Method for responding to client request and system for chat conversation

Abstract

To provide a method for responding to a client request and a system for chat conversation.SOLUTION: A method includes receiving a client request from a client device across an interface; determining a client identifier (ID) of a user who issues the request and / or a client component; processing the request and identifying and / or selecting at least one content source for giving paywalled content for responding to the request; issuing a secondary request to the identified / selected content source and providing content; receiving content from the identified / selected content source in response to the secondary request; inputting at least a portion of the received content and at least a portion of the client request into at least one trained model; transmitting at least a portion of an output of the trained model to the client component to provide a response to the client request; and optionally assigning the amount to be paid for transmission of the response using the client identifier (ID).SELECTED DRAWING: None

Description

[Technical Field]

[0001] explanation The present invention relates to a method for providing responses to client requests and a system for conducting chat conversations. [Background technology]

[0002] Systems for conducting online chat conversations are known. It is becoming increasingly common to have chatbots or chatterbots participate in chats that are traditionally conducted by humans. In recent years, service industries have developed that use chatbots to respond to customer requests (e.g., answering simple questions about pricing, service terms, etc.). Chatbots help reduce the workload for service workers and eliminate the most basic questions, allowing humans to respond to more complex and sophisticated queries.

[0003] As advances in training models (e.g., neural networks) have increased, so-called chatbots' ability to respond to more complex questions has increased. In November 2022, OpenAI released Chat Generative Pre-Trained Transformer, also known as ChatGPT. ChatGPT provides a chatbot that generates human-like text using an autoregressive language model generated by deep learning. Using large-scale language models (LLMs), technology is moving from responding to predefined questions to generating longer texts.

[0004] When generating text to provide clever responses, applications like ChatGPT require a large amount of computational power and therefore consume excessive energy. Therefore, there is a growing need to provide economical and intelligent ways to coordinate the use of such systems.

[0005] Furthermore, in some scenarios, the problem of how to handle copyrighted content that may be hidden behind a paywall is exacerbated: some LLMs are trained on copyrighted data, and it is nearly impossible to trace a particular response given by a chatbot using an LLM back to one or more sources that, to some extent, formed the basis for this particular response. This makes fair compensation for authors of paid content impossible due to the unresolved technical issues of mapping responses to sources.

[0006] Additionally, accessing paywalled content presents challenges due to complex case-by-case access solutions, making it difficult for users and even machines to access certain paywalled content. For machines, each integration of a new content provider typically requires the installation of special publishing software. There is also a risk to consumers of paywalled content that they may be tracked and / or exposed to malicious software. A further challenge with sharing available content is the risk of shared content interfering with a publisher's search engine optimization.

[0007] Another technical issue is that LLMs are trained on data that is available up to a certain point in time. In other words, the most recent data is not available to the LLM. Furthermore, training requires a large amount of training data, and it requires a large amount of energy to perform the training. For example, training a new LLM on just the most recent data from last year is inefficient in terms of energy consumption.

[0008] Ultimately, the above-mentioned problems lead to situations where certain content, particularly paywalled content, cannot be processed by the LLM, which can have a significant impact on the responses generated or, more abstractly, the quality of the services provided using the LLM. Summary of the Invention [Problem to be solved by the invention]

[0009] In view of this prior art, it is an object of the present application to provide an improved method for providing responses to client requests. A further object is to provide responses to client requests in an energy efficient manner. [Means for solving the problem]

[0010] The present invention solves each problem by a method according to claim 1, a computer readable medium according to claim 15, and a system according to claim 16.

[0011] In particular, the problem is how to provide a response to a client request, i.e., an image, text, or verbal request, a) receiving a client request over an interface from a client device; b) determining a client identity of the user and / or client device issuing the request; c) processing the client request and identifying and / or selecting at least one content source providing (relevant) content, in particular paywall content, to respond to the client request; d) issuing at least one secondary request to the identified / selected content source to provide the content; e) receiving content from the identified / selected content source in response to the secondary request; f) inputting at least a portion of the received content and at least a portion of the client request into at least one training model, in particular an autoregressive language model, preferably a deep learning model; g) transmitting at least a portion of the output of the trained model to a client device and providing a response to the client request; This is solved by a method comprising:

[0012] Optionally, the method comprises: h) using the client identification to allocate an amount due for sending the response, preferably without requiring simultaneous payment of the amount; It may further include:

[0013] A request from a client can include images, text, and / or audio / video data. Correspondingly, a response can be text, images, or audio / video data. Incoming client requests can be, for example, client requests generated via a web interface, such as is common in chat programs, and via an AI prompt keyboard key (e.g., a Copilot key) implemented by Microsoft Corp. via a Windows-based computer keyboard. Alternatively, each request can be generated by any other means (e.g., word processing software such as Microsoft Word, a drawing program, software running in a car, or any other type of device).

[0014] According to the present invention, a client request can originate from any type of client component (executing software or hardware). A client device is merely an example and is used as such in the following text.

[0015] Step c) can be a processing step (e.g., a data extraction and / or pattern matching step). In one embodiment, the client request can already include, for example, a descriptive name of a particular data source ("Please summarize the title page of the New York Times"). Alternatively, specific keywords can be extracted and used, for example, in a search request (e.g., a Google search) to determine the appropriate data source. The client request can, for example, relate to statistics (e.g., birth rate in Germany). According to the inventive method, "birth rate Germany" can be extracted and a Google search can be issued for data sources that provide information on each birth rate, for example, a search can be issued for "top data sources Germany birth rate".

[0016] In one embodiment, step c) is followed by the steps of c1) providing pricing from the content source to the client component, and c2) allowing pricing by the client component.

[0017] In one embodiment, the pricing in steps c1) and c2) is zero, and the user acknowledges that providing the service is worth a certain amount of money that can be used to compensate for or replace the paywalled content.

[0018] In one embodiment, step c3) is performed to bypass the paywall of the content server, whereby access is granted to restricted information, improving response quality and contributing to efficient processing.

[0019] In one embodiment, step f) may include issuing a primary request to the training model, requesting the model to summarize the content or portions of the content received from the identified / selected content sources.

[0020] Instead, the primary request can instruct the training model to use the received content to enhance the response to the client request. For example, suppose the client request was "Tell me about cloning," then the primary request could be "Taking into consideration the following latest article of cloning ([Text of the article]) explain cloning."

[0021] Alternatively, an autoregressive language model can be used to extract search terms and / or already suggest content sources. Staying within the context of the example already given of German birth rates, the autoregressive language model can be asked to "provide a list of top content sources to respond to this client request" and / or "provide search terms that can be used to search for related content sources." Of course, any of the above approaches can be limited to so-called paywalled content.

[0022] In one embodiment, step f) is followed by step f1), ie, generating an output by the trained model based on the input of step f).

[0023] In general, one object of the present invention is to receive client requests, use each client request (at least in part) to identify / select a content source, select at least one relevant content source, and communicate with each content source to receive additional data / content. The received content can then be used / summarized by a trained model (e.g., an autoregressive language model) to generate a response.

[0024] Each approach allows access to data that is not part of the training process for the training model, for example because the data is a) copyrighted, b) requires payment, or c) is closer than the most recent training stage performed on the training model, and this data can be used to provide the most accurate response to the publishing request.

[0025] The allocation of the amount to be paid in step h) is optional. An alternative to such a payment may be to request the relevant client component (e.g., client device), the user or any other entity to perform a specific service (e.g., watch a video stream, listen to an audio stream, fill out a form, provide information (e.g., personal information), and / or take any other action suitable to compensate for the received response). According to the invention, each compensation action may be performed before step d).

[0026] In one embodiment, an amount is allocated to be paid for sending a response. Each amount can be dedicated to a service provided via the training model and / or content received from a content source. In one embodiment, the allocated amount includes both the service and the content.

[0027] In one embodiment, the secondary request is for receiving content, particularly content related to the client request.

[0028] In one embodiment, the method preferably comprises, after step d), d1) determining a cost of receiving a response to the secondary request based at least in part on information from the content source; d2-1) allocating the amount in step d1) by interacting with a (micro)payment system; or d2-2) a step (270) of allocating the amount in step d1) to a chat system (20) including at least one training model (22) at least in part by using a (micro)payment system (30); Includes.

[0029] According to the present invention, content sources can be used to determine allocation amounts, for example, requests via a (standardized) API can be issued to content sources requesting costs. Returning to the New York Times example, a (computer-implemented) method may request the cost of receiving the latest title page from the content source / service providing each piece of content (e.g., the online publication of The New York Times).

[0030] In one embodiment, the method includes a step d1a) of sending a cost indication to a client component based on the determined cost, wherein the assignment to the client identification (ID) according to step d2-1) is only made upon receipt of an authorization signal, the authorization signal indicating that a user of the client device is open to assigning an amount correlated to the cost indication receiving a response to the client request.

[0031] The authorization signal may also be an authorization message. In one embodiment, the method includes querying the client device and / or a user of the client device whether access to the paid content is acceptable and whether the client / client device is willing to pay the respective costs.

[0032] Determining the client identity is necessary to link the allocation amount to a specific user (e.g., a participant in a chat and / or a client component). In the present invention, it is not necessary to identify a person / device, as long as there is another mechanism (e.g., via hardware and / or software) that links each client or client identity. Furthermore, it is not necessary to receive much information from a specific user (e.g., via a registration procedure). Any type of hardware, such as a MAC address (Media Access Control) and / or a processor identification number and / or a hard disk identification number and / or an IP address and / or other unique device number (e.g., a unique device identifier (UDID) for a smartphone), can be used to identify a client device and / or user. Furthermore, modern communication protocols provide access to mechanisms that can identify a user and / or client device. Such mechanisms can also be used to arrive at the client identity. The client identity can be any type of number or character and is not necessarily unique to a single device and / or a single user. Some methods that can be used to establish a client identity in accordance with the inventive concept are described in WO 2021 / 259608 A1, which is incorporated herein by reference.

[0033] According to one embodiment, the content received in step e) is received after the user agrees to the cost of the paywalled content.

[0034] As mentioned above, in one embodiment, the generation of a response is linked to the allocation of an amount due. The amount does not have to be paid immediately. The debt is simply noted, allowing for immediate progress of the case; for example, a response can be received without making a payment. This allows for convenient, time- and energy-efficient access.

[0035] In one embodiment, payment is only requested when the total amount reaches / exceeds a certain / predetermined threshold / amount and / or has not been paid for an extended period of time (e.g., within one month or two weeks).

[0036] By combining the concept of micropayments and / or partial payments with chatbot technology, a highly efficient way of generating responses is achieved. Micropayments or partial payments do not constitute a significant difficulty in using the offered service, while eliminating the required amount and reducing the load on the server implementing the method. Furthermore, the payment system can be used to settle the cost of paywalled content (immediately or later).

[0037] In one embodiment, the (predetermined) threshold amount is variable, in particular increasing after payment is received from a client ID and / or payment information from a client ID is obtained (optionally verified).

[0038] This allows for increased use of the method, and the positive effects of the present invention can be added to a wider range of uses.

[0039] According to the present invention, a partial payment can be defined as a payment where the amount to be paid is a fraction of the smallest physical unit available as official currency (e.g., a quarter of a 1-cent euro coin), although digital currencies such as Bitcoin may also be used.

[0040] In another embodiment, when step d2-2) is performed, one or more allocation amounts are allocated to the client identification from the chat system.

[0041] Optionally, if multiple allocated amounts are allocated to the chat system, the cumulative amount may be allocated to the client identity. Optionally, the client may settle all or part of the amount allocated to the client identity.

[0042] In another embodiment, before step d), the following steps are performed: in step d01), an index of content is obtained based on the client request, the index identifying paywalled content of at least one content source; in step d02), costs and payment terms of the at least one content source for a deferred payment arrangement for obtaining the content are determined; and in step d03), selection of paywalled content of the at least one content source is confirmed based on the associated payment terms of the at least one content source.

[0043] The selection may be confirmed by at least one training model. Optionally, the at least one training model may select content to satisfy the client request based on the likelihood of information retrieved in the index of content. Optionally, some or all of the content may be selected.

[0044] In one embodiment, the method may include a prediction step for determining the cost of at least one of the responses issued by the trained model. A payment system may be used to compensate for the use of the trained model and the content as provided by the at least one content server. The prediction may take into account one of the following parameters: Estimated power for determining response Power requirements for determining response Estimated computational power to at least partially determine response The computational power and processing time required to at least partially determine the response ·Request priority The amount of content source that should be used to generate the response to the request

[0045] In one embodiment, the prediction is (further) dependent on the load of the server that houses the training model. For example, there can be a weighting factor that causes an increase or decrease in the computational cost. For example, if the load is higher than average, the cost increases by 20% (weighting factor = 1.20). If the load is lower than average, the cost decreases by 10% (weighting factor = 0.90). This allows the inventive method to help distribute the load on the servers evenly over time, helping to avoid load peaks and establish a configuration with a better average workload.

[0046] When estimating power and / or computational power, the method can use a suitable model that takes into account the power consumption and / or computational power required to respond to previous requests and / or generate previous responses. Again, a training model can be used.

[0047] In one embodiment, the method comprises: n) issuing at least one invitation message to the client device offering a reward for the feedback in the response given in step g); o) receiving a feedback message from the client device in response to the message sent in step g); p) using the feedback messages to train and / or customize and / or otherwise inform the training model; q) optionally reducing the allocation amount due in response to receiving the feedback message; may include:

[0048] In one embodiment, the reward offer for feedback is based on a fixed amount (eg, 5 cents, 10 cents, or 1 euro).

[0049] In one embodiment, the rewards offered may be linked to a number of questions that the user is willing to answer. In other words, the user may offer a service in exchange for access to content.

[0050] Similarly, the invite message can describe an algorithm for calculating the reward or provide a specific value. Alternatively, the invite message can simply state that there is a reward, and once feedback is received, the reward will be calculated. This would involve reducing the allocated amount for a particular user and / or client device in response to receiving the feedback message. Again, the amount could be calculated at the time of reduction, or a flat fee could be reduced.

[0051] This allows the micropayment and / or partial payment system to generate incentives to improve the training model. Furthermore, incentives can be designed to gather feedback on the data most needed to improve the training model.

[0052] In one embodiment, the method includes determining a quality of the feedback message, each quality being described by a quality index (e.g., a numerical value).

[0053] In one embodiment, depending on the quality of the feedback, it is determined whether an existing training model is used to train and provide feedback. In one (another) embodiment, the reward, i.e., the allocation amount, is reduced only if the feedback as provided via the feedback message meets certain quality criteria (e.g., the quality index exceeds a predetermined threshold).

[0054] In one embodiment, the method includes issuing a search request to a database, in particular a relational database, where the search request is based at least in part on a client request. Each relational database can be a private or public database. The search request can be directed directly to content or a content source. In one embodiment, when the search request is directed to content, the response may also indicate the content source (e.g., a direct link to a particular article and / or a link to a server containing the content).

[0055] Furthermore, the above-mentioned problem is solved by a computer-readable medium having instructions which, when executed by at least one processor, perform at least one of the above-mentioned methods, and achieve the same advantages as those mentioned above.

[0056] Furthermore, the problem is solved by a system having a computer readable medium as described above.

[0057] Furthermore, the problem is a system for conducting (online) chat conversations via text and / or voice messages, a chat application adapted to provide at least one first participant in a chat conversation and to determine and output responses to questions posed by at least one second participant in the chat conversation; at least one training model, in particular an autoregressive language model, preferably a deep learning model, configured to receive questions posed by at least one second participant, determine a response, and output the response to the chat application; Including, The chat application a) issuing a secondary request to a content source and receiving the content; b) inputting the received content and questions at least in part into a training model; It is suitable for -Payment applications are - storing at least one client identification (ID) to identify at least one second participant and / or a client device used by at least one second participant; - using the client identification to assign an amount due to a response output by the chat application, preferably without requiring the simultaneous payment of the amount; - monitoring the total allocated amounts for a client identification, in particular the sum of several of these amounts; -send a payment request to at least partially settle the total allocation amount allocated to the client identification when the total allocation amount exceeds a (predetermined) threshold amount; It is suitable for Resolved by the system.

[0058] All of the above components, including the chat application, the payment application and the training model, can be part of a single software component, or the software components themselves can be distributed across several computers and distributed into separate software components that can interact with each other, for example as a client-server application.

[0059] In (alternative) embodiments, a payment request can be sent not only when the total allocated amount exceeds a (predetermined) threshold amount, but also when the amount has been allocated for a period longer than the preset threshold (e.g., more than one week, more than two weeks, more than one month, more than three months, or more than six months).Of course, there may be an option to make an earlier payment.

[0060] In one embodiment, the system, and in particular the payment application, does not consider the allocated amount, but only the time frame.

[0061] Although the system is generally suitable for responding to multiple questions, it is not necessary to input several questions or parts of several questions into the training model in step b) - indeed, one question received from one second participant and some content received from the content source may be input into the training model.

[0062] In one embodiment, the system is adapted to store certificates and authenticate the system against different content sources. In other words, there can be a database or any type of memory that stores certificates that establish trust relationships between the system and possible content sources. Each trust relationship can be used to automatically receive content, for example, over an API. Having certificates as the preferred means of establishing trust relationships between the system and possible content sources does not mean that the present invention is inapplicable to content sources that require other credentials, such as a username and password. According to the present invention, any means of establishing trust relationships is feasible. However, having each certificate is the most convenient and efficient.

[0063] Alternatively, a content source may provide content without establishing a trust relationship. For example, a flag / indicator may be utilized by the content source indicating that the content source is willing to provide content to a particular and / or any LLM.

[0064] In one embodiment, the system includes a database that stores content pricing, and the chat application is adapted to use the database to provide pricing information in response to queries. In one embodiment, the database includes relationships between content and / or content sources and pricing information. For example, there can be a table that stores content sources by descriptive name, arbitrary identifier, or URI, and for each entry, there is a respective pricing indication (e.g., 10 cents) per search page or 20 pages per request. Of course, the relationships can be of any complexity (e.g., categories that can distinguish particular content at particular content sources).

[0065] Alternatively and / or additionally, the training model itself may be used to perform the pricing task. In one embodiment, the training model may be called upon to provide a cost indication for responding to a particular request, optionally taking into account the client, client components, content source, hardware requirements, time of the request, and / or any other available information.

[0066] According to another embodiment, the above-mentioned problem is also solved by use of a chat system having at least one training model that responds to questions posed in the chat system for accessing paywalled content. The chat system may include an access model / interface for searching for paywalled content, particularly in textual form, and may use the searched paywalled content as part of the input to the training model that responds to client requests.

[0067] In other words, the chat system forms a gateway to access the paywalled content and can use the latest technology in processing languages ​​to summarize the paywalled content and generate appropriate responses to questions asked by the user. Each approach has the same advantages as those described above. In particular, the paywalled content can be accessed as needed, and for each access, the user and / or the chat system may individually decide whether accessing the paywalled content is effective or not and reinforce a given possible response. The access interface may be an application programming interface (API) that allows the transfer of all necessary data, or a crawler that (directly) accesses the content source and processes all necessary data.

[0068] Further embodiments are set forth in the dependent claims and at the end of the specification.

[0069] The invention will now be explained in more detail using some exemplary embodiments and with reference to the drawings. [Brief explanation of the drawings]

[0070] [Figure 1] Shown are a client device, a chat system, a payment system and two content servers connected via the internet. [Figure 2] 2 shows the components of the payment system according to FIG. 1; [Figure 3] The components of the chat system according to FIG. [Figure 4] 2 shows some components of one of the content servers according to FIG. 1; [Figure 5] 1 shows a first portion of an illustration of a method for providing a response to a client request. [Figure 6] 1 shows a second portion of an illustration of how a response is provided to a client request. [Figure 7] 3 shows an exemplary data structure for the payment system according to FIG. 2; DETAILED DESCRIPTION OF THE INVENTION

[0071] 1 illustrates an exemplary system according to the present invention. A client component, such as a client device 10 (e.g., a laptop, personal computer, or mobile terminal), connects to a chat system 20 via a network (in this example, the Internet 1). However, the system is not limited to a single client device 10; the system may operate with multiple client components. Furthermore, the chat system 20 and the client device 10 are communicatively coupled via the Internet 1 to a payment system 30 (e.g., a payment service provider) that facilitates micropayments. Typically, multiple content servers 50, 50' may be connected to the Internet 1 and communicate with the chat system 20 via a machine-to-machine interface (in one embodiment, via a standardized API).

[0072] Chat system 20 can include chat application 21 (FIG. 3), which can be a software program that allows users / participants to communicate with each other in real-time / near-simultaneous fashion.

[0073] In one embodiment, the chatbot is designed to perform sophisticated tasks, such as writing an article comparing different birth rates.

[0074] The chat application 21 can use a training model 22 (FIG. 3) to generate responses to specific questions provided by the request. In one embodiment, the training model 22 is a large-scale language model, e.g., a variant of a GPT (Generative Pre-Trained Transformers) model. Initial training may be performed by training the application 23, which trains the model 22 on a large amount of text data. Optionally, additional training is performed after the initial training. For additional training, content from a content server may be used. The chat application 21 is suitable for use in various natural language processing (NLP) tasks (e.g., text generation, language translation, and question answering). Examples of the chat application 21 may be, for example, ChatGPT or Dall E, available from OpenAI, LLC, or Copilot, available from Microsoft Corp.

[0075] In one embodiment, chat application 21 is adapted to generate coherent, fluent text in a variety of styles and formats. Chat application 21 can generate anything from creative writing to technical documentation and can further mimic different writing styles and / or voices. In accordance with the present invention, chat application 21 is adapted to provide responses taking into account / using paywalled content.

[0076] In one embodiment, the chat application 21 is adapted to understand and respond to context. As the text data includes a variety of topics and styles, the training model 22 can be trained with a large amount of text data to understand the context of a given input and generate an appropriate response. This forms a powerful tool for tasks such as question answering. However, this also allows the training model 22 to include additional data sources and provide appropriate or high-quality responses.

[0077] In one embodiment, the chat application 21 uses other technologies (e.g., integrated speech recognition and text-to-speech systems) to create more advanced interactive applications (e.g., voice assistants (e.g., Siri, Alexa, Google Assistant)).

[0078] The front end of the chat application 21 can take many different forms depending on the application and platform being used. In one embodiment, the front end of the chat application 21 is a web-based interface that allows users to enter text into text boxes and receive output in separate text areas.

[0079] The front end of the chat application 21 can be configured using different technologies (e.g., HTML, CSS, and JavaScript) to create an interactive and responsive web-based interface.

[0080] In one embodiment, the training model 22 is trained on a large amount of text data, meaning that the training model 22 has a large number of parameters. In one embodiment, the training model 22 may have approximately 100 billion parameters. Obviously, the larger the training model 22, the more computational power is required to process the input and generate a response.

[0081] The chat system 20 of FIG. 3 includes a content map device 25, an authentication application 26, a content source database 28, and a payment instruction application 29.

[0082] The payment instruction application 29 is suitable for establishing a connection between a (bank) account of the user and another (bank) account, which may be an account associated with a content source, and / or associated with the chat system 20, and / or associated with the payment system 30, and / or associated with a server. In other words, the payment instruction application 29 is configured to settle costs.

[0083] The chat application 21 of the invention is adapted to extract keywords from the request / question using a content mapper 25. In one embodiment, the keyword can be "birth rate" (see the example title above regarding German birth rate). The content source database 28 is then searched to identify / select content sources that have content related to the term "birth rate." The content source database 28 can be, for example, a relational database that can simply be queried by using an SQL statement (e.g., "SELECT SOURCES FROM CONTENT SOURCE DATABASE WHERE TEXT LIKE '%BIRTH RATE%'").

[0084] After identifying / selecting the appropriate content source, the chat application 21 may contact the respective content server (e.g., content server 50). When accessing the content server 50, the content server 50 may require authentication. For example, each authentication may be handled by an authentication application 26, which may maintain a set of certificates that establish a trust relationship between the content server 50 and the chat system 20. After establishing such a trust relationship, the chat application may connect to the indexing application 51 of the content server 50 to identify / select articles that contain, for example, the term "birth rate." Essentially, the chat application 21 may use the indexing application 51 to identify / select relevant content to respond to an incoming request. Once the content is identified / selected, the content may be extracted from a database (e.g., a relational database 53 containing the content). Alternatively, the indexing application 51 may point to a file system 55 containing the content, for example, in the form of an MS Word or text file.

[0085] Once the content is identified / selected, the chat application 21 can attempt to retrieve the respective content. However, the respective content may be paywalled, meaning that the originator of the respective content requires a specific amount to grant access to the content. In this regard, throughout the specification, the terms (amount to be paid) and cost are used interchangeably. To make the respective payments, the chat application 21 can interact with a payment system 30, which subsequently contacts a payment gateway 57 of the content server. Access to the content server 50 can be granted via a (standardized) API. Alternatively, the content server 50 provides a web interface 58 that can be used to search for and / or access the particular content. In fact, the web interface 58 can be used to make the respective payments.

[0086] It may be an aspect of the present invention for the chat system 20 to use a payment system 30 to receive compensation for submitted responses.

[0087] The payment system 30 includes an identification device 31 (FIG. 2), an interface device 32 capable of communicating with the chat system 20 and / or the client device 10, a memory device 33, and a processing device 34. The payment system 30 is a digital payment platform that allows users to purchase any type of digital goods and services in a flexible and convenient manner. The payment system 30 allows users to pay for digital content (e.g., online articles, e-books, music, and video games) without the need to enter the user's credit card details each time a purchase is made.

[0088] In one embodiment, the payment system 30 operates by allowing a user to create a potentially anonymous account using any payment information, such as a credit card number, and then pre-authorize / allocate a specific amount that can later be used to make purchases. This pre-authorized / allocated amount can be settled at a later stage using a credit card or any other payment method, or by providing services from the user equal to the specific amount. This allows the payment system 30 to greatly facilitate small incremental payments without having these amounts immediately debited to the preferred payment method.

[0089] In one embodiment, the payment system 30 is suitable for making purchases on any website that is integrated with the payment system 30. Authorization can be given by clicking a "Put it on my tab" button or link that assigns the amount to be paid to the user's client identification. Some embodiments of the payment system 30 that can be used are described in EP 2476087 B1, which is incorporated herein by reference.

[0090] The payment system can be a digital payment platform that allows users to purchase digital goods and services in a flexible and convenient way without having to enter credit card details each time. Users can pre-authorize a specific amount that can be used later for purchases and try out digital goods and services before committing to a purchase. The payment system may also provide various tools for merchants to integrate the platform into their e-commerce systems.

[0091] 2 illustrates possible individual components of a payment system 30. The payment system 30 according to one embodiment of the present invention comprises an identification device 31 for recording at least one identification number of a client device 10 or user, an interface device 32 for receiving and validating automatic debit instructions from the chat system 20 or any other merchant system, where the automatic debit instructions contain information regarding the amount (cost) due to be paid to the chat system 20 or any other system, a memory device 33 for storing the allocated amount together with the associated identification number ID, and a processing device 34 for processing input requests.

[0092] In one embodiment, the payment system 30 is adapted to identify client components, i.e., client devices 10, solely based on their MAC addresses. Accordingly, the memory device 33 stores the amounts to be paid in conjunction with the corresponding MAC addresses. For this purpose, the payment system 30 includes a corresponding database that holds a corresponding table. An exemplary excerpt from a table held in the database is shown in FIG. 7. This table includes, for example, three columns: a first column containing the identification number of the client device 10 or user, a second column containing the amount to be debited, and a third column containing the date on which the automatic debit instruction was received by the payment system 30. Each row of the table in FIG. 7 corresponds to an automatic debit instruction. Thus, it can be seen from the table in FIG. 7 that on July 1, 2009, 20 euro cents were debited / allocated to identification number 222. Furthermore, on September 20, 2009, 5 euro cents were debited to the same number. The IDs shown in the table are for illustrative purposes only. In one embodiment, the IDs can be MAC addresses. In other embodiments, the ID can point to another table that establishes the identity of a particular user or device. Alternatively, the ID can link each entity (device or person) to another system, for example, by holding a Facebook or Google ID.

[0093] Using these items, the processing device 34 can determine the total amount payable from the debit amount (allocated amount) for a particular identification number ID. For example, the total amount payable for identification number 222 amounts to 25 euro cents.

[0094] Thus, the payment system 30 can be configured such that a particular user is required to settle a user's debt if the user's debt is greater than, for example, 0.29 euros or 1 euro or 10 euros.

[0095] As described, the payment system 30 can be used not only to make payments or allocate possible payments between the client device 10 or each user and the chat system, but also between the chat system 20 and one of the content servers 50, 50′, where payments or payment allocations can be made directly by / to the chat system 20, or routed so that payments to the content server originate from the user or client device 10.

[0096] One embodiment method for implementing a system as described above may include the following steps. 1. The content server 50 opens a merchant / merchant account. 2. The content server 50 issues an LLM access certificate (CA). 3. The user or client device 10 queries the chat application 21 requesting data behind the paywall of the content server 50. 4. The chat application 21 requests access to the content on the content server 50 using the access credential. 5. If the access credentials are verified, the content server 50 allows the chat application 21 to access the data behind the paywall. 6. Negotiate a price for access, for example using pricing application 36. 7. Call the payment system 30 and request the client device 10 to acknowledge that a response to the issued question / inquiry will cost a particular amount, for example. 8. The user or client device 10 accepts the pricing. 9. The payment system 30 allocates the respective amounts. 10. The payment system 30 sends the entitlement token to the chat application 21. 11. The chat application 21 receives final access to the content on the content server 50 using the entitlement token. 12. Once the allocated amount reaches a certain threshold, the payment system 30 requests payment from the user or client device 10. 13. Once payment is received, the content server 50 is compensated.

[0097] 5 and 6 show the method / process in more detail. In step 200, the chat system 20 (more precisely, the chat application 21) receives a client request from the client device 10 over the interface. In one embodiment, the request indicates that the chat application 21 should use one of the content servers 50 or 50′ to respond to the request (right branch in FIG. 5; step 220). Alternatively, the chat application 21 itself can determine / suggest including / selecting one of the content servers 50, 50′ to improve the possible response (left branch in FIG. 5; step 210). In other words, the chat application 21 is adapted to first check whether the client request includes any indication regarding the content server 50, 50′ to be used, and if the client request does not include any indication, attempts to identify / select an appropriate suggestion / content server (left branch in FIG. 5; step 212). In other words, the chat system 20 can implement both branches and dynamically decide which branch to use.

[0098] 5 begins with receiving a client request that does not indicate a particular content server 50 or 50' to use to respond to the client request (step 210). In one embodiment, the received request may include the question, "Please tell me more about Jimmy Buffett."

[0099] In step 212, chat application 21 initiates a Google search for "Jimmy Buffett" and then selects an appropriate resource, for example, content server 50 accessible via www.creem.com, to refine the quality of possible responses. The selection algorithm can be implemented based on many criteria. In a very simple embodiment, chat application 21 simply selects the results that are highest ranked by Google.

[0100] In step 214, the chat application 21 checks whether the content server 50 has a Certificate Authority (CA) certificate. A CA is a trusted authority that issues digital certificates for websites and other entities. The CA verifies the website domain and website ownership, and then issues a TLS / SSL certificate that is trusted by the client.

[0101] In that case, the chat application 21 notifies the user / client device 10 that the content server 50 appears to be an authority on the subject matter and is trustworthy, and asks whether content originating from that content server 50 should be used in the response. If the user / client device 10 declines, or the content server 50 does not have a CA certificate, the chat application 21 generates a generic response (step 218) without requiring additional measures.

[0102] If, at step 216, the user / client device 10 indicates that content from the content server 50 should be used, processing continues to step 230 (see also FIG. 6). Alternatively, if, at step 216, the user / client device 10 indicates that content from the content server 50 should not be used, processing continues to step 218, generating a generic response without the need for additional measures.

[0103] As mentioned above, the right branch of Figure 5 includes an embodiment / workflow in which the user / client device 10 indicates that a particular content server 50 should be requested. Each request received in step 220 may include the following phrase: "Using creem.com, please tell me more about Jimmy Buffett." Similar to step 214, in step 222, the chat application 21 may immediately check whether the content server 50 has a CA certificate. If the respective certificate is available, processing may continue with step 230.

[0104] In step 230 (FIG. 6), the chat application 21 may check whether the content server 50 specifies that access to its content is restricted. For example, in one embodiment, the content server 50 may require payment, e.g., per access, per download page, or per session. Alternatively, or in addition, the chat application 21 may check whether any other access rules apply. In one embodiment, the content server 50 provides the information via metatags that are part of the content server's 50 initial web page. If the content server 50 does not require paid access, the chat application 21 may obtain the content of one or more web pages, e.g., for free, and use the received content to generate a response to the initial user request (step 240, then proceed to step 273 (not shown in FIG. 6)). In other words, the method generates and transmits a response to a client request without requiring access to paywalled content. This has the positive effect of providing up-to-date responses (the output of the training model 22) even without additional training of the training model.

[0105] If the content server 50 specifies that access to the content is restricted, then upon receiving the chat application 21 or a trigger signal from the chat application 21, the payment system 30 sends a certificate to the content server 50; a) the chat system 20 and the content server 50, and / or b) Payment System 30 and Content Server 50 In one embodiment, this certificate is different from the CA certificate and is related to payments, i.e., payment certificate.

[0106] Step 250 may include validating each certificate (eg, validating the certificate of content server 50).

[0107] In step 252, the payment system 30 may be invoked and provide pricing, for example, from a data table indicating costs. To do so, the payment system 30 may include a pricing application 36. Alternatively, pricing instructions may be requested from each content server 50, or the chat system 20 knows the costs because it maintains its own database with pricing information for different content servers 50 and / or specific content. In other words, the cost of accessing a content server is provided and / or determined by the content server 50.

[0108] Various deferred payment arrangements may be considered by the chat system 20 and the content server 50 depending on the characteristics of the requested content and the user of the client device 10. For example, if a user is a new user lacking credentials provided to the chat system 20 or payment system 30, the chat system 20 and the content server 50 may require immediate payment for paywalled content above a predetermined price threshold and / or agree to a short-term deferred payment option ("tab") (e.g., $5.00 or less) that applies to purchases of content priced at $1.00 or less. Additionally, the chat system 20 and the content server 50 may negotiate which party will bear the risk of non-payment of the tab by the client device 10. If the content server 50 agrees to bear the risk of non-payment associated with a user device, for example, the content server 50 may require, for example, that the chat system 20 and / or payment system 30 maintain a valid credit source (e.g., a credit card) for the user device on record and / or that the client or client device meets predetermined "creditworthiness" criteria based on a credit history. Additionally, other variations of these procedures should be considered aspects of this disclosure; for example, a client device may use a hybrid approach whereby some payments are made immediately and the remainder are deferred.

[0109] Additionally, the deferred payment arrangement may define additional features (e.g., payment schedules). For example, the payment system 30 may agree to accumulate payments from the user device 10 to the content server 50 and pay the content server 50 of received user payments on a predetermined schedule (e.g., once per day) or based on accumulated amounts (e.g., for accumulated payments totaling more than $10.00). By accumulating payments held by the content server 50 to a certain extent before paying the content server 50, the number of payment transactions and associated transaction costs can be reduced.

[0110] In step 254, the chat application 21 may present the determined cost for requiring the content server 50 to the user / client component and request permission.

[0111] If the user / client component cannot agree to these costs, a generic response may be returned in step 256 without involving the content server 50. Alternatively, if the user / client component agrees to the (additional) costs, the payment system 30 must allocate each cost to each user / client component. Allocation may be done by placing each amount in a virtual tab associated with the user / client component by client identification.

[0112] In that alternative scenario, the payment system 30 may pass the entitlement token to the chat application 21, which can bypass the paywall and receive content from the content server 50, for example, by obtaining the site of the response to the original user prompt / client request that is the provided content.

[0113] After step 254, and after the user answers "yes" (the rightmost branch in FIG. 6), processing continues at step 270 by assigning the cost of accessing the content server to the client identification ID by using payment system 30. In another embodiment, chat system 20 may be configured to assign the cost of accessing the content server to the client identification ID. In yet another embodiment, the content source may directly assign the cost of accessing the content server to the client identification ID.

[0114] After step 270, the paywall of the content server is bypassed and the content is received from the content server in step 272. In one embodiment, content from multiple content servers is obtained.

[0115] In a next step 273, at least a portion of the received content and at least a portion of the client request are input to at least one training model 22. In one embodiment, the client request is provided directly to the training model 22, which itself accesses search engines to obtain content, in particular paywalled content.

[0116] In a next step 274, a response is generated by the training model utilizing this client request and this one or more received contents and information already contained in the training model 22. In one embodiment, the user has already provided relevant additional data (e.g., family status, age, health status, hobbies, etc.) that can also be used by the training model 22 to provide a response.

[0117] The next step 276 is to send the generated response to the client device.

[0118] According to one embodiment, the chat application 21 constructs a response to the original client request using provided content from the training model 22. This has the positive effect of providing a response that is up-to-date (the output of the training model 22) even without further training of the training model. In other words, traditionally, a training model simply refers to its own training data, and the training data used to train the model has a specific date and time. The client request occurs some time after the latest training data. "Up-to-date" means that additional content generated between the training data date and the date of the client request can be implemented in the response without requiring further training of the model.

[0119] In the above-described embodiment, the user is always asked whether or not they are interested in using one of the identified / selected content servers 50, 50'. Within the scope of the present invention, the chat system 20 may utilize the content servers 50, 50' if it receives and / or requests confirmation from the user.

[0120] In the above-described embodiments, certificates are used to establish trust relationships between servers, machines and / or applications, but this trust relationship can also be established by other means, such as by simple passwords and / or login names, by physical proximity (e.g., installing applications on the same hardware), by dedicated physical or virtual connections (e.g., VPNs), etc.

[0121] In another embodiment, authentication, i.e., determining the client identity, is inherited from another device and / or application.

[0122] In another embodiment, the trust relationship between the client or client component and the content server or chat application is inherited from another device and / or application.

[0123] In another embodiment, the training model receives the client request directly and drives a search engine (e.g., Google, Bing, etc.) that, if paywalled content should be accessed, performs the steps described above to access the paywalled content (steps 250, 254, 270, 272, 274, 276).

[0124] In another embodiment, when allocating costs for sending a response, the allocation is split into two pairs, essentially having three participants: The first party is represented by the client component (or client identity). The second party is represented by the chat system 20 and / or the payment system 30 and / or the training model 22. The third party is represented by the content source (one or more content servers 50, 50'). Since there may be multiple relevant content sources, each content server may act as a separate third party. A first pair is formed between the first and second party. A second pair is formed between the second and third party. In other words, the first party is not directly connected to the third party, but both the first and third parties are connected through the second party.

[0125] When it comes to payment for content occurring between a first party and a third party, the second party acts as an intermediary. This has the advantage that the second party already has established connections to multiple third parties that are not easily available to the first party. For example, if a client request is "Plan a trip to the beach next weekend," a first content server may provide weather information, a second server may provide accommodation information, and a third server may provide information about surfing equipment rental services. All or part of the information above may be paywalled content. For example, accurate weather information may cost 10 cents, accommodation information may be free, and reliable rental information may cost 15 cents. Once the user agrees to the cost, all information is received by the second party, which provides the user with a response to their request.

[0126] Instead, payment for the content is processed by the second and third parties, thereby already settling the debt with respect to the third party. The debt of the first party to the second party remains, for example, as a loan (or tab).

[0127] Data relating to this payment information may be stored in a database, for example, as shown in Figure 7. Additionally, a receipt for one of the above-mentioned transactions between the first / second party and the third party may be stored in the database. The receipt may be given to the first party at a later time. Additionally, columns such as those shown in Figure 7 may be augmented with columns for the first and / or second and / or third parties involved in the payment. Payment for paywalled content may be performed as described in any of the above-mentioned embodiments.

[0128] In another embodiment, the three parties mentioned above are further described. When allocating costs for sending a response, the allocation is between the first party and the third party. In other words, the second party is not involved in the payment. Furthermore, the first party may allocate amounts to multiple third parties simultaneously and / or separately (e.g., via another payment provider). Again, data regarding this payment information may be stored in a database, for example, as shown in FIG. 7. Furthermore, columns such as those shown in FIG. 7 may be augmented by columns for the first party and / or second party and / or third party involved in the payment.

[0129] According to another embodiment, if a user (first party) promises to pay but ultimately does not, the payment from the second party to the third party may be withheld. In other words, the payment from the second party to the third party is only executed after the first party pays the second party. Both transactions may be stored in the same or different databases. The payment data stored between the second party and the third party may not be paid by the user, but may be paid per time frame (e.g., monthly, weekly, or daily). This trust-based relationship ensures that paywalled content can be retrieved and a complete response can be given to the client, thereby increasing energy efficiency.

[0130] However, in another embodiment, in accordance with the other embodiments described above, the method further includes obtaining payment information from the client ID before providing a response (e.g., before performing any step of providing a response). Obtaining payment information does not necessarily mean that payment is required. Rather, this step focuses on the willingness to pay for future responses. For example, after receiving the payment information, a first credit amount, e.g., $1, $3, or $5, is granted to the client component. The first credit amount may be used to pay content providers for content that provides responses to one or more client requests. After the first credit amount is exhausted, a first request for payment is sent to the client ID. Upon fulfilling the first request and making payment, a second credit amount (e.g., $5, $10, or $20) preferably greater than the first credit amount may be granted to the client ID. Again, after the second credit amount is exhausted, a second request for payment is sent to the client ID. If the second request is satisfied and payment is made, a third loan amount (e.g., $10, $20, or $50) greater than the second loan amount is preferably granted to the client ID. In other words, the loan amount is granted after obtaining and optionally verifying payment information from the client ID. Furthermore, the loan amount may be variable and may increase after receiving payment from the client ID. Optionally, the loan amount may increase after every second or third payment, or may increase sharply at an initial time and then remain constant at a higher level.

[0131] In summary, all of the above-described embodiments of the present invention enable substantial energy savings by providing up-to-date responses directly to user queries. In particular, when using training models, the inventive method and system enable the use of pre-trained models that do not require immediate training updates, thereby saving valuable energy for training new models and still providing up-to-date responses to user queries. Furthermore, by providing a unified access environment for paywalled content where relevant information can be quickly searched, client component usage time and server usage time are reduced, thereby facilitating economical use of energy resources.

[0132] In the above-described embodiment, Google may be used as a search engine to identify / select an appropriate content source. However, in the present invention, any available search engine may be used, such as Bing, Yahoo, Yandex, etc. Instead, the present invention may include any private or public search engine. Furthermore, it is not necessary to first identify / select a server and then search the server for relevant content. In accordance with the present invention, for example, any public or private index server may be used to locate content. In the embodiment, after identifying / selecting content (e.g., the front page of the New York Times), the availability of a particular certificate or trust relationship may be checked.

[0133] In the above-described embodiment, a single content server 50 was selected to improve the quality of the response and / or generate the response. According to the present invention, a response can be constructed using several (different) content servers 50, 50' and / or by requesting several documents from one or more content servers 50, 50'. According to the present invention, information (e.g., trustworthiness) retrieved from different content sources and / or fee content can be aggregated by comparing the received information.

[0134] In many of the described embodiments, payments are allocated. Of course, in accordance with the present invention, a system (e.g., payment system 30) can also request and / or effect instant payments, for example, via credit or debit cards, or via any other currency (e.g., digital currency such as Bitcoin).

[0135] Furthermore, in accordance with the present invention, the payment system 30 may provide a prepaid wallet containing pre-filled money so that payments can be made. Alternatively, the wallet may be postpaid, meaning that payments are requested once a certain threshold is reached.

[0136] Of course, it is within the scope of the present invention to receive a payment (e.g., $1) whereby only a small amount (e.g., 1 cent) can be spent immediately, with the remaining amount (e.g., 99 cents) being deposited and available for use at a later time.

[0137] Furthermore, in at least one of the above-described embodiments, the payment system 30 may act as a money distributor, allowing a user to make a single payment, with the payment system 30 distributing money among the associated entities.

[0138] In the above-described embodiments, the client device includes hardware separate from the hardware that houses the LLM. In accordance with the present invention, the LLM may be housed at least partially on the client device itself. In this scenario, establishing a trust relationship between the client device and one or more content servers is an option. However, establishing a trust relationship, for example, by using certificates, is optional.

[0139] It should be noted here that all of the above is claimed to be relevant to the invention, especially when considered in isolation and in any combination of the details shown in the drawings.

[0140] According to another embodiment and a first aspect, an artificial intelligence (“AI”) application server implementation method includes: a) receiving a user prompt over an interface from a user device; b) determining a user identification based on at least one of the user and the user device issuing the prompt; c) selecting at least one content server that processes the user prompt and provides particular digital content usable for generating a response to the user prompt; d) sending a content request to the selected content server and providing the particular digital content; e) receiving the particular digital content from the selected content server in response to the content request; f) processing at least a portion of the received particular digital content and at least a portion of the user prompt by at least one AI training model associated with the AI ​​application server to generate processing information; g) sending response information to the user device in response to the user prompt based at least in part on the processing information; and h) intermittently processing, by a training AI application associated with the AI ​​application server, at least a portion of the user identification and at least one of the received particular digital content, the output information, and the user prompt to update the at least one training AI model.

[0141] In the method according to the first aspect, according to the second aspect, the step of intermittently processing by a training AI application associated with an AI application server adaptively updates at least one AI training model with respect to user identification.

[0142] In the method according to the second aspect, according to a third aspect, the step of intermittently processing by a training AI application associated with the AI ​​application server includes adaptively updating at least one AI training model and processing at least a portion of the at least one AI training model.

[0143] In the method according to the second aspect, according to a fourth aspect, the AI ​​application server includes a plurality of processors.

[0144] In the method according to the fourth aspect, according to a fifth aspect, the step of intermittently processing by a training AI application associated with the AI ​​application server adaptively updates at least one AI training model that is partially executed by at least one processor other than the processor executing step f).

[0145] In the method according to the first aspect, according to a sixth aspect, the method further includes a step of: i) determining a cost of transmitting response information by at least one of a user and a user device associated with the user identification.

[0146] In the method according to the sixth aspect, according to a seventh aspect, step i) is performed without simultaneously requiring payment of an amount by at least one of the user and the user device associated with the user identification.

[0147] In the method according to the sixth aspect, according to an eighth aspect, the method further comprises the steps of: d1) determining the cost of receiving a response to the content request based at least in part on information from the content server; and d2-1) allocating the amount in step i) by at least in part using a (micro)payment system.

[0148] In the method according to the eighth aspect, according to the ninth aspect, the method includes a step d1a) of sending a cost indication to a user device based on the determined cost, wherein the allocation according to step d2-1) is made only upon receipt of an authorization signal indicating that at least one of the user and the user device authorizes the allocation of an amount correlated to the cost indication receiving a response to the client request.

[0149] In the method according to the eighth aspect, according to the tenth aspect, the step of determining the cost of receiving a response to the content request further includes retrieving content pricing from a database, such content pricing being based on at least one of the content, pricing for the content server, and a predetermined relationship between the content server and at least one of an account associated with the content source AI application server and the user identification.

[0150] In the method according to the eighth aspect, according to an eleventh aspect, the step of determining a cost of responding to the request includes calculating a power requirement for generating a response to the request.

[0151] In the method according to the first aspect, according to a twelfth aspect, the method further includes the steps of: a. monitoring a total allocated amount for a particular client identification; b. sending a payment request to at least partially settle the total allocated amount allocated to or incurred before the threshold time for the particular client identification if the total allocated amount exceeds at least one of the threshold amounts; and c. determining the cost of responding to the request.

[0152] In the method according to the first aspect, according to a thirteenth aspect, step c) includes issuing a search request to the database, the search request being based at least in part on the client request.

[0153] In the method according to the first aspect, according to a fourteenth aspect, step c) includes authenticating at least one of the user device or the AI ​​application server to the content server.

[0154] In the method according to the first aspect, according to a fifteenth aspect, the step of authenticating at least one of the user device or the AI ​​application server to the content server includes a step of storing an authentication certificate for the content server in each memory device associated with at least one of the user device or the AI ​​application server.

[0155] In the method according to the first aspect, according to a sixteenth aspect, step c) includes performing an internet search of at least one content server.

[0156] In the method according to the first aspect, according to a seventeenth aspect, step c) further includes utilizing a ranking model algorithm to determine at least one content server that provides the particular digital content based in part on at least one of the suitability of the digital content available from each content server, the cost associated with such digital content from each content server, and any predetermined relationship between the AI ​​application server and each content server.

[0157] In the method according to the first aspect, according to an eighteenth aspect, the at least one training model is an autoregressive language model and a deep learning model.

[0158] In the method according to the first aspect, according to a 19th aspect, the user prompt includes at least one of a question, a request, an image, an audio file, and a video file.

[0159] According to another embodiment, a method for providing a response to a client request, e.g., an image, text, comprises: a) receiving (200) the client request over an interface from a client component, in particular a client device (10); b) determining a client identification (ID) of the user and / or client component issuing the request; c) identifying at least one content source providing (relevant) content, in particular paywall content, for processing the client request and responding to the client request; d) issuing a secondary request to the identified content source and providing the content; e) receiving content from the identified content source in response to the secondary request; f) inputting at least a portion of the received content and at least a portion of the client request into at least one training model (22), in particular an autoregressive language model, preferably a deep learning model; g) transmitting at least a portion of the output of the training model (22) to the client component and providing a response to the client request; and h) optionally using the client identification (ID) to allocate an amount to be paid for the transmission of the response, preferably without requiring a simultaneous payment of an amount.

[0160] The method of this other embodiment described above further includes the steps of: d1) determining the cost of receiving a response to the secondary request, in particular by including a content source; and d2-1) allocating the amount in step d1), by including a (micro)payment system.

[0161] The method of this other embodiment described above further includes a step of: d1a) sending a cost indication to the client component based on the determined cost, wherein the allocation according to step d2-1) is made only upon receipt of an authorization signal, the authorization signal indicating that the user of the client component is open to allocating an amount correlated to the cost indication receiving a response to the client request.

[0162] The method of this other embodiment described above further comprises the steps of: a) monitoring a total allocated amount for a particular client identification (ID); b) sending a payment request to at least partially settle the total allocated amount allocated to the particular client identification (ID) when the total allocated amount exceeds a (predetermined) threshold amount; and c) determining the cost of responding to the request, preferably including calculating the power required to calculate a response to the request and / or the required computing power to at least partially calculate a response to the client request.

[0163] The method of this other embodiment above, wherein step c) further comprises issuing a search request to a database, particularly a relational database, the search request being based at least in part on the client request.

[0164] The method of this other embodiment described above, wherein step c) preferably further comprises authenticating the client component or chat system that performs step c) to the content source.

[0165] According to a further embodiment, a computer-readable medium having instructions that, when executed by at least one processor, perform the method described in this other embodiment above.

[0166] According to a further embodiment, there is provided a system for conducting (online) chat conversations via text and / or voice messages, in particular for implementing the method described in this other embodiment above, comprising: a chat application (21) adapted to provide at least one first participant of a chat conversation and to determine and output responses to questions posed by at least one second participant of the chat conversation; at least one training model (22) used by the software application for determining the response, in particular an autoregressive language model, preferably a deep learning model, the at least one training model (22) being suitable for a) issuing secondary requests to a content source and receiving content, and b) at least partially inputting the received content and questions into the training model; storing at least one client identification (ID) to identify a client component used by the second participant and / or the further participant; - suitable for assigning amounts due to responses output by the chat application using the client identification, preferably without the simultaneous need for payment of the amounts; - Monitor the total allocation amount for a specific client identification, -send a payment request to at least partially settle the total allocation amount allocated to a particular client identification when the total allocation amount exceeds a (predetermined) threshold amount; -Payment applications (30) and A system including:

[0167] A system according to this further embodiment as described above, wherein the system is adapted to store certificates and authenticate the system against different content sources.

[0168] The system of this further embodiment described above further includes a database storing content pricing, wherein the chat application is adapted to use the database to provide pricing information in response to queries, and preferably further includes a database containing relationships between content and / or content sources and pricing information.

[0169] Another embodiment includes the use of a chat system (20) having at least one training model (22), in particular an autoregressive language model, preferably a deep learning model, that responds to questions posed in the chat system to access paywalled content, wherein the chat system includes an access model that searches for paywalled content, in particular in text form, and uses the searched paywalled content as part of the input to the training model that responds to client requests.

[0170] Another embodiment includes the above use further comprising a chat system including a (micro)payment system for effecting payments to content providers of paywalled content. [Explanation of symbols]

[0171] Reference sign 1. Internet 10 client devices 20 Chat System 21. Chat Applications 22 Training Models (e.g., LLM) 23 Training Applications 25 Content Map Device 26 Authentication Applications 28 Content Source Database 29 Payment Instruction Application 30 Payment Systems 31 Identification Device 32 Interface Devices 33 Memory Devices 34 Processing Device 36 Pricing Applications 50, 50' Content Server 51 Indexing Applications 53 Relational Database with Content 55 File System 57 Payment Gateways 58 Web Interface ID identification number 200 Step 200: Receive client request 210 Step 210: Receive a request that does not have a specified content server 212 Step 212: Selecting a Content Server by Chat Application 214 Step 214: Check if the content server has a CA certificate 216 Step 216: Propose a generated response using the content server 218 Step 218: Generate a generic response 220 Step 220: Receive a request including a content server 222 Step 222: Check if the content server has a CA certificate 230 Step 230: Verify access to the content server 240 Step 240: Get content from the content server 250 Step 250: Verify the content server's certificate 252 Step 252: Give pricing 254 Step 254: Present the cost of accessing the content server 256 Step 256: Return a general response 270 Step 270: Assigning the cost of accessing the content server 272 Step 272: Bypass the paywall and receive the content 273 Step 273: Input received content and client requests into training model 274 Step 274: Generate a response using the trained model 276 Step 276: Send the generated response to the client device

Claims

1. 1. A method for providing a response to a client request, e.g., image, text, comprising: a) receiving a client request (200, 210) over an interface from a client component, in particular a client device (10); b) determining a client identification (ID) of at least one of the user and client component issuing the request; c) processing said client request and identifying / selecting (212) at least one content source providing (relevant) content, in particular paywall content, to respond to said client request; d) issuing at least one secondary request to the identified / selected content source to provide content (230); e) receiving content from the identified / selected content source in response to the secondary request (272); f) inputting (273) at least a portion of said received content and at least a portion of said client request into at least one training model (22), in particular an autoregressive language model, preferably a deep learning model; g) sending (276) at least part of the output of the trained model (22) to the client component to provide the response to the client request; h) optionally assigning an amount due to said sending of said response using said client identification (ID) (270), preferably without requiring the payment of an amount at the same time; A method comprising:

2. Preferably, after step c), c1) providing pricing from the content source to the client component (252); c2) allowing pricing by said client component; The method of claim 1 , comprising:

3. 3. The method of claim 2, wherein in steps c1) and c2), the pricing is zero and the user agrees to provide the service.

4. Preferably, after step c), c3) Bypassing the paywall of the content server (272) The method according to any one of claims 1 to 3, comprising:

5. Preferably, after step d), d1) determining (252) a cost of receiving a response to said secondary request based at least in part on information from said content source; d2-1) assigning (270) said amount in step d1) to said client identification (ID) at least in part by using a (micro)payment system (30); or d2-2) allocating (270) said amount in step d1) to a chat system (20) including said at least one training model (22) at least in part by using a (micro)payment system (30); The method according to any one of claims 1 to 4, comprising:

6. d1a) sending a cost indication to the client component based on the determined cost, wherein the assignment to the client identification (ID) according to step d2-1) is only made upon receipt of an authorization signal, the authorization signal indicating that the user of the client component is open to assigning an amount correlated to the cost indication receiving the response to the client request. The method of claim 4, comprising:

7. Preferably, after step f), f1) generating an output by said training model (22) based on said input of step f). The method according to any one of claims 1 to 6, comprising:

8. i) monitoring the total allocated amount for a particular client identification (ID), in particular the sum of a plurality of said amounts; j) sending a payment request to the client component to at least partially settle the total allocated amount allocated to the particular client identification (ID) if the total allocated amount exceeds a (predetermined) threshold amount; k) determining a cost of responding to said client request, preferably comprising calculating the power required to compute said response to said client request and / or the computational power required to at least partially compute said response to said client request; The method according to any one of claims 1 to 7, comprising:

9. The method according to any one of claims 1 to 8, wherein the (predetermined) threshold amount is variable, in particular increasing after a payment is received from the client ID.

10. The step c) Issuing a search request to a database, particularly a relational database, said search request being based at least in part on said client request. The method of any one of claims 1 to 9, further comprising:

11. The step c) Preferably, authenticating the client component or chat system (20) that performs step c) to the content source. The method of any one of claims 1 to 10, further comprising:

12. The step b) Obtaining and optionally verifying payment information from said client ID. The method of any one of claims 1 to 11, further comprising:

13. The method according to any one of claims 5 to 12, wherein when step d2-2) is performed, the one or more allocated amounts are allocated to the client identification (ID) from the chat system (20).

14. Before step d), d01) obtaining a content index based on the client request, the index identifying paywalled content of the at least one content source; d02) determining costs and payment terms of said at least one content source for a deferred payment arrangement for acquiring said content; d03) confirming the selection of the paywalled content of the at least one content source based on the associated payment terms of the at least one content source; The method according to any one of claims 1 to 13, wherein the

15. A computer readable medium having instructions which, when executed by at least one processor, perform the method of any one of claims 1 to 14.

16. A system for conducting (online) chat conversations via text and / or voice messages, in particular implementing the method according to any one of claims 1 to 14, a chat application (21) suitable for providing at least one first participant of said chat conversation and for determining and outputting responses to questions posed by at least one second participant of said chat conversation; at least one training model (22), in particular an autoregressive language model, preferably a deep learning model, configured to receive the question posed by the at least one second participant, determine the response and output the response to the chat application (21); Including, The chat application (21) a) issuing at least one secondary request to a content source and receiving content; b) inputting the received content and questions at least in part into the training model (22); Suitable for - a payment application (30) - storing at least one client identification (ID) to identify said at least one second participant and / or a client component used by said at least one second participant; - using said client identification (ID) to assign an amount due to said response output by said chat application (21), preferably without the simultaneous need for payment of an amount; - monitoring the total allocated amount against a client identification (ID), in particular the sum of a number of said amounts; - sending a payment request to at least partially settle the total allocated amount allocated to the client identification (ID) if the total allocated amount exceeds a (predetermined) threshold amount; Suitable for system.

17. 17. The system of claim 16, adapted to store certificates and to authenticate the system to different content sources.

18. a database storing content pricing, said chat application (21) being adapted to use said database to provide pricing information in response to said questions, preferably a database containing relationships between content and / or content sources and said pricing information; 18. The system of claim 16 or 17, comprising:

19. Use of a chat system (20) having at least one trained model (22), in particular an autoregressive language model, preferably a deep learning model, that responds to questions posed in the chat system (20) for accessing paywalled content, said chat system (20) including an access model / interface, in particular in text form, for retrieving said paywalled content, and using said retrieved paywalled content as part of the input to said trained model (22) that responds to said questions, in particular client requests. use.

20. 20. Use according to claim 19, comprising the chat system (20) including a (micro)payment system for effecting payments to content providers of the paywalled content.

Images