Computational training and load distribution system and related method

Abstract

In present invention, a system for training and allocating a computational load and a related method are provided. The computational load is generated by artificial intelligence systems or by a user's software and comprises computational tasks. Said system comprises a plurality of devices, each comprising at least one processing unit that performs said computational tasks in connection with a web portal with an application. Said web portal with an application controls the availability of said processing unit to execute one or more computational tasks. Said execution of a computational task and the execution of training objectives generates a profit of an amount for a user owning said device, which can be spent on services and for making queries to said artificial intelligence systems and / or requesting allocations, among the system's devices, of one or more computational loads generated by said user's software.

Description

[0001] LEIBO. / 45e2025

[0002] “Computational training and load distribution system and related method”

[0003] Description

[0004] Field of the invention

[0005] The invention relates to the field of information systems and infrastructures. More specifically, the invention aims to provide a system for sharing the workload between various processors and which is capable of promoting the training of artificial intelligences with reward mechanisms.

[0006] Prior art

[0007] In an age where artificial intelligence (Al) is transforming every sector of the global economy, computing power becomes a crucial resource. However, access to this vital resource is currently monopolized by a few tech giants, leaving many startups, small and medium-sized enterprises (SMEs), and independent researchers out of the innovation race. The cost of a computing station that can provide a useful computational basis for artificial intelligence is very high, therefore, systems that provide for the distribution of the computational load among multiple stations / devices have begun to emerge.

[0008] In the field of computational load sharing, an example is the subject of patent CN108449401 A by GUAN XIANJIN et al. The invention provides a method of sharing computing power based on blockchain technology. The method involves the following steps: creating a computing power chain; receiving a computing power request from an artificial intelligence (Al) by the computing power chain; transmitting the request to a network node in the computing power chain, where the node provides computing power to resolve the request; and rewarding a node deemed to be the fastest with a token. The invention thus provides a solution to the large computational loads typically required by artificial intelligence and specifically rewards nodes that are able to provide greater speed and computing power.

[0009] Another example is the subject of patent application CN111949394A by LIANG YINGTAO et al. The invention provides a method and system for sharing computing power resources and LEIBO. / 45e2025 a storage medium. The method involves the following steps: sending a task request to a server node; obtaining blocks into which the task is divided and executing them in parallel; executing a computation task based on a task block to generate a computation result; verifying the computation result and returning a verified result to the server node; the task request is actively forwarded based on the computation nodes, and a large number of task blocks generated after processing and task division are distributed reasonably to all available computation nodes in a network.

[0010] Another example is the subject of patent US11977961B2 by ERGEN MUSTAFA. The invention relates to a method and system for distributing computing power to a plurality of end devices at different locations using a decentralized architecture. The method and system cluster a plurality of mobile nodes capable of providing highly virtualized computing and storage resources using an artificial intelligence (Al) model. Clustering is performed using two forecasting models: a mobility forecasting model and a theoretical framework. The mobility prediction model learns the timing and direction of movements, as well as the mobility patterns of each of the plurality of mobile nodes, to ascertain the computing capacity for the specified location at a given time. The theoretical framework performs sequential parallel conversion in Al model learning, optimization, and caching algorithms. After clustering, the mobile node cluster is used to process multiple workloads cooperatively for the set of end devices in the specified location.

[0011] The inventions reported so far, by way of non-exhaustive example, are representative of the inventive-technological framework available to date.

[0012] The inventions cited present technical problems related to the fact that although they provide solutions to the problem of distributing computational loads, they remain tied to standard computing sources and preferably to server farms or in any case to specific hardware devices. Some inventions also provide reward mechanisms, but these turn out to reward the fastest nodes (which contribute to the computation). While the inventions provided allow access to great computing power even to nodes and entities that do not possess it, they still depend on and heavily reward complex and expensive hardware devices, which consume a large amount LEIBO. / 45e2025 of energy. Furthermore, these inventions are based on traditional self-learning mechanisms for artificial intelligence. Furthermore, such inventions are often attributed to large companies that have enormous data centers and computing capacity, but they limit access to these resources and offer them at a prohibitive cost for many innovators, researchers, and small businesses. Furthermore, when computational loads are distributed in known inventions, it is impossible to be aware and certain that one or more computational operations will be completed by one or more nodes to which the load is distributed. In this way, the computational load calculations have a distribution among various devices that is extremely random, unpredictable and untraceable, resulting in problems certifying which devices perform which calculations and in problems related to the impossibility of defining a device as a beneficial and legitimate source of computation without referring exclusively to data relating to the quantity of calculations performed / data exchanged.

[0013] The present invention emerges as a revolutionary solution that democratizes access to computing power, transforming everyday devices into a global network of computational resources. The computational load is then divided among a large number of devices. The invention also activates a reward mechanism that is not simply aimed at rewarding the sources that provide greater computational availability, but is also aimed at exploiting and rewarding the plurality of users who interact on the system to train artificial intelligences faster, more precisely (as training is verified and carried out by humans), and in a more varied way.

[0014] The object of the present invention is to provide a system for training and distributing the computational load of artificial intelligences to multiple devices, which may also be portable devices, which is able to solve the above-mentioned problems.

[0015] The advantages offered by the present invention will be clearer in light of the detailed descriptions which follow.

[0016] Description of the invention

[0017] According to the present invention, a system is implemented which includes one or more artificial intelligences. The system trains said artificial intelligences and distributes one or more computational loads among them. For the present invention, it is intended that a LEIBO. / 45e2025 computational load comprises a plurality of computation tasks. In the present invention, the computational loads are those generated by queries made by users to said artificial intelligences, but also generated by said users’ own software. Said system comprises a plurality of devices, each comprising at least one processing unit. Said devices are connected to a web portal with an application. Said computational load is shared between the devices that host one or more client software. Said client software controls the availability of said processing unit to execute one or more computing tasks. Said client software causes said one or more computing tasks to be executed by said processing units. Said computing tasks use a defined percentage of the maximum computing capacity of said processing units. Said defined percentage is the maximum computing power value as a percentage of said maximum computing power of the processing unit, which said computing tasks are enabled to exploit on said device. Said execution of a computation task generates an earning of an amount of a token (which is preferably a digital exchange currency, but also a cryptocurrency and / or a score) for a user who owns said device. The user can spend quantities of said token by making queries to said artificial intelligences and / or requesting allocations, among the system’s devices, of one or more computational loads generated by said user’s software. The system further includes a web portal with an application featuring said artificial intelligences and said token. The web portal with application includes said client software installed on said devices. Said client software receives one or more computing tasks from said artificial intelligences.

[0018] The web portal with application further includes a user section where a user is shown a personal amount of said token. Said user defines said defined percentage of said maximum computing capacity of processing units of his / her device via said user section.

[0019] The web portal with application further includes a training section comprising training objectives. Said training objectives include labeling operations of artificial intelligence output, labeling and / or recognition of obj ects within images and / or videos, transcription of text present in images and / or speech. The results of said operations of said training objectives train said artificial intelligences. The operations of said training objectives are performed by one or more users. Said training section provides an increment of an amount of said token to a user who LEIBO. / 45e2025 completes a training objective.

[0020] The web portal with application further includes one or more query sections that provide, to said users, interfaces on said devices. Users make one or more queries to said artificial intelligences via said interfaces of said query sections, where some queries require spending an amount of said token.

[0021] The web portal with application further comprises one or more decentralized blockchains recording said increments and said expenditures of quantities of said token.

[0022] The web portal with application further includes one or more coordination nodes that receive said computational loads from said portal with application. Each coordination node sends one or more availability requests to said client software. The coordination node receives information from the client software on the availability of the processing units to perform calculations. Said processing units are “available” when the difference between a maximum computing capacity of such processing units and an instantaneous usage percentage of such processing units is greater than or equal to such defined percentage. Said coordination node sends one or more computation tasks of said computational loads to said client software.

[0023] The web portal with application further includes one or more analysis algorithms that evaluate the stability of a connection with one or more devices. Said analysis algorithms send a transmission and / or a test computation task to said device, analyzing the reception, computation and response times thereof. Where said test transmission and / or said computation task are inputs that require the processing units of the devices to perform some elementary / simple computation operations. Such inputs are defined by users with system administrator rights. The analysis algorithms allow said coordination nodes to send computation tasks to said client software of said devices when said reception, computation and response times fall within accepted ranges. Said analysis algorithms prevent the coordination nodes from sending computation tasks to said client software of said devices when said reception, computation and response times fall outside said accepted ranges.

[0024] The present invention also relates to a method of operation of the described system. The method comprises the following steps: LEIBO. / 45e2025

[0025] - one or more training steps where a user accesses said training section of the web portal with application and performs said operations of one or more said training objectives; the results of said operations are controlled by users of the web portal with application and / or by users with administrator rights; a personal quantity of said token of the user being increased when said operations are correctly performed;

[0026] - a query step where a first device performs a query to one or more artificial intelligences and / or makes a request for distribution of one or more computational loads generated by its own software running on said first device; said query generates a computational load; said query step has a positive (I7) or negative (N) result; when a user who owns said first device has a quantity of said token greater than or equal to the quantity of tokens to be spent to carry out said query and / or distribution request, the outcome is positive and the method continues with a request step, otherwise the outcome is negative (TV) and the Web portal with application sends a notification to said user in a notification step;

[0027] - said request step in which one or more coordination nodes send said one or more availability requests to said client software of second devices; said request step has a positive (I7) or negative (TV) outcome; when the difference between said maximum computing capacity of said processing units and said instantaneous usage percentage of said processing units is greater than or equal to said defined percentage, the outcome is positive (I7) and the method continues with the calibration step, otherwise the outcome is negative (TV) and the method waits for a predefined time and repeats said request step;

[0028] - a calibration step in which said analysis algorithms send said transmission and / or test computing task to said second devices analyzing said reception, calculation and response times; said calibration step has a positive (I7) or negative (N) outcome; when said reception, calculation and response times fall within said accepted ranges, the outcome is positive (F) and the method continues with a task sending step, otherwise the outcome is negative (N), and the method continues with a reporting step in which the Web portal with application sends a notification on said second device regarding the impossibility of proceeding with a performance of calculation tasks; LEIBO. / 45e2025

[0029] - one or more of said task sending steps in which said coordination nodes send said calculation tasks to said client software of said second devices; said processing units of said second devices perform said calculation tasks; said client software returns to said coordination nodes one or more results of said calculation tasks; said coordination nodes send said results to said first device. Therefore, the first device receives the results of its query to the artificial intelligence which have been obtained through processing carried out jointly between the second devices. In some embodiments of the present invention, said second devices comprise the first device, implying that the processing unit of the first device is also exploited to perform some computation tasks related to said query.

[0030] The system of the present invention is designed to incentivize users to share their unused computational resources and actively participate in the network through gamification and staking mechanisms. Users can earn Tokens not only based on the quantity and quality of computing power provided, but also by improving their tier within the platform. Additionally, a staking system allows users to earn interest on their tokens with an additional reward multiplier based on the staking level reached. Computing power provided by users is preferably measured in standardized units, such as FLOPS (Floating Point Operations Per Second). The client software installed on users' devices continuously monitors the amount of shared computing resources, ensuring accurate and transparent metering.

[0031] In preferred embodiments of the present invention, users may earn experience points for each computational power contribution made or for each training objective accomplished. By accumulating experience points, users level up, unlocking additional rewards and bonuses.

[0032] Users register on the web portal and download the application and client software. The client software is installed on users' devices (PCs, laptops, servers, smartphones, tablets, smartwatches, and others) and begins monitoring and sharing available computing resources. The client software uses the device's unused computing power to perform computational tasks requested by the network. The client software periodically sends detailed reports to the central network of the web portal application, indicating the amount of computing resources provided and the quality of service (uptime, speed, etc.). LEIBO. / 45e2025

[0033] In preferred embodiments of the present invention, users choose to stake their tokens directly from their wallet on the web portal application. Staking returns are calculated based on the amount of tokens staked, the time spent staking, and the user's staking level. Higher-level users benefit from multipliers on reward distribution, further increasing their earnings.

[0034] Earned tokens are transferred to users’ wallets within the web portal with application. Where users can choose to hold tokens as a long-term investment, exchange tokens for cryptocurrencies, use tokens to pay for services on the web portal with the application, or to purchase additional computing power.

[0035] The web portal and application is designed to scale with increasing user numbers and computing power demands, ensuring the reward mechanism remains effective even as the network grows exponentially.

[0036] The rewards mechanism is designed to be fair, transparent, and incentivized, encouraging active user participation and ensuring proportional distribution of rewards. This approach not only values the contribution of each user, but also creates a sustainable and collaborative ecosystem that fosters innovation and technological progress. Users can turn unused computing power into a valuable resource, earning tangible rewards and contributing to a more equitable and innovative future.

[0037] In the system and related method described, the various devices of the various users that collaborate to share a computational load are tracked and constantly monitored. Scores are also provided, checks are made, and precautions are taken to significantly increase the likelihood that a computation task started on a device will be completed by the device itself. All these precautions are related to the monitoring of the location of the devices provided in some embodiments of the invention, the monitoring of their battery, also provided in some embodiments of the invention, as well as the monitoring of network conditions. In this way, computational load calculations have a predictable distribution that can also be guided by directing particularly complex computational tasks to devices for which these precautions yield positive results. In this way, it is also possible to trace the origin of the data exchanged between devices and the web portal with the application, ensuring greater security. LEIBO. / 45e2025

[0038] The advantages offered by the present invention are evident in the light of the description presented thus far and will be even clearer from the accompanying figures and the related detailed description.

[0039] The invention will hereinafter be described in at least a preferred embodiment thereof by way of non-limiting example with the aid of the accompanying figures, in which:

[0040] - FIGURA 1 shows a general view of a system 100 according to the present invention is shown;

[0041] - FIGURA 2 shows a view of a signaling and regulating device 120 and a device 110 according to the present invention;

[0042] - FIGURA 3 shows a view of a signaling and regulating device 120 and a device 110 according to the present invention during a change of a defined percentage operated by means of a regulation element 122;

[0043] - FIGURA 4 shows a representative block diagram of a method 300 of operating a system 100 according to the present invention.

[0044] Detailed description of the invention

[0045] The present invention will now be illustrated by way of a purely non-limiting or binding example, resorting to the figures which illustrate some embodiments with respect to the present inventive concept.

[0046] With reference to FIG. 1, a general view is shown of a system 100 according to the present invention. In FIG. 1 as in the following description, the embodiment of the present invention currently considered the best is illustrated.

[0047] As shown in FIG. 1, the present invention relates to a system 100 which comprises, trains one or more artificial intelligences 101 and distributes one or more computational loads between them. One or more computational loads are generated by queries that users make to artificial intelligences. Other computational loads are generated by one or more of said users’ software. Each computational load comprises a plurality of computation tasks. Where computation tasks are operations that are performed by processing units such as binary calculation operations. LEIBO. / 45e2025

[0048] The system 100 comprises a plurality of devices 110 each comprising at least one processing unit. Said devices 110 are connected to a web portal with application 200. Said computational load is distributed among said devices 110. The devices 110 host one or more client software 201 that control the availability of said processing unit to execute one or more computing tasks. Said client software 201 causes said one or more computing tasks to be executed by said processing units. Said computing tasks use a defined percentage of the maximum computing capacity of said processing units. Said defined percentage is the maximum computing capacity value (calculated as a percentage of said maximum computing capacity of the processing unit) that said computing tasks are enabled to exploit on the processing unit of said device 110. Said execution of a computation task results in an earning of an amount of a token 202 for a user owning said device 110. Said user spends quantities of said token 202 by making said queries to said artificial intelligences 101 and / or requesting a distribution of one or more computational loads generated by said software of said user, between the devices 110 of the system 100. In this way, for example, a user who needs to carry out an analysis that consumes significant computing resources and therefore generates a huge computational load, can decide to spend said tokens 202 by making a request to distribute said one or more computational loads among the various devices 110 of the system 100. In the embodiments shown in FIG. 1, devices 110 are a fixed computer (center in FIG. 1) a portable laptop (left in FIG. 1) and a smartphone (on the right in FIG. 1).

[0049] The system 100 further comprises a web portal with application 200 in turn comprising said artificial intelligences 101 and said token 202. The Web portal with application 200 also includes:

[0050] - said client software 201 installed on said devices 110; said client software 201 receiving one or more computation tasks from said artificial intelligences 101. The client software 201 is installed on the device 110 with the prior consent of a user who downloads the application 200 of the web portal with application 200 on the device 110 and with it said client software 201;

[0051] - a user section 203 in which a user is shown a personal (own) quantity of said token 202. LEIBO. / 45e2025

[0052] Said user defines said defined percentage of said maximum computing capacity of processing units of his / her device 110 via said user section 203. Specifically, the user section 203 includes interfaces and controls through which the user can define said defined percentage. By way of non-limiting or binding example, the user may define that said defined percentage is equal to 20%, which implies that said computing tasks never use a computing capacity of the processing unit of the user's device 110 that exceeds 20% (percentage with respect to the maximum capacity);

[0053] - a training section 210 including training objectives. The training objectives involve classifying the output of the artificial intelligences 101 into predefined categories. The training objectives further include requests to label images as “dog,” “cat,” or “bird” based on the visual content of one or more images and / or videos. The training objectives further include identifying and labeling specific objects such as “cars,” “people,” “buildings,” and others within an image or video. For example, the training objectives include a transcription of text contained in images or videos, thus helping to create annotated datasets useful for training optical character recognition (OCR) and automatic speech transcription models. The training objectives further include evaluating the quality of the outputs generated by the artificial intelligences 101 models, providing feedback on the accuracy and reliability of the predictions generated by the artificial intelligences 101 models. The results of said operations of said training objectives are used to train, i.e. provide knowledge bases for said artificial intelligences 101. The operations of said training objectives are performed by one or more users. The training section 210 provides an increment of an amount of said token 202 to a user who completes a training objective as a reward mechanism;

[0054] - one or more query sections 220 providing, to said users, interfaces on said devices 110. Said users make one or more queries to said artificial intelligences 101 via said interfaces of said query sections 220. Some queries require spending an amount of said token 202;

[0055] - one or more decentralized blockchains 230 that record said increases and said expenditures of the quantity of said token 202 in order to give a unique and certain value to said token 202; LEIBO. / 45e2025

[0056] - one or more coordination nodes 240 that receive said computational loads from said application portal 200. Said coordination node 240 sends one or more availability requests to said client software 201. Said coordination node 240 receives information from said client software 201 on the availability of said processing units for the execution of calculations. Such processing units are considered “available” when the difference between their maximum computing capacity and their instantaneous usage percentage (usage by any process running at that instant by the processing unit) is greater than or equal to such defined percentage. Said coordination node 240 sends one or more computation tasks of said computational loads to said client software 201. Said coordination nodes 240 are preferably organized on coordination stations comprising one or more processing and sorting devices that perform operations of receiving computational loads, sending said availability requests and receiving information on said availability. Such processing and sorting devices are preferably servers with computing units and software that govern said operations. The sending of calculation tasks is carried out using specific connections via the Internet that are believed to be known by an expert in the field and which will therefore not be described in detail;

[0057] - one or more analysis algorithms 250 that evaluate the stability of a connection with one or more devices 110. Said analysis algorithms 250 send a transmission and / or a test calculation task to said device 110 analyzing its reception, calculation and response times. Where said transmission and / or said test computation task are inputs that require the processing units of the devices 110 to perform some elementary / simple computation operations. Such inputs are defined by users with system administrator rights 100. Said analysis algorithms 250 allow said coordination nodes 240 to send computation tasks to said client software 201 of said devices 110 when said reception, computation and response times fall within accepted ranges. Said analysis algorithms 250 prevent the sending of calculation tasks to said client software 201 of said devices 110 when said reception, calculation and response times fall outside said accepted ranges. Where said “accepted ranges” are time ranges that are manually defined by users with administrator rights of system 100 and / or LEIBO. / 45e2025 defined / modified / updated by the analysis algorithms 250 based on the communication experiences of system 100.

[0058] In some embodiments of the present invention, said processing units of said devices 110 include graphics processing units (GPUs) and central processing units (CPUs). Where for GPUs, the maximum computing power is preferably calculated based on the dedicated vRAM (Video RAM) of the device 110, where 100% of the maximum computing power corresponds to a situation where said vRAM is fully utilized.

[0059] In some embodiments of the present invention, said device 110 includes a geolocalization unit that acquires data on a location of said device 110. Said geolocalization unit sends said location data to said analysis algorithms 250 of the web portal with application 200. Said analysis algorithms 250 further analyze said position data and one or more variations of said position data, where the variation of said position data provides information regarding a rate of travel. Sending a first transmission and / or test computation task from said analysis algorithms 250 to said device 110 results in the acquisition of an initial position by said geolocalization unit. This initial position is acquired by the analysis algorithms 250. Said analysis algorithms 250 send said transmissions and / or test calculation tasks to said device 110 with a time frequency that is defined by users with administration rights on the web portal with application 200. Said analysis algorithms 250 increase said temporal frequency of sending said transmissions and / or sent test computation tasks and / or send a signal to said coordination nodes 240, as the distance of said device 110 from said initial position increases and / or as the speed of movement of the device 110 as measured by said changes in said position data increases. Said signal sent to said coordination nodes 240 produces a reduction or interruption in the sending of computing tasks to said client software 201 of said device 110. Preferably, ranges for distance from the starting position and speed are defined in the web portal with application 200. By “defined ranges” we mean maximum and minimum limits that have been defined by users with administrator rights for the web portal with application 200 and / or by the analysis algorithms 250 themselves on the basis of training and historical analyses. These defined ranges are related to position data, therefore for example to movements that occur within specific areas, within a distance radius LEIBO. / 45e2025

[0060] (preferably expressed in meters) from said initial position, and are related to variations in data, such as the speed of movement of a device 110 (and therefore of the user who owns it). A defined range preferably refers to movements performed at speeds below 5 km / h and at distances from a first acquired position, contained within 100 m. This defined range is indicative of a static or quasi-static movement condition of the device 110, that is, indicative of the fact that the device 110 is stationary or almost stationary or that it is in any case moving very slowly. In such a motion condition the analysis algorithms 250 send a limited number of transmissions and / or test computation tasks. Said number of transmissions and / or test calculation tasks is therefore related to a non-limiting time frequency of, for example, 30 minutes to verify the permanence of the network conditions measured by sending said first transmission and / or test calculation task. In movement conditions in which a speed of movement greater than 80 km / h is detected, for example, understanding that the device 110 is moving on a means of transport, the analysis algorithms 250 send more frequently (by increasing the temporal frequency) said transmissions and / or test calculation tasks to prevent a large and sudden movement of the device 110 from causing it to reach areas in which reception and / or network coverage is limited.

[0061] In some embodiments of the present invention, said application web portal 200 sends an acquisition request to a user on a device 110. Said acquisition request requires a time interval in which to send said computation tasks to said device 110. Said acquisition request further requires one to confirm or modify the defined percentage for the said time interval. Said analysis algorithms 250 acquire said time interval, said defined percentage and acquire said reception, calculation and response times of said device 110. Said analysis algorithms 250 evaluate which and how many computing tasks said device 110 can carry out in said time interval. Said analysis algorithms 250 communicate to said coordination nodes 240, said calculation tasks (which and how many) can be performed by said device 110. By way of preferential example, the acquisition request provides an interface through which a user can indicate via an appropriate digital key / drop-down menu or text field a time when the device 110 will be available (and therefore free from other processes) to perform said calculation LEIBO. / 45e2025 tasks and said defined percentage. By way of non-limiting or binding example, a user who is about to do a two-hour physical activity in a gym leaving his / her device 110 unused, may not only provide said time interval equal to two hours, but may also increase said defined percentage knowing that the device 110 must remain unused.

[0062] In some embodiments of the present invention such as that shown in FIGGS. 2 and 3, the system 100 further comprises a signaling and regulation device 120 connected to said device 110 via a wired connection (e.g. with a USB connection as shown in FIGGS. 2 and 3) or even wireless (e.g. Bluetooth, not shown in the figures), and further connected to said web portal with application 200 and preferably to the user section 203. Said signaling and regulating device 120 comprises a first and a second light bar 121’, 121”. Said first light bar 121 ’ lights up indicating said defined percentage, while said second light bar 121” lights up indicating an instantaneous percentage of total usage of one or more processing units of said device 110. The signaling and regulating device 120 comprises one or more regulating elements 122 which are sliders, levers, washers and / or analog and / or digital buttons. Said adjustment elements 122 modify said defined percentage when operated by a user and the modification of said defined percentage is shown on said first light bar 121’ which lights up accordingly. Said processing units of device 110 calculate said instantaneous percentage of total usage and instantly display and update said instantaneous percentage of total usage on said second light bar 121”. In this way, the user who experiences slowdowns in the normal use of the device 110 can immediately verify whether the instantaneous percentage of total use is equal to or close to 100% (indicative of an excessive number of processes running simultaneously on the device 110) and can consequently see whether the reduction of the defined percentage (operated with the adjustment elements 122) produces an improvement in the performance of the device 110 by reducing the instantaneous percentage of total use. Furthermore, in this way, the user can immediately change the percentage defined in relation to even short periods of inactivity, without having to log in to their user section 203. For example, if the user has to take a 15- minute break away from the device 110, he / she can quickly move a control element 122 to increase the value by a defined percentage, for example, to 80%, and immediately restore that LEIBO. / 45e2025 value, by doing the reverse operation, to a value acceptable to him / her upon returning from the break.

[0063] In some embodiments of the present invention such as that shown in FIGGS. 2 and 3, said signaling and adjustment device 120 further comprises a display 123 and said adjustment elements 122 are digital sliders, digital levers, digital washers and / or digital buttons displayed on said display 123.

[0064] In the specific case shown in FIGGS. 2 and 3, the signaling and regulating device 120 has a touchscreen display 123 which allows said regulating element 122 to be operated by touching it on the display 123 (as shown in FIG. 3). FIG. 2 shows a light bar 121’ indicative of a defined percentage of 27% and a light bar 121” indicative of an instantaneous percentage of total usage of 75%. FIG. 3 shows the same signaling and regulating device 120 as in FIG. 3 wherein a user’s hand touches the touchscreen display 123 to move said adjustment element (which is a slider in FIGS. 2 and 3) along the light bar 121’ upwards, thus producing an increase in the defined percentage that from 27% (FIG. 2) goes to 58% (FIG. 3). The resulting change in the defined percentage, shown in FIG. 3, produces an increase in the said instantaneous percentage of total usage which reaches 100%. In this way the user immediately sees what percentage is defined that generates an instantaneous percentage of total usage equal to or less than 100% or any other desired value.

[0065] In some embodiments of the present invention, one or more devices 110 include batteries. Said batteries provide power to said devices 110. Said analysis algorithms 250 further acquire, in these embodiments, data relating to a percentage of charge of a battery of said device 110 and data relating to the variation of said percentage of charge. Said analysis algorithms 250 analyze said variation of said charge percentage following said transmission and / or test calculation task performed by the processing unit of the device 110. Said analysis algorithms 250 predict percentage reductions in the charge of a battery of a device 110 and associate a predicted percentage reduction with each computational task. Said analysis algorithms 250 prevent said coordination nodes 240 from sending to client software 201 of a device 110 computation tasks having a predicted percentage reduction greater than, equal to or close to a residual percentage LEIBO. / 45e2025 of battery charge of the device 110. Where the expected percentage is “close” to the residual charge percentage, it is meant that the residual charge percentage is greater than the expected one and that it is preferably greater than a percentage value preferably within the range from 0% to 10%. In this way, the analysis algorithms 250 (communicating with the coordination nodes 240) prevent computation tasks from remaining unfinished due to a sudden shutdown caused by a battery running low or by battery protection mechanisms that lead to the shutdown of the device 110 before reaching a charge percentage of 0%. Furthermore, by doing so, interrupting the sending of computing tasks frees the processing units from their computational load and extends the remaining battery life.

[0066] The present invention further relates to a method 300 of operating the system 100 described. With reference to FIG. 4, a block diagram is shown representing a method 300 of operation which comprises the following steps:

[0067] - one or more training steps 301 wherein a user accesses said training section 210 of the web portal with application 200 and performs said operations of one or more said training objectives. The results of these operations are controlled by users of the web portal with application 200 and / or by users with administrator rights. When the quality of the results is confirmed, that is, the operations have been correctly executed, the user earns (sees an increase in) his / her own quantity of said token 202. Where “correctly executed” means that said users of the Web portal with application 200 and / or users with administrator rights consider said operations to be error-free;

[0068] - a query step 302 in which a first device 110 makes a query to one or more artificial intelligences 101 and / or a request for the distribution of one or more computational loads. Said query generates a computational load. The query step 302 has a positive (F) or negative (TV) outcome. When a user who owns said first device 110 has a quantity of said token 202 greater than or equal to the quantity of tokens 202 to be spent (of the cost in terms of tokens 202) to carry out said query and / or distribution request, the outcome is positive (F) and the method 300 continues with a request step 305. When said quantity of said token 202 is less than said quantity of tokens 202 to be spent (cost) to carry out said query and / or distribution LEIBO. / 45e2025 request, the outcome is negative (N) and the web portal with application 200 sends a notification to said user in a notification step 303 in which the insufficiency of tokens 202 is reported to proceed with the query and / or distribution request and in which actions for earning the token 202 can be recommended such as a purchase (exchange of money for token 202), the completion of training objectives or even the provision of a defined percentage of computing capacity of the processing unit of the device 110 to carry out one or more computing tasks;

[0069] - a request step 305 wherein one or more coordination nodes 240 send said one or more availability requests to said client software 201 of second devices 110. Said request step 305 has a positive (f) or negative (N) outcome. When the difference between said maximum computing capacity of said processing units and said instantaneous usage percentage of said processing units is greater than or equal to said defined percentage, the outcome is positive (I7)- When said difference is less than said defined percentage, the outcome is negative (TV) and the method 300 waits a predefined time and repeats said request step 305. Said “default time” is a time defined by a user with administrator rights in the web portal with application 200;

[0070] - a calibration step 310 wherein said analysis algorithms 250 send said transmission and / or test calculation task to said second devices 110 analyzing said reception, calculation and response times. Said calibration step 310 has a positive (f) or negative (N) outcome. When said reception, calculation and response times fall within the said accepted ranges, the outcome is positive (f) and the method 300 continues with a task sending step 320. When said reception, calculation and response times fall outside said accepted ranges, the outcome is negative (N), and the method 300 continues with a reporting step 311 in which the web portal with application 200 sends a notification to said second device 110 regarding the impossibility of proceeding with the execution of calculation tasks;

[0071] - one or more task sending steps 320 wherein said coordination nodes 240 send said computation tasks to said client software 201 of said second devices 110. Said processing units of said second devices 110 perform said computing tasks. Said client software 201 LEIBO. / 45e2025 returns to said coordination nodes 240 one or more results of said computation tasks and said coordination nodes 240 send said results to said first device 110.

[0072] In some preferred embodiments of the present invention, said second devices 110 comprise the first device 110, implying that the processing unit of the first device 110 is also used to perform one or more tasks of calculating the computational load resulting from said query of the user using the first device 110 to said artificial intelligences 101.

[0073] In some embodiments of the present invention in which the system 100 comprises battery- powered devices 110, in said calibration step 310, said analysis algorithms 250 further acquire data relating to a charge percentage of a battery of said second device 110 and data relating to the change in said charge percentage following said transmission and / or test calculation task. Said analysis algorithms 250 predict a reduction in the percentage of charge of a battery of said second device 110 caused by each computational load calculation task required by said query and / or by said distribution request. Said calibration step 310 has a negative outcome (N) further when said analysis algorithms 250 foresee for said calculation tasks a percentage reduction greater than, equal to or close to a residual percentage of battery charge of said second device 110.

[0074] Finally, it is clear that modifications, additions or variations that are obvious to a person skilled in the art may be made to the invention described so far, without thereby departing from the scope of protection provided by the appended claims.

Claims

LEIBO. / 45e2025Claims1. System (100) comprising one or more artificial intelligences (101); said system (100) training said artificial intelligences (101) and allocating one or more computational loads to them; one or more computational loads being generated by queries to said artificial intelligences (101); each computational load comprising a plurality of computational tasks; said system (100) being characterized in that it comprises:- a plurality of devices (110) each comprising at least one processing unit; said devices (110) being connected to a web portal with application (200); said computational load being distributed among said devices (110); said devices (110) hosting one or more client software (201); said client software (201) controlling an availability of said processing unit to perform one or more computation tasks; said client software (201) causing said one or more computation tasks to be performed by said processing units; said computation tasks using a defined percentage of a maximum computation capacity of said processing units; said defined percentage being the maximum computation capacity value in percentage with respect to said maximum computation capacity of the processing unit, which said computation tasks are enabled to exploit on said device (110); said execution of a computation task resulting in a gain of an amount of a token (202) for a user owning said device (110); said user spending amounts of said token (202) by making queries to said artificial intelligences (101) and / or making requests for the distribution of one or more computational loads generated by said user’s software, among the devices (110) of the system (100);- a Web portal with application (200) comprising said artificial intelligences (101) and said token (202); said Web portal with application (200) further comprising: o said client software (201) installed on said devices (110); said client software (201) receiving one or more computation tasks from said artificial intelligences (101) and / or said software from said users; o a user section (203) in which a personal quantity of said token (202) is shown to a user; said user defining said defined percentage of said maximumLEIBO. / 45e2025 computation capacity of processing units of his / her device (110) via said user section (203); o a training section (210) comprising training objectives; said training objectives comprising operations of labeling output of the artificial intelligences (101), labeling and / or recognition of objects within images and / or videos, transcription of text present in images and / or speech; the results of said operations of said training objectives training said artificial intelligences (101); said operations of said training objectives being performed by one or more users; said training section (210) providing an increment of an amount of said token (202) to a user completing a training objective; o one or more query sections (220) providing, to said users, interfaces on said devices (110); said users making one or more queries to said artificial intelligences (101) via said interfaces of said query sections (220); some queries requesting an expenditure of an amount of said token (202); o one or more decentralized blockchains (230) recording said increments and said expenditures of quantities of said token (202); o one or more coordination nodes (240) receiving said computational loads from said application portal (200); said coordination node (240) sending one or more availability requests to said client software (201); said coordination node (240) receiving information, from said client software (201) on the availability of said processing units to perform calculations; said processing units being available when the difference between a maximum computing capacity of said processing units and an instantaneous usage percentage of said processing units is greater than or equal to said defined percentage; said coordination node (240) sending one or more computing tasks of said computational loads to said client software (201); o one or more analysis algorithms (250) evaluating a stability of a connection with one or more devices (110); said analysis algorithms (250) sending aLEIBO. / 45e2025 transmission and / or a test computing task to said device (110) analyzing its reception, calculation and response times; said analysis algorithms (250) allowing said coordination nodes (240) to send calculation tasks to said client software (201) of said devices (110) when said reception, calculation and response times fall within accepted ranges; said analysis algorithms (250) preventing the sending of calculation tasks to said client software (201) of said devices (110) when said reception, calculation and response times fall outside said accepted ranges.

2. System (100), according to the preceding claim 1, characterized in that said processing units of said devices (110) comprise graphics processing units (GPU) and central processing units (CPU).

3. System (100), according to the preceding claim 1 or 2, characterized in that said device (110) comprises a geolocalization unit acquiring data on a position of said device (110); said geolocalization unit sending said position data to said analysis algorithms (250) of the web portal with application (200); said analysis algorithms (250) further analyzing said position data and one or more variations of said position data; sending a first transmission and / or test calculation task from said analysis algorithms (250) to said device (110) resulting in the acquisition of an initial position by said geolocalization unit; said initial position being acquired by said analysis algorithms (250); said analysis algorithms (250) sending said transmissions and / or test calculation tasks to said device (110) with a temporal frequency; said analysis algorithms (250) increasing said temporal frequency of sending said transmissions and / or test calculation tasks sent and / or sending a signal to said coordination nodes (240), as the distance of said device (110) from said initial position increases and / or as the speed of movement of the device (110) measured by said variations of said position data increases; said signal sent to said coordination nodes (240) producing a reduction or an interruption of the sending of calculation tasks to said client software (201) of said device (110).

4. System (100), according to any preceding claim, characterized in that said web portalLEIBO. / 45e2025 with application (200) sends an acquisition request to a user on a device (110); said acquisition request requesting a time slot in which to send said computation tasks to said device (110); said acquisition request further requesting to confirm or modify said percentage defined for said time slot; said analysis algorithms (250) acquiring said time slot, said defined percentage and acquiring said reception, computation and response times of said device (110); said analysis algorithms (250) evaluating which and how many computation tasks said device (110) can perform in said time slot; said analysis algorithms (250) communicating said computation tasks performable by said device (110) to said coordination nodes (240).

5. System (100), according to any of the preceding claims, characterized in that it further comprises a signaling and regulating device (120); said signaling and regulating device (120) being connected to said device (110); said signaling and regulating device (120) being further connected to said web portal with application (200); said signaling and regulating device (120) comprising a first and a second light bar (121’, 121”); said first light bar (121”) lighting up indicating said defined percentage; said second light bar (121”) lighting up indicating an instantaneous percentage of total use of one or more processing units of said device (110); said signaling and regulating device (120) comprising regulation elements (122); said regulation elements (122) being sliders, levers, washers and / or buttons; said regulation elements (122) modifying said defined percentage; said modification of said defined percentage being displayed on said first light bar (121’); said processing units of the device (110) calculating said instantaneous percentage of total use and instantly displaying and updating said instantaneous percentage of total use on said second light bar (121”).

6. System (100), according to the preceding claim 5, characterized in that said signaling and adjustment device (120) further comprises a display (123); said adjustment elements (122) being digital sliders, digital levers, digital washers and / or digital buttons displayed on said display (123).

7. System (100), according to any of the preceding claims, characterized in that one orLEIBO. / 45e2025 more devices (110) comprise batteries; said batteries providing power to said devices (110); said analysis algorithms (250) acquiring data relating to a percentage of charge of a battery of said device (110) and data relating to the variation of said percentage of charge; said analysis algorithms (250) analyzing said variation of said charge percentage following said transmission and / or test calculation task performed by the processing unit of the device (110); said analysis algorithms (250) predicting reductions in charge percentages of a battery of a device (110) and associating a predicted percentage reduction to each calculation task; said analysis algorithms (250) preventing said coordination nodes (240) from sending to client software (201) of a device (110) calculation tasks having a predicted percentage reduction greater than, equal to or close to a residual battery charge percentage of the device (110).

8. Method (300) of operating the system (100), according to any of the preceding claims, characterized in that it comprises the following steps:- one or more training steps (301) where a user accesses said training section (210) of the web portal with application (200) and performs said operations of one or more said training objectives; the results of said operations being controlled by users of the web portal with application (200) and / or by users with administrator rights; a personal quantity of said token (202) of the user being increased when said operations are correctly performed;- a query step (302) where a first device (110) performs a query to one or more artificial intelligences (101) and / or a request for distribution of one or more computational loads; said query generating a computational load; said query step (302) having a positive (Y) or negative (N) result; when a user who owns said first device (110) has a quantity of said token (202) greater than or equal to the quantity of tokens (202) to be spent to carry out said query and / or distribution request, the outcome is positive (Y) and the method (300) continues with a request step (305); when said quantity of said token (202) is less than said quantity of tokens (202) to be spent to carry out said query and / or distribution request the outcome is negative (N) and the Web portal withLEIBO. / 45e2025 application (200) sends a notification to said user in a notification step (303);- a request step (305) in which one or more coordination nodes (240) send said one or more availability requests to said client software (201) of second devices (110); said request step (305) having a positive (Y) or negative (N) outcome; when the difference between said maximum computing capacity of said processing units and said instantaneous usage percentage of said processing units is greater than or equal to said defined percentage, the outcome is positive (Y); when said difference is less than said defined percentage, the outcome is negative (N) and the method (300) waits for a predefined time and repeats said request step (305);- a calibration step (310) in which said analysis algorithms (250) send said transmission and / or test computing task to said second devices (110) analyzing said reception, calculation and response times; said calibration step (310) having a positive (Y) or negative (N) outcome; when said reception, calculation and response times fall within said accepted ranges, the outcome is positive (Y) and the method (300) continues with a task sending step (320); when said reception, calculation and response times fall outside said accepted ranges, the outcome is negative (N), and the method (300) continues with a reporting step (311) in which the Web portal with application (200) sends a notification on said second device (110) regarding the impossibility of proceeding with a performance of calculation tasks;- one or more task sending steps (320) in which said coordination nodes (240) send said calculation tasks to said client software (201) of said second devices (110); said processing units of said second devices (110) performing said calculation tasks; said client software (201) returning to said coordination nodes (240) one or more results of said calculation tasks; said coordination nodes (240) sending said results to said first device (110).

9. Method (300) of operation, according to the preceding claim 8, of a system (100), according to the preceding claim 7, characterized in that in said calibration step (310), said analysis algorithms (250) further acquire data relating to a percentage of charge of aLEIBO. / 45e2025 battery of said second device (110) and data relating to the variation of said percentage of charge following said transmission and / or test calculation task; said analysis algorithms (250) making a prediction of a reduction in percentage of charge of a battery of said second device (110) caused by each computational load calculation task required by said query and / or by said distribution request; said calibration step (310) having a negative outcome (N) further when said analysis algorithms (250) predict a percentage reduction greater than, equal to or close to a residual percentage of battery charge of said second device (110).

Images