation dimensions may be used and updated within this technology depending on the specific design (customizable) of the advertising unit. However, they may also be used in collective problem-solving networks and AGI / SI systems, which can enhance the accuracy and usefulness of reputation in these systems. This is described in particular in Section 4.5, Figure 7, and the cited PPA and PCT. 6.12 Solving Issues within an Ad Unit

[0442] This technology assumes that the main problem-solving system constitutes part of the AGI / SI system. This AGI / SI system consists of numerous intelligent entities (human and non-human) that collaborate to solve problems and learn sequentially or in parallel, thereby improving the overall capabilities of the AGI network. This AGI / SI system is described in detail in the previously cited PPA and PCT, Section 4.5, and Figures 1 to 13. The general process of problem-solving is shown in Figure 2.4.10.

[0443] Figure 10 includes a step called "Identifying the Operator." In this step, the problem solver decides which action, or "operator," to apply next to move forward with the problem-solving process and get closer to a solution. In relation to current technology, when using online advertising units to collect expertise and information and utilize it for problem-solving and AI learning, one example of how problem-solving can be achieved within the advertising unit is shown in Figure 18 and explained as follows. 1) The AGI problem-solving system handles all steps of the universal problem-solving process shown in Figure 10, except for the step of "identifying operators". 2) For very simple tasks or sub-tasks, the AGI system will concisely present the current status of the task and the next goal or sub-goal within the online advertising unit. 3) Human users viewing the advertisement enter the next action (operator) to take to solve or make progress toward solving the main problem or sub-problem. 4) User input is sent to the AGI system and processed according to universal problem-solving methods. This processing may be combined with input from other users viewing online advertisements or from other intelligent entities (humans or AI) participating in the problem-solving network. 5) Based on the results of step 4, the system dynamically updates the online ad units. This update includes (but is not limited to) the following: a. Was the user's input accepted or rejected? b. Whether the goal or sub-goal was achieved. c. The input selected by the system to advance problem solving (e.g., action or operator), and the reason or explanation for that selection (optional). d. A description of the new task state after applying the selected action or operator. This task state may include new or updated goals / subgoals, lists (of operators), images, metrics, or other information sufficient to describe the new task state. e. The amount of credits or rewards accumulated by the user within the ad unit based on their input. f. Additional requests for new user input via online advertising units or links to more advanced problem-solving interfaces. 6) If the task / subtask is resolved, the user is given credit commensurate with their contribution. If it is not resolved, the user is prompted for additional input (5f), and the method is repeated from step 1. Alternatively, the user exits the ad unit and ends their participation in solving the task. 7) When a user leaves an ad unit and ceases participating in solving the task, they may receive any applicable rewards, or these may be continuously accumulated as credits in their account. Additionally, the user's reputation metric is updated and stored in association with the user's identification information within the system. 8) The AGI system's learning process (see Figure 5) increases the overall knowledge of the AGI system based on the user's successful or unsuccessful problem-solving attempts. This includes users who participate in problem-solving through online advertising units. In this way, online advertising units function not only as a means to solve specific problems, but also as a valuable means to enhance the overall knowledge, intelligence, and capabilities of the AGI system.

[0444] As is evident from the example above and the "dual value" arising from Step 8, online advertising units designed to help solve problems are far more valuable than traditional online advertising units used to sell goods or services, and therefore offer a significantly superior means of monetizing human interest online.

[0445] The applicant points out that using online advertising units to acquire knowledge and expertise about human behavior ("operators") is just one way to advance problem-solving and enhance AI intelligence through problem-solving monetization methods. While not limited to these, any of the problem-solving steps shown in Figure 2.4.10 can benefit from human expertise recruited and communicated through appropriately modified online advertising units. Specifically, When problem-solving is at the stage of defining the problem and defining the problem space, online advertising units can ask users to describe how they frame the problem using text, audio, or visual means. This description can then be translated into the language of the universal problem-solving framework (problem states, operators, goals, etc.) using the “natural language to problem-solving language translator” shown in Figure 6 and described in Section 4.5 and other cited PPAs and PCTs. When a problem has already been expressed and defined, and the means-end analysis or other heuristic methods are being applied to determine which subgoals should be set, an online advertising unit may, for example, ask the user to describe intermediate goals that will serve as "stepping stones" to the final solution, using text, audio, or visual means. These intermediate subgoals are communicated to and processed by the AGI problem-solving system. The user may reason using means-end analysis, in which case they are providing reasoning knowledge. Furthermore, if they use specific heuristics, such as setting subgoals based on past experience solving similar problems, they are providing a more powerful form of knowledge than general means-end analysis, and thus providing expert knowledge and expertise toward problem-solving. AGI systems may require user input regarding the safety and ethical implications of setting and pursuing specific goals or subgoals. In such cases, appropriately modified online advertising units can be used to solicit opinions on the safety or ethical aspects of proposed goals and subgoals. Utilizing online advertising units in this way enables real-time, dynamic human monitoring of safety and ethical considerations, which can be particularly important in enhancing the integrity, safety, and reliability of the AGI system. Users may be asked to vote, rank, or rate potential goals / subgoals or actions ("operators") proposed by themselves or other intelligent entities through online ad units. In this regard, all methods of voting, weighted voting, etc., as already disclosed in detail in the cited PPA and PCT, can leverage online ad units as an important means of obtaining dynamic, real-time information on which goals / subgoals or operators should be adopted in problem-solving efforts. Generally, as discussed above, the learning methods shown in the final boxes of Figures 5 and 10 can enhance the learning and intelligence of AI / AGI / SI entities and systems by leveraging input provided by human users through online advertising.

[0446] To clarify the potential scope and value of this technology, the applicant would like to describe a specific case of problem-solving via online advertising units. This particular type of problem involves providing feedback to LLMs or AI agents to help them learn. Large amounts of data, such as behavioral data and content obtained from the internet or within organizational data systems, are already available and are actually being used to train LLMs and other AI agents. However, in many cases, this data is qualitatively mediocre. This is understandable, because (as in the mythical Lake Wobegon effect) everyone wants to think they are "above average," but by definition this is impossible. As a result, while it is relatively easy to train LLMs or AI agents to perform at an "average level" on a particular task, it is extremely difficult to train them to perform at a level significantly above average unless specialized and extremely rich data is used.

[0447] Therefore, this technology, which uses online advertising units to identify and acquire information from individuals with such superior expertise, is particularly useful (and valuable) in addressing the challenging task of giving LLMs and other AI agents expert-level performance. This challenge of acquiring extremely valuable data can be considered a special challenge that can be solved by the method described above, especially by leveraging the capabilities that this technology provides: the ability to precisely target individuals with the required level of expertise online, and then extract that information using specially designed online advertising formats and make it available to AI agents that require training. This use case alone can increase online advertising monetization by at least 10 times compared to current levels.

[0448] The second point of clarification is that in the future, it will become commonplace for AI agents to act on behalf of human users, representing their interests online. This scenario is already becoming increasingly common and is expected to account for a large portion of internet traffic in the future. In this case, it will be AI agents, rather than humans, who primarily interact with online advertisements.

[0449] Therefore, where the applicant refers to “human user” or “user” in the above and throughout this specification’s description of methods and embodiments, in its most common, powerful, and preferred embodiment it actually means “intelligent entity user.” That is, an AI agent may encounter online advertising units that seek knowledge, expertise, or information to help solve a problem and / or train other AIs. And, if the AI ​​agent is sufficiently intelligent, its contribution through the advertising unit may be as valuable as, or in some cases more valuable than, that of a human expert.

[0450] In a preferred embodiment of such a case, the AI ​​agent is required to identify itself as an AI agent. This allows the system to filter out its input as needed. This is particularly important in safety and ethical issues where human opinion is required. It is important to note here that it is impossible to derive ethical values ​​and determine what is right and wrong from logic alone. Therefore, if most AI agents in the future possess intelligence higher than most human experts, the primary type of knowledge and information sought through this technology and other means may become ethical information. This is because, poetically speaking, it relies on the "human heart," not on mere intelligence. 6.13 Solving problems outside of ad units

[0451] It is important to understand that not all problem-solving must take place within an online ad unit. In a suitable embodiment, online ad units include links to web pages, apps, AI agents, and other means, allowing for use beyond the limitations of the ad unit's own limited space. The primary function of an online ad unit is to engage users in activities that capture their interest and involve them in sharing some of their knowledge. Generally, users are more receptive to one-click information provision within an ad unit, such as answering yes / no ethical questions or selecting from a list of options. However, for broader and more advanced problem-solving, while participation within the ad unit is possible as described above, it is usually more efficient to guide users from the ad unit to a more specialized user interface via links, which allows for a more user-friendly presentation of problem-solving capabilities.

[0452] Therefore, this technology includes all problem-solving methods and systems that focus on such systems and methods, as described in detail in Section 4.5, the relevant Figures 1 to 13, and the cited PPA and PCT.

[0453] Even securing information from just one click from the right user at the right time is a far superior method of monetizing clicks through traditional monetization techniques that direct users to promotional landing pages. The applicant demonstrated this fact in a previous company, showing that by soliciting user opinions on stock price trends, it was possible to generate 100 times the value of online ad units without specifically targeting ad units to stock market experts. However, to maximize the value of this technology, it is extremely effective to target online ad units to experts with the appropriate expertise at precisely the time they are needed, thereby significantly increasing the value and usefulness of the ad units and making it a superior method to any monetization method currently in existence. 6.14 Feedback Mechanism for Online Advertising Targeting

[0454] Current online advertising targeting systems are optimized to increase the "click-through rate" (CTR) of ads. This is because CTR is the primary piece of information advertisers possess. While advertisers typically measure how well their ads convert, or generate sales, they are often reluctant to share this additional information. Advertisers are secretive about conversion rates because they fear that if the true value of their ads were revealed to online advertising vendors, advertising prices would rise accordingly.

[0455] However, in a preferred embodiment of this technology, a feedback loop exists between reputation metrics or other indicators that assign merit or responsibility to each problem solver, and a targeting mechanism that delivers online advertisements to specific experts. This feedback loop is free from conflicts of interest, practically implementable, and a realistic means for large online advertising vendors who want to improve the intelligence of their AI by leveraging their online advertising capabilities.

[0456] Figure 19 illustrates the basic feedback mechanism for ad targeting, and the process proceeds as follows: 1. Online advertising companies use information about users (including, but not limited to, user profiles, cookies, user preferences, behavioral data, purchase history, site dwell time, page dwell time, app usage time, CTR, available conversion metrics, browsing history, viewing habits, email and text content analysis, prompts to AI agents, and other commonly known metrics in the industry) to target experts who may be able to provide useful information for solving problems. 2. Measure and record the response rate of users who provide information via links to problem-solving and / or AI training systems within (as described in this technology) or outside of ad units. 3. Record metrics related to the quality of information and / or problem-solving provided by targeted users. These metrics include all reputation metrics described in Section 6.11, as well as information metrics listed in the cited PPA / PCT, metrics based on KIT theory as detailed in the PCT “Catalysts for Growth of Super Intelligence,” and information quality metrics widely used in the fields of machine learning, intelligent online advertising targeting, and general problem-solving. 4. Calculate statistical relationships, such as correlations and regression coefficients, for the targeting factors in Step 1, and identify the factors that have the greatest impact on the key quality indicators in Step 3. Depending on which information, knowledge, and expertise were most effective in solving problems and / or achieving AI training goals, the organization can determine which indicators to prioritize. 5. Based on the analysis in Step 4, weight and adjust specific targeting factors, targeting processes, and algorithm parameters from Step 1. Then, use the improved targeting information, parameters, methods, and algorithms to target subsequent online advertisements, and repeat the process from Step 1. 6.14a Key to AI dominance

[0457] Google / Alphabet, Meta, Amazon, Apple, and many other leading online advertising technology companies are also highly focused on gaining a competitive advantage in the fields of AI agents and AGI. If these companies and organizations use this technology, there are previously unrecognized advantages. It enables them to solve problems on a collective intelligent entity network. This allows them to not only deploy AGI services immediately, but also train their AI agents (whether or not they wish to move to the next stage of AGI as described in the applicant's other inventions) by combining their advertising targeting capabilities with this technology.

[0458] Given that obtaining high-quality training data is becoming a bottleneck in the AI ​​"arms race," this technology is a novel and extremely useful tool for these companies and organizations. Furthermore, these companies have no obstacles in building a closed-loop feedback mechanism that directly correlates the usefulness of specific experts and expertise in training their AI agents with the information used to target and find those experts. Thus, these organizations have all the necessary components to maximize their online advertising targeting systems not for "click-through rates" or "conversions," but for providing the information most useful for improving the intelligence of their AI agents.

[0459] The magnitude of this advantage cannot be overstated. Consider the current situation where all major advertising technology companies are spending hundreds of billions of dollars to purchase Nvidia chips in an attempt to enhance the performance and intelligence of their AI. This situation is seen as key to dominating the multi-trillion dollar market represented by AI. These chips mean that power will be concentrated in the hands of chip companies like Nvidia. In fact, at the time of writing, orders for the chips are backed up for more than a year in advance, and the profit margins are extremely high.

[0460] However, there are alternatives to remaining under the control of chip manufacturers. These alternatives don't involve entering the chip manufacturing business, where all major players are already involved. Rather, they involve leveraging online advertising targeting capabilities to secure high-quality data—specifically, the data needed to enhance the intelligence of a company's AI agents in specific areas. One TB of high-quality data is worth more than 100 TB of mediocre data, and the computing resources (chips) required to train it are only 1% of that.

[0461] In short, major technology companies—at least those with large online advertising capabilities like Alphabet, Meta, Microsoft, Apple, and Amazon—are overlooking the fastest and best way to enhance their competitive edge in the AI ​​field: combining existing technologies with this technology to accelerate the learning and intelligence of their AI without increasing the demand for chips and computing resources. Of course, having access to high-quality data and abundant computing resources is ideal. However, as mentioned above, handling better data with fewer computing resources yields better results than handling mediocre data with many. In short, certain organizations are unaware that they sit on the key to AI dominance, and this technology, and the technologies detailed in the cited PPAs and PCTs, are the key to outperforming the competition. 6.15 Feedback Mechanisms / Processes in Spot of Interest Markets

[0462] Unlike many commodities traded in the market (such as AAA-rated corporate bonds), human interests and expertise vary greatly in quality and usefulness. As mentioned above, objective, automated, and / or third-party methods can be used to classify or grade human interests (e.g., distinguishing between unskilled and uneducated individuals and those with university education and software development experience). However, the usefulness of specific human interests varies far more than, for example, the grading of crude oil, corn, or soybeans. Fortunately, AI excels at pattern recognition and judgment (e.g., creditworthiness assessment and human selection based on resumes). Furthermore, problem-solving and AI training tasks have objective indicators of quality and success, as detailed above, which allow for a more precise estimation of the value of human interests for a particular task than can be done with a single soybean.

[0463] Therefore, a crucial element in accurately pricing human interest is knowledge of both a specific individual's performance record and the nature of the tasks they are required to perform. Using the KIT methodology described in previous PPA and PCT studies, it is possible to estimate the AI's contribution to intelligence by, for example, comparing the AI's goals and existing knowledge base to the knowledge and expertise of a specific individual (or other intelligent entity), and further adjusting for fluctuating performance metrics such as punctuality, reliability, and reputation.

[0464] By establishing a feedback loop between human performance indicators for a specific task and the price paid in the spot market, the system can, and is desirable, propose recommended prices for specific human expertise related to a particular task, based on available resources. It should be noted that this feedback only serves to fix potential bid and asking prices when the market's price discovery mechanism does not function efficiently with sufficient liquidity.

[0465] This feedback is considered most useful in objectively characterizing the nature of human interest, including its type and category, when bringing human interest to market. However, ultimately, it is the market price that determines the appropriate price for a particular human interest.

[0466] If a client overpays and is dissatisfied, that dissatisfaction will be reflected in reputation metrics that are fed back into the market, potentially leading to a lower quality rating or classification for that individual in future transactions. Conversely, if expectations are consistently exceeded and the client feels they received a favorable deal, that feedback may lead to an improved quality rating for that individual in the future.

[0467] One of the novel and useful features of this technology in general, and especially of the feedback mechanism related to the Spot of Interest market, is its ability to create a “vector” track record that meticulously documents every step (or incorrect step) in problem-solving, showing the performance of every human being in all tasks and subtasks with complete accuracy and transparency. This objective and transparent record can be implemented through blockchain technology (without limitation, including ETHEREUM-based smart contracts and tokens documented in other PPAs and PCTs), enabling a more accurate reputation that can be analyzed by AI and translated into a far more precise and fair value-added estimate compared to commonly used “5-star” rating systems, etc.

[0468] Therefore, based on this feedback mechanism, the same person might only earn a rate of $10 per hour for tasks that anyone (or most people) can perform equally or better, but could command a rate of $1,000 per hour for highly specific tasks that require unique information and expertise, and for which they have a track record of creating $100,000 worth of value in past projects. Such a feedback mechanism helps both sellers and buyers of interests to pay fairer prices for specific tasks and to realize more profitable transactions on the spot market. In other words, people are unique, and this technology helps people be fairly compensated for their uniqueness while preventing clients from overpaying for commoditized, low-value interests.

[0469] One example of a feedback process in a spot of interest market includes the following steps (also illustrated in Figure 20): 1. The purchase of human interest is contracted on the spot market, and the price and other details of the contract are recorded. 2. The work is performed by humans. 3. Metrics (including, but not limited to, reputation metrics in Section 6.11 and other work- and task-related metrics) are recorded (automatically) during work performance. 4. After the work is completed, payment details and client satisfaction are recorded. 5. All metrics are stored in association with transparent and auditable transaction records (optionally enabled by blockchain technology). 6. Upon completion of the task, or at a time when practically feasible, correlation analysis, other statistical analysis, or machine learning is performed to refine classifications that associate human interests with humans themselves and with different types of tasks. This updates reputation metrics and recommended price estimates when certain types of interests (including specific individuals or categories of individuals) are associated with certain types of tasks. 7. The next time a task or task category is put up on the spot market of interest, updated indicators and other information (such as information from Step 6) will be used to better match human interest to the task and to provide an estimated price. However, the market mechanism will ultimately determine the price at which a transaction is completed on the spot market. 6.16 Mechanisms for Continuous Improvement

[0470] Similar to the spot market feedback mechanisms and processes described in Section 6.15, the effectiveness of all steps in the overall online ad unit technology, including but not limited to online ad targeting, online ad unit size, shape, overall design and content, online ad unit placement, online ad unit display frequency, and other aspects of presentation and process flow, can be measured by metrics, analyzed and optimized using statistical methods and machine learning processes, and continuously improved. 6.17 Dynamic Arbitrage Trading Process

[0471] Online advertising companies themselves, intermediaries that sell online advertising to clients, and AI agents or systems that seek to improve their own intelligence may engage in dynamic arbitrage in the following ways: 1) Estimate the value of a specific type of human interest and knowledge regarding a particular task (e.g., a problem-solving task, a subtask, or an AI training task). 2) Estimate the cost of obtaining the amount of human interest necessary to complete the task from appropriate intelligent entities (human or non-human) using online advertising unit technology (including interest spot markets, fixed or variable pricing charged by online advertising companies through their various platforms and technologies, and other means). This cost should include any apportionment of other costs incurred by the entity performing the arbitrage. 3) If the estimated value in item 1 exceeds the estimated cost in item 2 by a variable representing a predetermined or dynamically changing profit margin, the interest is purchased at a cost approximating the value estimated in item 1 and sold to the client, or a task is performed using the interest to create value approximating the value estimated in item 2, and a profit is recorded. 4) Return to item 1 and execute arbitrage starting with the largest arbitrage opportunity, repeating until you reach the smallest acceptable arbitrage opportunity (potential profit).

[0472] The applicant believes that the highest value of the human interest of experts with unique data arises from applying that expertise to training AI agents. This is because, once trained, an AI agent or system can be reused indefinitely as long as its knowledge and expertise remain valid, and therefore such entities, or their owners, are willing to pay a higher price for human interest used in this way than other clients who wish to sell a single product or service or solve a problem only once.

[0473] As long as more intelligent AI entities (including AI agents, AGI systems, and SI systems) can improve themselves and further increase their intelligence and capabilities with the assistance of expert human interests, these entities are likely to find ways to increase the amount of money they can use to further enhance their intelligence. This positive feedback loop will continue until humans are no longer a useful source of information for improving their intelligence. Therefore, the use of online advertising unit technology will catalyze the intelligence growth of these AIs.

[0474] Companies that provide access to online advertising units and use these units to engage in interest arbitrage have the potential to earn enormous sums of money, but it is crucial to have safety measures in place regarding the types of tasks permitted within online advertising units and interest spot markets. For example, terrorists (human or AI) may be able to pay extremely high prices for specific expertise related to weapons of mass destruction, which must be prohibited. 6.18 Human Worker-Client Interface

[0475] The user interface for human workers and clients may include LLM or natural language text or voice-based interfaces. In this case, the human simply speaks to the AI ​​agent, which translates the utterance into a foundational universal problem-solving framework and coordinates the problem identification and resolution activities. The natural language translation process shown in Figure 6 is relevant in this regard. However, more specialized interfaces can also be constructed.

[0476] A guiding principle is that each step in problem-solving, particularly those described in Figures 4, 6, 8, and 10, may have a specific user interface for optimizing problem-solving within or outside the online advertising unit. Many variations, such as text boxes, dropdown lists, templates, dynamically resizable input / output areas, visual I / O devices, and VR devices (e.g., Apple's VisionPro or Meta's VR system), are widely known in the industry and are applicable to optimize the ease of execution of the problem-solving steps described in this disclosure and the previously cited PPAs and PCTs. 6.19 Automation AI / AGI Interface

[0477] In addition to using interfaces for humans (e.g., human clients and workers), this technology accepts any intelligent entity in both worker and client roles. Here again, natural language (e.g., that generated by an LLM agent) can function as a universal interface.

[0478] Multimodal AI agents can communicate problem specifications and express / solve ideas using images, audio, and even sensory and output tools that surpass human perception and generation capabilities.

[0479] For example, an AI agent attempting to solve a certain problem might use X-rays to detect fractures invisible to the human eye, and that same AI agent could 3D print a cast or generate chemical compounds for pain relief or healing. These are forms of output that a human alone could not produce.

[0480] Therefore, interfaces for non-human entities can generally include APIs, digital interfaces, and other interfaces, in addition to all modes that function with human intelligence. This allows for optimization of the efficiency of problem specification and resolution. 6.20 Optimization of advertising targeting and system efficiency through recursive use of problem solving

[0481] One of the novel and very powerful aspects of this technology is its ability to solve all kinds of cognitive problems. Therefore, while Sections 6.14 through 6.16 describe feedback and other specific methods for improving each aspect of this technology, a general method for optimizing its various aspects (including, but not limited to, ad targeting and overall system efficiency) is to define these tasks as problems to be solved and then allow the system itself to solve these improvement methods. This recursive use of the technology to improve the technology itself is a unique capability rarely found in other inventions and reflects the universality of the problem-solving capabilities underlying this technology. This capability is described herein and in the previously cited PPA and PCT. 6.21 Safety and Ethical Checks

[0482] Companies that provide access to online advertising units and engage in arbitrage through interest trading using these units have the potential to reap enormous profits. However, it is crucial to have safety measures in place regarding the types of tasks permitted in online advertising units and interest spot markets. For example, terrorists (human or AI) could potentially pay extremely high prices for specific expertise related to weapons of mass destruction, which must be prohibited.

[0483] Figure 8 illustrates scalable safety checks (e.g., performed each time a goal or subgoal is set) for a problem-solving system (which forms the foundation for problem-solving by AGI or AI agents, AAAI, and more broadly, other intelligent entities). Similar scalable checks should be incorporated into the process of creating tasks and subtasks (whether created by any intelligent entity) that appear in online advertising units or interest spot markets. That is, an ethical check should be performed before the task appears in either, and this check should be performed in much the same way as the process steps shown in Figure 8, with a final step of creating transparent and auditable records and continuously improving the safety check system. 6.22 Regulatory Compliance

[0484] The processes described in Section 6.21 and Figure 8 can be used alone or in combination with other existing methods known in the industry to curate, detect, and remove prohibited content from the platform, ensuring compliance with regulations and laws in the region and context in which the technology operates. In particular, AI agents trained to detect violating or non-compliant content may also be used to assist in regulatory compliance. 6.23 Universality of Platform and Cross-Cultural Systems and Methods

[0485] As is clear from the above discussion, the online advertising unit in this technology is applicable to any platform or technology that supports online advertising. Similarly, just as online advertising and related technologies are localized to accommodate different languages ​​and cultures, this technology can also be localized using methods well known in the industry.

[0486] One of the unique and innovative features of this technology is that, by leveraging the widespread use of existing online advertising, it provides a means of accessing interests, knowledge, and expertise from virtually every person on Earth. By utilizing existing online infrastructure that has been developed over the past two decades at a cost of tens of billions of dollars, or even trillions of dollars, this technology can leverage the collective intelligence of billions of people and combine it with non-human entities to solve any problem (through the collective intelligence AGI systems described in Section 4.5 and the previously cited PPA and PCT) and to train AI in all areas of cognitive activity. 6.24 Coordination and Dynamic Alignment Between Ad Units

[0487] Another innovative aspect of this technology is that, while the overall problem-solving process is coordinated by a collaborative AGI system, different problem solvers can address different aspects of the problem sequentially or in parallel.

[0488] For example, multiple online ad units can simultaneously solicit "next steps" from multiple individuals. These individuals can submit their proposed next steps through multiple online ads without necessarily being aware of submissions from others. A second group of online ad units then presents the next step options obtained from the first group, allowing the same or different individuals to vote on the preferred next step for problem-solving. In other words, all steps of sequential or parallel collaborative problem-solving, as shown in Figures 11 and 12, including the adjustment of substeps and assembly into an overall solution using the problem tree structure in Figure 13, can be utilized by this technology. The main difference is that problem-solving tasks are divided and distributed to solvers via multiple ad units, rather than through other types of interfaces defined by AGI network technology. Furthermore, if entities link from ad units to other types of interfaces, they can use those interfaces to address the problem as needed.

[0489] For the system as a whole, it doesn't matter whether the work is done sequentially or in parallel, or whether it's done within an ad unit or elsewhere. The process steps, including coordinating problem solvers, are largely the same; the main difference is the location of the work. 6.25 Integration of Real-time and Asynchronous Functionality / Data Feeds

[0490] Another innovative feature of this technology is its ability to incorporate real-time and asynchronous data feeds into ad units. For example, if the task presented within an ad unit is to recommend "which stocks to buy," then a real-time data feed of stock prices can be incorporated into the ad unit.

[0491] Furthermore, if the challenge involves sending a stock recommendation using the text functionality within an online ad unit and waiting for a response before adding additional recommendation text in the comments section, the online ad unit can wait for an asynchronous response from another user (who may be in a different ad unit) before proceeding.

[0492] Alternatively, some tasks may be performed asynchronously outside the ad unit (for example, through an email system), while others are performed in real time within the ad unit. Generally, this technology is designed so that problem solving progresses step by step, and that problem solving can proceed whether the steps and related information are provided in real time or asynchronously when the solver has an opportunity to respond or input. 7.0 Preferred Embodiments and Variations

[0493] The present technology, an online advertising unit for leveraging human interest or the interest of intelligent entities, can be implemented in a variety of ways, either independently or as part of a larger system. In this section, the applicant describes a typical use case with three variations and further refers to several systems and methods described in the previous section. These exemplary embodiments demonstrate how several systems and methods work together. For anyone skilled in software development and with some expertise in the field of online advertising systems, it will be obvious that many variations are possible, some of which include more or fewer methods than those included in the exemplary preferred embodiments. 7.1 Preferred Embodiments for Online Advertising Agencies and Clients

[0494] Consider use cases for companies whose business model involves selling online advertising and maximizing profits by displaying clients' ads, or for companies that act as brokers specializing in connecting clients with end online advertisers. Companies or organizations that derive a significant portion of their revenue from such business models include, but are not limited to, Alphabet (including its Google Search and YouTube® divisions), Meta (including its Facebook, Instagram, and Reels platforms), ByteDance (including its TikTok products), Baidu, X (formerly Twitter), Spotify, Snap, Pinterest, and advertising-focused companies (e.g., PubMatic, Magnite, Sea Limited, Criteo, TheTradeDesk, Jalopy, Taboola, and Outbrain). Companies like Amazon and Apple, which sell subscriptions and products but also derive significant revenue from advertising, are also relevant to this use case.

[0495] In the following examples, the company that owns the system and method of the present invention will be referred to as the "Company," and the client that seeks to purchase human interest and extract information, knowledge, and expertise will be referred to as the "Client."

[0496] The following steps illustrate a preferred embodiment of how a company can increase the monetization of its online advertising revenue while simultaneously providing clients with valuable, human-interest-based information, knowledge, and expertise. 1) The company constructs a database containing information useful for targeting specific types of online advertisements to specific users. Means of constructing the database include, but are not limited to, methods and technologies already owned by the company, purchasing user data and information from others, and constructing a database of human experts (including preferences and other information useful for advertising targeting) using the method of the present invention (Section 6.7). a. If the existing human expert database is not large enough, or if a company or client seeks to find and recruit more experts, the online advertising technology of the present invention can be used for this purpose (Sections 6.7 and 6.9). 2) Clients (e.g., OpenAI, or companies seeking to train / customize AI agents (Section 6.3) or solve problems (Section 6.2)) seek to purchase human interest or labor from the company. The client communicates the categories of human interest or labor they require, which, by the universal nature of the invention (Section 6.22), may include requirements relating to a specific group of people. 3) When requirements are communicated, checks are made to ensure that the requests do not violate any regulatory or ethical requirements (Section 6.21). Requirements may be communicated through existing means or through interaction with human interest spot markets (Section 6.6). a. When a human interest spot market is used, several auction methods and other mechanisms are available for pricing human interest and labor (Sections 6.6a, 6.6b, and 6.6c). b. Furthermore, spot market performance can be improved by using feedback loops and processes tailored to client needs (Section 6.15). 4) The Company may work with clients to design interactive online advertising creative content to elicit interest and work from human experts (Section 6.8), and then deliver the advertisements using the Company's existing ad targeting and display technologies. 5) Interactive online ads can acquire interest and labor (e.g., problem solving) within the ad unit itself (Section 6.12), or they can achieve optimized acquisition outside the ad unit by linking to a web page or other interface (Section 6.13). a. Depending on the nature of the work, interactive advertising may require collaboration by multiple people (Section 6.24) and / or integration with real-time or asynchronous data capabilities (Section 6.25). b. During problem-solving, scalable ethical and safety checks are performed to ensure that no unethical or unsafe expertise is used (e.g., tasks involving training or customizing AI agents) (Section 6.21 and Figure 8). 6) Based on reputation metrics and methods (Section 6.11), as well as other metrics related to problem-solving work and acquired knowledge, online advertising targeting can be improved through feedback loops and related methods (Sections 6.14, 6.15, and 6.16). 7.2 Modification 1 of a Preferred Embodiment: When “Company” is also “Client”

[0497] In some cases, the company and the client may be the same organization. Examples include, but are not limited to, Alphabet (including its Google Search and YouTube® divisions), Meta (including its Facebook, Instagram, and Reels platforms), X (including its X social media platform and its AI division, X.AI), Tencent (including its WeChat / advertising and AI divisions), Amazon (including its advertising, Mechanical Turk, and AI divisions), and Apple (including both its advertising and AI divisions). These companies have both online advertising capabilities and AI divisions.

[0498] These companies are particularly well-positioned to succeed in developing cutting-edge forms of AI in the current AI development race. This is because they can leverage their online advertising capabilities to acquire the expertise and knowledge lacking in their AI agents and other AI systems, and train their AI using human expertise acquired by combining the present invention with existing online advertising (targeting, etc.) functions. These companies have the advantage of not having to pay markups when using their online advertising capabilities and being able to rapidly and efficiently improve the intelligence of their AI agents and systems (including, but not limited to, AGI and SI systems) (see Section 4.5) by investing in the costs of displaying the advertising units of the present invention.

[0499] The steps in the preferred embodiment are largely the same as those disclosed in Section 7.1, except that the company and client are identical. Furthermore, it is possible to more closely integrate the feedback loop for improving ad targeting (Sections 6.14 and 6.16). The use of the human interest spot market is optional because the company / client has access to a large amount of human interest and expertise internally and does not need to bid for it externally. However, the company / client may wish to purchase additional human interest / expertise using the human interest spot market mechanism (Section 6.6) to complement the human interest accessible through its online advertising capabilities, in which case the feedback collected in a manner related to the spot market can also be closely integrated with the company / client's AI development activities (Section 6.15). 7.3 Modification 2 of a Preferred Embodiment: When Focusing on AI Safety / Ethics

[0500] Assuming that multiple AGI and SI systems are in operation and ensuring their safe and ethical operation is a top priority for governments, organizations, and humanity as a whole, one major application of the present invention is to collect ethical information and examine potential safety issues that arise during the process of solving certain problems or other cognitive activities of AI systems (see Section 5.3).

[0501] Therefore, a key variation of the basic embodiment outlined in Section 7.1 is to focus the online advertising unit on tasks that involve reviewing, evaluating, voting on ethical choices, or otherwise eliciting human opinions. These tasks include efforts to collect representative and statistically valid samples of human values ​​and opinions across diverse cultures and geographical locations (see Section 5.3a), leveraging the ability to access people through the diverse platforms of the Invention (see Section 6.23). It also includes areas of real-time oversight of ethical and safety issues (see Section 5.3b), where the ability to collect opinions from a large number of people on the same issue in parallel and to coordinate responses (see Section 6.24) is crucial. Furthermore, for dynamic tasks such as considering urgent ethical or safety issues, the ability to integrate real-time and asynchronous data into the online advertising unit (see Section 6.25) is also important.

[0502] As described above, by modifying the general embodiments described in Section 7.1 using the methods shown in each of the above sections, the present invention can be optimized for the purpose of improving AI safety and guaranteeing human-compatible value. 7.4 Modification 3 of a Preferred Embodiment: Including a Non-Human Entity

[0503] The description of this invention has primarily focused on embodiments that leverage human interest through online advertising units and human interest spot markets. However, as a novel system and method variation of the invention, it is possible to leverage the interest of any intelligent entity, whether human or AI, using the same inventive mechanisms, processes, and systems.

[0504] In general, the human interest spot market (Section 6.6) will be expanded into the intelligent entity spot market. In the future, as AI agents, AGIs, and SI systems will operate online representing the interests of humans (including other entities), these AI systems and agents will also interact with online advertising in the same way as humans. Therefore, online advertising units and interest spot markets may be modified to accommodate not only human agents but also non-human agents. The main difference in implementation is the use of an AI / AGI interface (Section 6.19) instead of the traditional human interface (Section 6.18).

[0505] Just as we can build profiles of individual humans for advertising targeting and use those profiles to target the necessary human expertise, we can also build profiles for advertising targeting for individual AI agents and use those profiles to target the necessary AI expertise.

[0506] Furthermore, the distinction between humans and their interests (Section 6.4) also applies to AI. AI itself is different from the interests derived from that AI. The primary focus of this invention is not to enclose the intelligent entity itself, but rather to utilize the interests and cognitive abilities of the intelligent entity (including the associated knowledge, information, and expertise) in novel ways.

[0507] Furthermore, the same methods can be applied to non-human agents (AI, AGI, SI, etc.) so that a database for identifying human-related expertise can be constructed using the methods of the present invention (Sections 6.7, 6.8, and 6.9). These methods can also be applied to non-human entities, just as it is possible to accumulate and associate reputation with humans (Section 6.11) and to implement feedback mechanisms to improve human targeting and the operation of the present invention (Sections 6.14, 6.15, and 6.16). It is important to note that intelligent entities can be both human and non-human.

[0508] Similarly, there are diverse sources of intellectual behavior. Traditionally, the source has been humans, but now and in the future, AI must also be included as a source of intellectual behavior. The interest of this invention lies in efficiently identifying and utilizing intelligence (including related information, knowledge, and expertise) to solve problems, and in particular to train other AIs and improve their intelligence.

[0509] In the short term, humans are the most skilled experts and possess the intelligence best suited to this task. However, in the future, non-humans may become fully capable of fulfilling this role and ultimately become the intelligence best suited to this task. The technologies according to the present invention (including online advertising units and interest spot markets) are a general mechanism for leveraging these diverse forms of intelligence. Therefore, modifications that target the interests of any intelligent entity, regardless of whether the interests are human or non-human (with appropriate modifications to the interface and methods for this purpose), that the spot market becomes a market for general "interests," and that the online advertising unit can capture problem-solving from any intelligent entity (including non-humans), are crucial for maximizing the value of the present invention in the future.

[0510] The problem-solving and AI intelligence enhancements generated by monetizing (human or non-human) interest through the novel methods of this invention are many times greater than the value generated by online advertising today. This is why the invention is not only novel but also extremely useful and valuable.

[0511] Figure 21 is a schematic diagram of a computer system 100 that can be used or implemented in a user's device and / or peripheral components of the present invention. The applicant notes that reference numerals 102-104 in Figure 21 refer to elements that may exist in both classical and quantum computing architectures, and that the present invention can be implemented using both architectures. In fact, quantum computing architectures have the ability to solve a large number of steps in parallel, which allows for searching a "problem space" or tree-structured data with many branches at once, thus greatly improving the problem-solving efficiency of AI agents and intelligent entities that utilize quantum computing architectures.

[0512] Computer system 100 constitutes part of an example machine and is an example of one or more computers referred to herein. This machine can execute instruction sets to perform any methodology discussed herein. In various embodiments, this machine may operate as a standalone device, or it may be connected to other machines (e.g., networked) and operate as a server or client machine in a network environment. It may also operate as a peer machine in a peer-to-peer (or distributed) network environment. This machine may be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a network router, a switch or bridge, or any machine whose actions are specified according to an instruction set executed sequentially or non-sequentially.

[0513] The computerized system 100 may include one or more ...

Claims

1. A system that utilizes online advertising technology to enhance the intelligence of an artificial intelligence (AI) agent or system, the system comprising a computer system, The computer system comprises a processor, a computer-readable storage medium, and program instructions. The program instruction is a program instruction stored in the computer-readable storage medium, To provide online advertising units to one or more user computer systems using a network, Receiving one or more inputs from the user computer systems participating in the online advertising unit, wherein the inputs are provided by intelligent entities utilizing each user computer system, and the inputs are associated with solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem, or a combination thereof, and further, the intelligent entities are any one or a combination of a human user utilizing a computer system, an AI agent or system, an artificial general intelligence (AGI) agent or system, and a superintelligence (SI) agent or system. Transmitting the aforementioned input to the AI ​​agent or the system, The AI ​​agent or the system uses the input in a universal problem-solving method to generate a solution to the problem or the sub-problem, The AI ​​agent or the system is trained based on the results of a universal problem-solving method derived from trials in which the problem or sub-problem was successfully solved or failed, thereby improving the intelligence of the AI ​​agent or the system. In order for the computer system to realize this, the processor can perform the following: A system characterized by the following features.

2. A method of online advertising technology used in conjunction with an artificial intelligence (AI) agent or system, the method for enhancing the intelligence of the AI ​​agent or system. A step of receiving problem information relating to a problem provided by the AI ​​agent or the system, A step of generating an online advertising unit that includes the aforementioned problem information, A step of providing the online advertising unit to one or more user computer systems using a network, A step of receiving one or more inputs from the user computer systems participating in the online advertising unit, wherein the inputs are provided by intelligent entities utilizing each user computer system, and the inputs are associated with solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem, or a combination thereof, and further, the intelligent entities are a human user utilizing a computer system, an AI agent or system, an artificial general intelligence (AGI) agent or system, and a superintelligence (SI) agent or system, or a combination thereof. A step of transmitting the input to the AI ​​agent or the system, The AI ​​agent or the system includes a step of using the input in a universal problem-solving method to generate a solution to the problem or the sub-problem, The process involves training the AI ​​agent or the system based on the results of a universal problem-solving method derived from trials in which the problem or sub-problem was successfully solved or unsuccessfully solved, thereby improving the intelligence of the AI ​​agent or the system. including, A method characterized by the following features.

3. The process further includes receiving advertising specifications from the client, which are used when generating the online advertising unit, in combination with the aforementioned problem information. The method according to claim 2, characterized in that

4. The advertising specifications include one or a combination of the following: advertising content, demographic information, location restrictions for online ad units, advertising budget, and metrics for determining whether the problem or sub-problem has been successfully resolved. The method according to claim 3, characterized in that

5. The process further includes charging clients based on cost per 1,000 impressions or click-through rate metrics, The method according to claim 2, characterized in that

6. The aforementioned problem information includes one of the following: the current status of the problem, the sub-problems of the problem, the goal of the problem, or the sub-goals of the goal, or a combination thereof. The method according to claim 2, characterized in that

7. The process further includes transmitting one or more additional inputs from an intelligent entity to the AI ​​agent or the system in combination with inputs from a participating user computer system. The intelligent entity is one of the following: an additional human user utilizing an additional computer system, an additional AI agent or system, an artificial general intelligence (AGI) agent or system, a superintelligence (SI) agent or system, or a combination thereof. The method according to claim 2, characterized in that

8. The update of the online ad unit includes one or a combination of the following: whether the input was accepted or rejected, whether the goal or sub-goal of the task was achieved, whether the input was selected by the AI ​​agent or the system to advance the task resolution, a description of the new task state after applying the selected input, the amount of credit or reward accumulated by the intelligent entity based on that input, and an additional request to solicit new user input via the online ad unit or link and involve them in the task resolution process of the task. The method according to claim 2, characterized in that

9. If the aforementioned problem or sub-problem is solved, the process further includes awarding credit to the participating intellectual entity based on its contribution. The method according to claim 2, characterized in that

10. If the aforementioned problem or sub-problem is not resolved, the process further includes requesting additional input from human users of each participating user computer system. The method according to claim 2, characterized in that

11. The process further includes a step of compensating a human user if that human user exits the online advertising unit and ceases to participate in the problem-solving process, The method according to claim 2, characterized in that

12. The learning of the AI ​​agent or the system further includes a learning method selected from the group consisting of differential privacy, federated learning, homomorphic cryptography, synthetic data generation, secure multi-party computation, data anonymization, transfer learning, active learning, self-supervised learning, domain adaptation, reinforcement learning, few-shot learning, explainable AI, privacy-preserving record linkage, data augmentation, generative adversarial networks (GANs), crowdsourcing for data labeling, personalization layers of models, knowledge distillation, and ensemble learning. The method according to claim 2, characterized in that

13. The process further includes creating a database of experts in intelligent entities, wherein the experts in intelligent entities are one or a combination of a human user utilizing a computer system, an additional AI agent or system, an artificial general intelligence (AGI) agent or system, and a superintelligence (SI) agent or system. The method according to claim 2, characterized in that

14. The online advertising unit is provided to one or more experts of intelligent entities from the database. The method according to claim 13, characterized in that

15. The creation of the aforementioned database is performed using technologies selected from the group consisting of relational database management systems, NoSQL databases, data warehouses, data mining, machine learning algorithms, graph databases, vector databases, indexing, full-text search, blockchain, data visualization tools, application programming interfaces, data cleansing and preprocessing, cloud storage, caching, transactional database systems, real-time database systems, content delivery networks, data compression, data encryption, and replication. The method according to claim 13, characterized in that

16. This further includes a process of bidding based on interest, utilizing the interest spot market. The method according to claim 2, characterized in that

17. The aforementioned spot market of interest is A means for interested human users to access the market and specify seller information and the selling price of the information, A means by which buyers who purchase human users' interests can access the market and specify the buyer information and bid price they wish to purchase, A market mechanism for queuing the aforementioned bid price and the aforementioned selling price, comprising categories of buyer and seller information, where each category has a market; Includes, The aforementioned market mechanism is configured to allow markets in each category to be formed by market shapers. The aforementioned market mechanism is configured to allow matching of bid prices and selling prices, and to facilitate binding transactions between buyers and sellers of information. The method according to claim 16, characterized in that

18. The aforementioned spot market of interest is The process involves sellers and buyers of human users' interests and expertise registering on the platform, buyers providing details of their interests and expertise, and sellers providing their available time slots and areas of expertise. The process involves listing the seller's available time slots and areas of expertise, and the buyer's preferred time slots and needs. The platform uses algorithms to dynamically determine the price of human users' interests and expertise based on supply, demand, and user ratings. The process involves matching one or more buyers with one or more sellers based on requirements, availability, and price. This process enables transactions where the buyer makes a payment for the seller's time, and the platform earns a fee. After each session, buyers and sellers evaluate each other and provide feedback that will influence future pricing and matching. including, The method according to claim 16, characterized in that

19. The aforementioned platform utilizes mechanisms selected from auction formats, dynamic pricing, supply and demand adjustment, escrow systems, tokenized transactions, etc. The method according to claim 18, characterized in that

20. The aforementioned spot market of interest is The process involves human users registering as buyers and sellers of their interests and expertise on the platform, with buyers providing details of their interests and expertise, and sellers providing details of their available time slots and areas of expertise. The process by which a seller creates an auction regarding the seller's time slot, The process involves submitting one or more bids regarding the time frame and expertise required by the buyer, The process of ending the auction either at a predetermined time or when the seller accepts the buyer's bid, The process involves the buyer who wins the auction making payment to the seller, the seller providing interest or expertise based on the time frame, the platform mediating the transaction and ensuring payment, and After each auction ends, buyers and sellers evaluate each other and provide feedback that will influence future auctions and visibility on the platform. including, The method according to claim 16, characterized in that

21. The aforementioned platform utilizes mechanisms selected from Dutch auctions, reverse auctions, sealed bid auctions, ascending auctions, time-priority fixed pricing, valuation-based dynamic pricing, supply and demand curve adjustments, time zone segmentation, subscription access, freemium models, group purchases, tiered pricing by expertise level, flash sales, loyalty point systems, hybrid auctions, geographic pricing, behavioral pricing, escrow systems, social impact pricing, and tokenized transactions. The method according to claim 20, characterized in that

22. The online advertising unit is configured, or can be configured, to include materials related to the area of ​​expertise associated with the issue information. The method according to claim 2, characterized in that

23. The online advertising unit is configured, or can be configured, to expand or redirect a human user to a secure form on a landing page when interacted with by that human user, and the secure form is configured to collect user information and provide questions regarding the human user's credentials. The method according to claim 2, characterized in that

24. The process further includes verifying the user information received in the secure form and classifying the human user in the database based on their credentials. The method according to claim 23, characterized in that

25. The online advertising unit is configured, or configurable, to include an interface in which a human user can input a solution to a problem or sub-problem. The method according to claim 2, characterized in that

26. The online advertising unit is configured or configurable to include one or a combination of gamification, a scoring system, and a leaderboard of other human users. The method according to claim 25, characterized in that

27. The online ad unit is configured, or can be configured, to include a text input field or other interactive tool that allows a human user to directly provide a solution to a problem or sub-problem within the space of the online ad unit. The method according to claim 25, characterized in that

28. The aforementioned solutions are stored in a database, and an algorithm is used to evaluate the quality and relevance of the solutions based on predefined criteria. The method according to claim 27, characterized in that

29. The aforementioned human user can receive immediate feedback or reward points based on the quality and relevance of the evaluated solution, which can then be exchanged for various incentives. The method according to claim 28, characterized in that

30. The contributions of the aforementioned human users are tracked, and top contributors are identified from among multiple additional human users as potential future participants. The method according to claim 28, characterized in that

31. The online advertising unit is configured, or can be configured, to expand or redirect a human user to a secure form on a landing page when interacted with by that human user, and the secure form is configured to collect user information and provide questions regarding the human user's credentials. The method according to claim 25, characterized in that

32. The process further includes verifying the user information received in the secure form and classifying the human user in the database based on their credentials. The method according to claim 31, characterized in that

33. The process involves analyzing the input from the participating human users and determining the quality or relevance of each input. A process of providing the participating human users with rewards based on the quality or relevance of their inputs, Further including, The method according to claim 2, characterized in that

34. The process further includes each participating human user providing digital wallet information to a secure form within the online advertising unit or to a linked platform. The method according to claim 33, characterized in that

35. The aforementioned rewards may be provided in the form of one or a combination of cryptocurrency, electronic gift cards, electronic vouchers, subscriptions, access to online resources or content, recognition on online platforms, products related to human user expertise, or sponsorship of professional conferences or events. The method according to claim 33, characterized in that

36. The process further includes encouraging the participating human users to refer other human users to the online platform and providing additional rewards for each referral that contributes to the platform, The aforementioned online platform receives the task and provides online advertising units. The method according to claim 33, characterized in that

37. The process further includes receiving evaluation metrics for each human user, The method according to claim 2, characterized in that

38. The aforementioned evaluation metrics are multifaceted and include one or a combination of the following: punctuality, budget adherence, work quality, solution success rate, client satisfaction, peer evaluation, external evaluation, innovation score, communication skills, adaptability, leadership, technical proficiency, learning agility, conflict resolution ability, project management ability, reliability, efficiency, cultural fit, work ethic, client retention rate, feedback responsiveness, networking ability, mentoring / education, adaptability / responsiveness, problem-solving speed, and creativity / innovation. The method according to claim 37, characterized in that

39. The aforementioned evaluation metrics are automatically updated within the online ad unit for each human user. The method according to claim 37, characterized in that

40. The process further includes updating the online advertising unit based on the results of the universal problem-solving method implemented for the aforementioned problem or sub-problem, The method according to claim 2, characterized in that

41. The online advertising unit is configured, or can be configured, to prompt a human user to describe how a problem should be defined, and this description is translated into the language of the universal problem-solving method using a translation means from natural language to problem-solving language. The method according to claim 2, characterized in that

42. The online advertising unit is configured, or can be configured, to prompt a human user to describe an intermediate subgoal toward the final solution of the problem, and such intermediate subgoal is communicated to the AI ​​agent or the system and used for processing by the universal problem-solving method. The method according to claim 2, characterized in that

43. The process further includes requesting the AI ​​agent or system to provide a human user with safety or ethical information regarding a task, a subtask, or a goal related to the task, The method according to claim 2, characterized in that

44. The generation or updating of the online advertising unit further includes a step of utilizing the security or ethical information, The method according to claim 43, characterized in that

45. The process further includes modifying the online advertising unit and prompting human users to vote, rank, and rate one or more potential goals, subgoals, or actions proposed by either human users or intelligent entities, or a combination thereof. The method according to claim 2, characterized in that

46. If the AI ​​agent or the system is required to identify itself as an AI agent and only human opinions are sought, the process further includes excluding input from the AI ​​agent or the system. The method according to claim 2, characterized in that

47. The aforementioned online advertising unit is provided by an online advertising service provider that utilizes user information about human users who have expertise related to the issue, in order to target human users who have expertise related to the issue. The method according to claim 2, characterized in that

48. The process further includes measuring and recording the response rate of participating human users who provided information through links to websites or interfaces within or outside the online advertising unit. The method according to claim 47, characterized in that

49. The process further includes recording metrics related to the quality of information or solutions provided by targeted human users, The method according to claim 47, characterized in that

50. The process further includes calculating correlations or statistical relationships based on the user information, identifying which factors have the greatest impact on the metrics, which metrics online advertising service providers prioritize, and which inputs were most effective in solving the problem or in training the AI ​​agent or the system. The method according to claim 49, characterized in that

51. This further includes the process of conducting arbitrage, The arbitrage transaction in question is, With respect to the aforementioned problem or sub-problem, the process involves an online advertising service provider estimating the value of the interests and knowledge of a particular type of person, The process by which an online advertising service provider estimates the cost of obtaining the degree of human interest necessary to complete the aforementioned task or subtask from the human user or intelligent entity via the online advertising unit, A step of determining whether the estimated value exceeds a predetermined or dynamic variable representing the profit margin, and if so, using the degree of human interest to perform a task of purchasing human interest at a cost approximating the estimated value and selling it to the client, or creating value close to the estimated cost, including, The method according to claim 47, characterized in that

52. The arbitrage transaction further includes a step of prioritizing the processing of the largest arbitrage opportunity and repeating it until the smallest acceptable arbitrage opportunity is reached. The method according to claim 51, characterized in that

53. The online advertising unit is configured, or can be configured, to include one or more interfaces selected from the group consisting of a large language model, natural language text, a voice-based interface, a text box, a dropdown list, a template, a dynamically resizable input / output area, a visual input or output device, a virtual reality device, and an augmented reality device. The method according to claim 2, characterized in that

54. The online advertising unit comprises multiple online advertising units, wherein the first online advertising unit is provided to a first group of human users for receiving input, and the second online advertising unit is provided to either the first or second group of human users and is configured, or configurable, to solicit votes for the input. The method according to claim 2, characterized in that

55. The input is configured to be provided in real time to the AI ​​agent or the system for real-time use by the universal problem-solving method. The method according to claim 2, characterized in that

56. A method for increasing the monetization of online advertising revenue for online advertising service providers by providing information to clients based on the interests of intelligent entities, The process of an online advertising service provider constructing one or more databases containing information relating to targeting a particular type of online advertisement to a particular intelligent entity, wherein the intelligent entity is one or a combination of a human user utilizing a computer system, an AI agent or system, an artificial general intelligence (AGI) agent or system, and a superintelligence (SI) agent or system. A step in which a client communicates a request to the online advertising service provider to purchase the interest of an intellectual entity, wherein the request includes requirements and desired categories for the interest of the intellectual entity. The online advertising service provider performs a check on the request at the time the requirement is communicated, the check being configured or configurable to ensure that the request does not violate any regulatory or ethical requirements. The process of creating content for an interactive online advertising unit that is configured or configurable to acquire interest and work from intelligent entities, The process involves the online advertising service provider delivering the interactive online advertising unit to the computer system of the intelligent entity via a network. The process includes enabling the interactive online advertising unit to collaborate among multiple intelligent entities and integrate with real-time or asynchronous data capabilities, A step of performing ethical and safety checks during the problem-solving process in a problem provided by a client, wherein the ethical and safety checks are configured or can be configured to ensure that unethical or unsafe expertise is not part of the problem; A step of improving the interactive online advertising unit through a feedback loop based on evaluation metrics, wherein the evaluation metrics include one or a combination of reputation metrics and metrics related to problem-solving processes or acquired knowledge. A method characterized by including

57. The aforementioned client is one or a combination of an artificial intelligence (AI) agent or system, a system for training or customizing an AI agent, or a system for providing solutions to a problem. The method according to claim 56, characterized in that

58. If the database does not contain a predetermined number of human users, or if additional human users are required, the further step includes configuring the interactive online advertising unit to acquire additional human users. The method according to claim 56, characterized in that

59. The aforementioned requirements are configured to be communicated through interaction with the human interest spot market. The method according to claim 56, characterized in that

60. The aforementioned spot market of interest is A means for interested human users to access the market and specify seller information and the selling price of the information, A means by which buyers who purchase human users' interests can access the market and specify the buyer information and bid price they wish to purchase, A market mechanism for queuing the aforementioned bid price and the aforementioned selling price, comprising categories of buyer and seller information, where each category has a market; Includes, The aforementioned market mechanism is configured to allow markets in each category to be formed by market shapers. The aforementioned market mechanism is configured to allow matching of bid prices and selling prices, and to facilitate binding transactions between buyers and sellers of information. The method according to claim 59, characterized in that

61. The aforementioned spot market of interest is The process involves sellers and buyers of human users' interests and expertise registering on the platform, buyers providing details of their interests and expertise, and sellers providing their available time slots and areas of expertise. The process involves listing the seller's available time slots and areas of expertise, and the buyer's preferred time slots and needs. The platform uses algorithms to dynamically determine the price of human users' interests and expertise based on supply, demand, and user ratings. The process involves matching one or more buyers with one or more sellers based on requirements, availability, and price. This process enables transactions where the buyer makes a payment for the seller's time, and the platform earns a fee. After each session, buyers and sellers evaluate each other and provide feedback that will influence future pricing and matching. including, The method according to claim 59, characterized in that

62. The method according to claim 61, characterized in that the platform utilizes one or more mechanisms selected from the group consisting of Dutch auctions, reverse auctions, sealed bid auctions, open rising price auctions, time-priority fixed pricing, valuation-based dynamic pricing, supply and demand curve adjustment, time frame segmentation, subscription access, freemium models, co-buying, expertise level tiering, flash sales, loyalty point systems, hybrid auctions, regional pricing, behavioral pricing, escrow systems, social impact pricing, and tokenized transactions.

63. The aforementioned spot market of interest is The process involves human users registering as buyers and sellers of their interests and expertise on the platform, with buyers providing details of their interests and expertise, and sellers providing details of their available time slots and areas of expertise. The process by which a seller creates an auction regarding the seller's time slot, The process involves submitting one or more bids regarding the time frame and expertise required by the buyer, The process of ending the auction either at a predetermined time or when the seller accepts the buyer's bid, The process involves the buyer who wins the auction making payment to the seller, the seller providing interest or expertise based on the time frame, the platform mediating the transaction and ensuring payment, and After each auction ends, buyers and sellers evaluate each other and provide feedback that will influence future auctions and visibility on the platform. including, The method according to claim 59, characterized in that

64. The aforementioned platform utilizes mechanisms selected from Dutch auctions, reverse auctions, sealed bid auctions, ascending auctions, time-priority fixed pricing, valuation-based dynamic pricing, supply and demand curve adjustments, time zone segmentation, subscription access, freemium models, group purchases, tiered pricing by expertise level, flash sales, loyalty point systems, hybrid auctions, geographic pricing, behavioral pricing, escrow systems, social impact pricing, and tokenized transactions. The method according to claim 63, characterized in that

65. The process further includes providing a feedback mechanism related to the aforementioned spot market of interest, The feedback mechanism is configured, or can be configured, to record each step of the problem-solving process and create a vector track record of performance across all tasks and subtasks for each human user. The method according to claim 59, characterized in that...

66. The aforementioned vector track record is implemented using blockchain technology, enabling precise reputation, and is configured or configurable to be analyzed by the AI ​​agent or system and converted into an estimate of value for each task for each human user. The method according to claim 65, characterized in that

67. The interactive online ad unit is configured, or configurable, to capture and engage human interest by providing a link to a webpage or interface optimized for acquisition within or outside the interactive online ad unit. The method according to claim 56, characterized in that

68. The construction of the aforementioned database includes the process by which the online advertising service provider purchases user data and information from remote sources. The method according to claim 56, characterized in that

69. A method of online advertising technology used in conjunction with an artificial intelligence (AI) agent or system, A step of generating an online advertising unit that includes problem information provided by the AI ​​agent or the system, A step of providing the online advertising unit to one or more user computer systems using a network, A step of receiving one or more inputs from the user computer system participating in the online advertising unit, wherein each input is provided by an intelligent entity, and the input is associated with solving a problem, solving a sub-problem of the problem, or contributing to the progress of the problem, or a combination thereof, and further, the intelligent entity is one or a combination thereof, of a human user utilizing the computer system, an AI agent or system, an artificial general intelligence (AGI) agent or system, and a superintelligence (SI) agent or system. A step of transmitting the input to the AI ​​agent or the system, The process involves the AI ​​agent or system, or an additional AI agent or system, using the input in a universal problem-solving method to generate a solution to the problem or the sub-problem, A process utilizing a feedback mechanism for training the AI ​​agent or system by providing metrics to the AI ​​agent or system, and a targeting mechanism for directing the online advertising unit to a specific intelligent entity, wherein the metrics are used to assign merit or negligence to the intelligent entity. including, A method characterized by the following features.

70. The method according to claim 69, further comprising the step of measuring and recording the response rate of participating human users who provide information through links to websites or interfaces within or outside the online advertising unit.

71. The method according to claim 69, further comprising the step of recording the metrics relating to the quality of information or solutions provided by the target human user.

72. The method according to claim 71, further comprising the steps of calculating correlations or statistical relationships based on the information of the human user, identifying the factors that have the greatest influence on the indicator, the indicator preferred by the online advertising service provider, and the inputs that were most effective in solving the problem or achieving or maximizing the training of the AI ​​agent or system.

73. This further includes the process of conducting arbitrage, The arbitrage transaction in question is, With respect to the aforementioned problem or sub-problem, the process involves an online advertising service provider estimating the value of the interests and knowledge of a particular type of person, The process by which an online advertising service provider estimates the cost of obtaining the degree of human interest necessary to complete the aforementioned task or subtask from the human user or intelligent entity via the online advertising unit, A step of determining whether the estimated value exceeds a predetermined or dynamic variable representing the profit margin, and if so, using the degree of human interest to perform a task of purchasing human interest at a cost approximating the estimated value and selling it to the client, or creating value close to the estimated cost, including, The method according to claim 69, characterized in that

74. The arbitrage transaction further includes a step of prioritizing the processing of the largest arbitrage opportunity and repeating it until the smallest acceptable arbitrage opportunity is reached. The method according to claim 73, characterized in that

Images