program

Abstract

To provide a program capable of objectively evaluating whether a user's behavior is "human-like".SOLUTION: In a system where a server 1 and a plurality of user terminals are connected to enable mutual communication via a communication network, a program inputs learning data including behavior data and correct answer data into system to train a first learning model and a second learning model. When unknown data is input, the program evaluates the behavior, outputs the first output data on the basis of the evaluation results, determines the authenticity of the behavior by the user using the first data, and generates second output data on the basis of the determination results. The program then performs output on the basis of the first output data and the second output data.SELECTED DRAWING: Figure 12

Description

[Technical Field]

[0001] The present invention relates to a program. [Background technology]

[0002] Conventionally, there is known a technique for recognizing user behavior using images. Then, various sensors are used to acquire data such as image data, video data, acoustic data, tactile data, kinesthetic data, temperature data, kinematic data, and wireless signal data, and scoring is performed. Specifically, a technique for realizing scoring using a contextually relevant scoring technique is known (for example, see Patent Document 1, etc.). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2017-525070 Summary of the Invention [Problem to be solved by the invention]

[0004] Advances in technologies such as AI (Artificial Intelligence, hereinafter referred to as "AI") and IoT (Internet of Things, hereinafter referred to as "IoT") have made it possible to recognize user behavior.

[0005] While software and other tools can automate various processes on behalf of users, there is a growing demand for users to perform actions that are difficult to replace with software, i.e., "human-like" behavior. Furthermore, there is a need to objectively evaluate whether a behavior is "human-like." However, conventional technologies cannot objectively evaluate whether a user is behaving in a "human-like" manner.

[0006] The present invention aims to objectively evaluate whether a user's behavior is "human-like." [Means for solving the problem]

[0007] In order to solve the above problem, the present invention provides a program that causes a computer to function as a generation means that generates output data according to the content of a judgment of the truth or falsity of a user's behavior when unknown data related to the user's behavior is input into a trained model that has learned the user's behavior data, and as an output means that outputs the output data. [Effects of the Invention]

[0008] According to the present invention, it is possible to objectively evaluate whether a user's behavior is "human-like" or not. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of a system configuration according to an embodiment of the present invention. [Figure 2] FIG. 10 is a diagram illustrating an example of pre-processing. [Figure 3] FIG. 10 is a diagram illustrating an example of execution processing. [Figure 4] FIG. 1 is a diagram illustrating a first example of overall processing of AI learning and execution. [Figure 5] FIG. 10 is a diagram illustrating a second example of the overall process of AI learning and execution. [Figure 6] FIG. 2 is a diagram illustrating an example of a hardware configuration of an information processing device. [Figure 7] FIG. 1 is a network diagram showing an example of an AI configuration. [Figure 8] FIG. 10 is a diagram illustrating an example of input to the first trained model. [Figure 9] FIG. 10 is a diagram illustrating an example of input to a second trained model. [Figure 10] FIG. 10 is a diagram illustrating an example of execution processing. [Figure 11] FIG. 10 illustrates an example of conversion to an in-game reward. [Figure 12] FIG. 10 is a diagram illustrating an example of overall processing. [Figure 13] FIG. 2 is a diagram illustrating an example of a functional configuration. [Figure 14] FIG. 10 is a diagram illustrating a configuration example using an auxiliary device. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment will be described with reference to the drawings.

[0011] [System configuration example] 1 is a diagram showing an example of a system configuration according to this embodiment. For example, as shown in FIG. 1, a system 1 mainly includes user terminals 20A, 20B, and 20C (hereinafter, these may be collectively referred to as "user terminals 20") and a server 11.

[0012] Hereinafter, the person who manages the server 11 will be referred to as the "administrator 5." Furthermore, the people who operate the user terminals 20A, 20B, and 20C will be referred to as the "user 4A," "user 4B," and "user 4C," respectively (hereinafter, these may be collectively referred to as the "user 4").

[0013] The administrator 5 is a person who operates the information processing service provided by the system 1. On the other hand, the user 4 is a person who uses the information processing service provided by the system 1. Furthermore, the administrator 5 and the user 4 differ in whether they operate an information processing device, the server 11, which is an example of a management device, or the user terminal 20.

[0014] Although the example shown in FIG. 1 has three user terminals 20 and one server 11, the number of servers 11, the number of user terminals 20, the number of administrators 5, and the number of users 4 are not important.

[0015] The server 11 and the user terminal 20 are connected to each other so as to be able to communicate with each other via a communication network 2. For example, the communication network 2 is the Internet, a mobile communication system (for example, a public line such as 4G (4th Generation, fourth generation mobile communication standard) or 5G (5th Generation, fifth generation mobile communication standard)), a wireless network such as Wi-Fi (registered trademark), or a combination of these.

[0016] The user terminal 20 downloads a program from the server 11 or accesses the server 11 to provide a service. Note that communication with the server 11 is not required to execute the process. That is, the user terminal 20 may download a program or install it from a medium to create an execution environment. Note that a game service is provided to the user 4, for example, by installing a program downloaded from the server 11 into the user terminal 20 or by accessing the server 11.

[0017] [Examples of AI learning and execution] Hereinafter, AI learns through "pre-processing." The AI ​​in the learning stage, i.e., "pre-processing," will be referred to as the "learning model A1." Then, once learning has progressed to a certain extent, the learning model A1 becomes the "trained model A2." Hereinafter, the execution stage in which output processing is performed using the trained model A2 will be referred to as the "execution process."

[0018] Note that this embodiment uses two AIs. Of the two AIs, the learning model A1 and the trained model A2 of one AI are referred to as the "first learning model A11," and the trained model A2 is referred to as the "first trained model A21." Similarly, of the two AIs, the learning model A1 and the trained model A2 of the other AI are referred to as the "second learning model A12," and the trained model A2 is referred to as the "second trained model A22."

[0019] In the following explanation, the first learning model A11 and the second learning model A12 are collectively referred to as the learning model A1. Meanwhile, the first trained model A21 and the second trained model A22 are collectively referred to as the trained model A2. Note that in the following explanation, the AIs are divided into two, but one AI may have two functions, or processing may be distributed among three or more AIs.

[0020] The "pre-processing" is performed before the "execution processing." However, the "pre-processing," i.e., the trained model A2 may continue to train after the "execution processing."

[0021] [Pre-processing example] 2 is a diagram showing an example of pre-processing. For example, the pre-processing is performed by the server 11.

[0022] The learning model A1 receives learning data D1 as input and performs learning. That is, the learning model A1 performs so-called "supervised" learning. Below, a first learning model A11 and a first trained model A21 will be described as examples.

[0023] The learning data D1 includes known behavioral data and the like, and is data associated with a "correct answer" for this behavioral data. Specifically, the learning data D1 includes behavioral data indicating the results and details of the behavior, such as eleventh data D11, twelfth data D12, thirteenth data D13, etc. (including a data group that compiles multiple data such as the eleventh data D11, the twelfth data D12, and the thirteenth data D13. Hereinafter, simply referred to as "behavioral data"), and correct answer data D20.

[0024] The learning model A1 learns a correspondence relationship that outputs correct answer data D20 in response to input of behavioral data, based on input of learning data D1.

[0025] The behavioral data will be described in detail later.

[0026] Furthermore, it is desirable that the learning model A1 learns using big data D4. For example, the big data D4 is data on the Internet. However, the big data D4 may also be data input by an administrator 5 or the like.

[0027] [Execution processing example] 3 is a diagram showing an example of the execution process. For example, the execution process is performed by the user terminal 20, or by the user terminal 20 and the server 11 working together.

[0028] The trained model A2 is in a state where the trained model A1 has been trained through pre-processing. That is, when the pre-processing shown in Figure 2 is executed, the trained model A2 is generated.

[0029] When unknown data D2 is input, the trained model A2 generates output data D3 for the unknown data D2.

[0030] The unknown data D2 is data for which the "correct answer" to the behavioral data is unknown at the time of input. Specifically, the unknown data D2 includes behavioral data (including a data group that compiles multiple data such as the 21st data D21, the 22nd data D22, and the 23rd data D23, etc.; hereinafter simply referred to as "behavioral data") such as the 21st data D21, the 22nd data D22, and the 23rd data D23. For example, the unknown data D2 is composed of data of the same type (different data format, etc., but different content) as the training data D1.

[0031] When the output data D3 is generated, the output data D3 is transmitted to, for example, the user terminal 20. Thereafter, the user terminal 20 outputs an output screen or the like to the user 4 based on the output data D3.

[0032] The output data D3 and the output based on the output data D3 will be described in detail later.

[0033] 4 and 5 are diagrams showing an example of the overall process of AI learning and execution.

[0034] FIG. 4 is a diagram illustrating a first example of overall processing of AI learning and execution.

[0035] FIG. 5 is a diagram illustrating a second example of the overall process of AI learning and execution.

[0036] The relationship between the pre-processing shown in Figure 2 and the execution processing shown in Figure 3 is as shown in Figure 4 for the first learning model A11 and the first trained model A21. On the other hand, the relationship between the pre-processing shown in Figure 2 and the execution processing shown in Figure 3 is as shown in Figure 5 for the second learning model A12 and the second trained model A22.

[0037] Note that the pre-processing and execution processing do not have to be performed in the sequential order illustrated in the figure. Therefore, it is not necessary to consecutively perform the period during which preparation is performed by the pre-processing and the subsequent period during which execution processing is performed. Therefore, once the trained model A2 has been created, the execution processing may be performed after a period of time has elapsed since the pre-processing. Furthermore, once the trained model A2 has been generated, the execution processing may be performed by reusing the trained model A2.

[0038] The learning data D1 and unknown data D2 are different between the learning process and the execution process. Also, the AI ​​starts as a learning model A1 in the learning stage, but after a certain amount of learning, it becomes a trained model A2. In this way, the trained model A2 trained using big data D4 as training data is what is known as "generative AI."

[0039] The eleventh data D11 included in the training data D1 and the twenty-first data D21 included in the unknown data D2 (hereinafter, the same relationship applies to the twelfth data D12 and the twenty-second data D22, etc.) are the same data type. That is, the eleventh data D11 and the twenty-first data D21 are both behavioral data. Note that the behavioral data may be shared by the two AIs (for example, copied and input to each AI), or different behavioral data (including cases where some of the behavioral data are different) may be input. Below, an example in which the behavioral data is shared by the two AIs will be described.

[0040] In the learning data D1, the "correct answer" is known, whereas in the unknown data D2, the "correct answer" is unknown. Specifically, the learning data D1 includes the correct answer data D20, whereas the unknown data D2 does not include the correct answer data D20. Therefore, in the learning data D1, the relationship between the behavioral data and the correct answer data D20 is known.

[0041] On the other hand, the unknown data D2 does not include the correct answer data D20, and the "correct answer" for the unknown data D2 is unknown. The trained model A2 generates output data D3 for the unknown data D2 based on the correlation between the training data D1 trained in pre-processing and the correct answer data D20.

[0042] Comparing the first overall process and the second overall process, the correct answer data D20 used in learning is different, i.e., the content of the output data D3 output in the execution process is different. Hereinafter, the correct answer data D20 used in learning the first learning model A11 will be referred to as "first correct answer data D51." On the other hand, the correct answer data D20 used in learning the second learning model A12 will be referred to as "second correct answer data D52."

[0043] The first correct answer data D51, the second correct answer data D52, and the output data D3 will be described in detail later.

[0044] The execution process may be partly a process that uses a table, etc. In this way, in a configuration that uses a table, a so-called rule-based configuration, the pre-processing is a process that prepares for inputting a table (also called a look-up table (LUT)), a mathematical formula, etc.

[0045] [Example of hardware configuration of information processing device] FIG. 6 is a hardware configuration diagram of an information processing device. The information processing device is a server 11, a user terminal 20, etc. Hereinafter, it is assumed that the information processing device has the same hardware configuration as the server 11. For example, the information processing device is a workstation or a general-purpose computer such as a personal computer. However, each information processing device may have a different hardware configuration. Furthermore, it is preferable that behavior be sensed by an IoT device, a wearable terminal, or the like.

[0046] The server 11 mainly includes a processor 111, a memory 112, a storage 113, an input / output interface 114, and a communication interface 115. Furthermore, each component of the server 11 is connected to a communication bus .

[0047] The processor 111 executes a series of instructions contained in a server program 11P stored in the memory 112 or the storage 113, thereby realizing processing and control.

[0048] The processor 111 is, for example, an arithmetic device and a control device such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an MPU (Micro Processing Unit), an FPGA (Field-Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), or a combination thereof.

[0049] The memory 112 is a main storage device that stores the server program 11P, data, etc. For example, the server program 11P is loaded from the storage 113. The data includes data input to the server 11 and data generated by the processor 111. For example, the memory 112 is a RAM (Random Access Memory) or other volatile memory.

[0050] The storage 113 is an auxiliary storage device that stores the server program 11P and data, etc. The storage 113 is, for example, a read-only memory (ROM), a hard disk drive, a flash memory, or other non-volatile storage device. The storage 113 may also be a removable storage device such as a memory card. As another example, the storage 113 may be an external storage device. With this configuration, for example, in a situation where multiple user terminals 20 are used, such as an amusement facility, it becomes possible to collectively update the server program 11P or data.

[0051] The input / output interface 114 is an interface that connects external devices such as a monitor, an input device (for example, a keyboard or a pointing device), an external storage device, a speaker, a camera, a microphone, and a sensor to the server 11.

[0052] The processor 111 also communicates with external devices via an input / output interface 114. The input / output interface 114 is, for example, a Universal Serial Bus (USB), a Digital Visual Interface (DVI), a High-Definition Multimedia Interface (HDMI (registered trademark)), a wireless terminal, or other terminals.

[0053] The communication interface 115 communicates with other devices (such as the user terminal 20) connected to the communication network 2. For example, the communication interface 115 is a wired communication interface such as a LAN (Local Area Network), or a wireless communication interface such as Wi-Fi (registered trademark) (Wireless Fidelity), Bluetooth (registered trademark), or NFC (Near Field Communication).

[0054] However, the information processing device is not limited to the above hardware configuration. For example, the user terminal 20 may further include a sensor such as a camera. Various data acquired by the user terminal 20 using the sensor may be transmitted to the server 11.

[0055] [Examples of training model and trained model configurations] 7 is a network diagram showing an example of the configuration of an AI. The learning model A1 and the trained model A2 are, for example, AIs configured as shown in the following network.

[0056] Hereinafter, the learning model A1 and the trained model A2 will be described as being implemented on the server 11, i.e., on the cloud. However, part or all of the learning model A1 and the trained model A2 may be implemented on the user terminal 20, etc.

[0057] The network 300 has, for example, an input layer L1, an intermediate layer L2 (also called a "hidden layer"), and an output layer L3.

[0058] The input layer L1 is a layer that inputs data.

[0059] The intermediate layer L2 converts the data input at the input layer L1 based on weights (e.g., coefficients used for multiplication) and biases (e.g., adding a constant), etc. The results of this processing at the intermediate layer L2 are transmitted to the output layer L3.

[0060] The output layer L3 is a layer that outputs output contents and the like.

[0061] Then, through learning, weight coefficients (for example, coefficients for input characters or images are changed based on learning) and parameters changed through learning are optimized. Note that network 300 is not limited to the network structure shown in the figure. In other words, AI may be realized by other machine learning methods.

[0062] For example, the AI ​​may be configured to perform preprocessing such as dimensionality reduction (for example, a process that converts a relationship of three or more dimensions into a relationship that can be determined by simple calculations of three or less dimensions) using "unsupervised" machine learning. The relationship between input and output is preferably processed using simple calculations such as linear expressions. This type of calculation can reduce computational costs.

[0063] In addition, the AI ​​may undergo processes such as dropout to reduce overfitting (also called "overfitting" or "over-fitting"). Other preprocessing such as dimensionality reduction and normalization may also be performed.

[0064] The AI ​​may have a network structure such as a CNN (Convolution Neural Network). In addition, for example, the network structure may have a configuration such as an LLM (Large Language Model), an RNN (Recurrent Neural Network), or an LSTM (Long Short-Term Memory). In other words, the AI ​​may have a network structure other than deep learning.

[0065] The AI ​​may also have hyperparameters. That is, the AI ​​may be configured such that some of its settings are configured by a user. Furthermore, the AI ​​may specify the features to be learned, or the user may set some or all of the features to be learned.

[0066] Furthermore, the learning model A1 and the trained model A2 may use other machine learning methods. For example, the learning model A1 and the trained model A2 may be unsupervised models that undergo preprocessing such as normalization. Furthermore, learning may be reinforcement learning (a learning method in which an AI makes a choice and is given an evaluation (reward) for the choice, resulting in a larger evaluation), etc.

[0067] In the learning, data expansion may be performed. That is, in order to increase the amount of learning data used in learning the learning model A1, preprocessing may be performed to expand one piece of experimental data or the like into multiple pieces of learning data. Increasing the amount of learning data in this way allows for further learning of the learning model A1.

[0068] Furthermore, the learning model A1 and the trained model A2 may be configured to perform transfer learning or fine tuning. That is, since the user terminal 20 often has a different execution environment for each device, the settings for each device may differ depending on the execution environment. For example, the basic configuration of the AI ​​may be trained on a different information processing device. Thereafter, each information processing device may be further trained or configured to be further optimized for its respective execution environment.

[0069] [Example of evaluation of human-likeness] Human-likeness, for example, can be evaluated in terms of embodiment, allowing for accurate assessment that a person is a human, rather than an AI or a robot. For example, when attempting to impersonate someone using software, it is difficult for the person being evaluated to take physical actions, as they do not physically exist.

[0070] For example, when a person exercises, the body moves, generating vibrations and sounds to a certain extent. There are also other vital reactions. On the other hand, when the presence is disguised, there is no real human body, so sensing results indicating movement, etc. cannot be obtained.

[0071] Therefore, by evaluating whether or not there is physical behavior that can be recognized as movement, etc., it is possible to accurately determine whether or not the behavior is human-like. Note that the determination may also be made based on items other than those related to physical behavior.

[0072] However, human-likeness is not limited to physicality, but can be anything that is difficult for AI, robots, or software to deal with.

[0073] It is desirable to further evaluate whether the behavior is socially and generally desirable. For example, moderate exercise is a healthy behavior. In addition, participating in events such as volunteer activities is often a socially desirable behavior. Furthermore, actions such as picking up trash or purchasing environmentally friendly products are often desirable behaviors. If it can be evaluated that these actions were performed, the evaluation result may be a high evaluation. For example, the learning model A1 learns which behaviors are more desirable using big data D4, etc.

[0074] [Example of behavioral data] 8 is a diagram showing an example of input to the first trained model. For example, behavioral data D201 is input as unknown data D2 as follows:

[0075] The behavioral data D201 is generated in accordance with the behavior of the user when the user behaves on the user terminal 20. The unknown data D2 is generated based on an operation by the user 4 (for example, inputting an exercise record, etc.), automatically generated in the background by the user terminal 20 in accordance with the behavior (for example, the number of steps or walking speed of the user 4 using an acceleration sensor and a position sensor), or acquired and input from another app, etc.

[0076] The behavioral data D201 is acquired, for example, by a sensor or the like (including the analysis results of analyzing sensor data). The behavioral data may also be input, for example, by the user 4 through a GUI (Graphical User Interface), so-called "self-reporting." Alternatively, the behavioral data D201 may be acquired from another app, or may be obtained by analyzing the content of transmissions on a network such as an SNS (Social Networking Service).

[0077] Specifically, by analyzing sensor data such as acceleration sensors, position sensors using GNSS (Global Navigation Satellite System), cameras, or microphones, it is possible to determine, for example, whether user 4 is exercising, traveling in a car, or communicating with others.

[0078] If image data of a receipt taken when User 4 makes a purchase at a retail store can be acquired, User 4's purchase history can be entered.

[0079] Once application history data showing the application history on the application site or image data of tickets, etc., can be acquired, a participation behavior record of user 4 can be entered. The event may be a real-world event or an event such as an online seminar.

[0080] When speech made by user 4 can be acquired by a voice sensor such as a microphone or over the phone, a record of speech made by user 4 can be input. Also, when the voices of multiple people are recognized, a record of communication during an actual conversation can be input. Note that the communication record may also be communication on the web or via a chat app, etc.

[0081] When transmission data etc. sent by user 4 can be acquired via a chat app, SNS or the Internet, that is, a network, the transmission record of user 4 can be input.

[0082] The user 4 may input his / her own behavior through so-called "self-reporting." For example, the behavior may be input through a dedicated GUI for inputting behavior, or when the user 4 inputs a journal or the like through another application, the journal data may be analyzed and used as a self-reported record.

[0083] When User 4 plays a game, if a game operation record (for example, data on various games) can be acquired, the game operation record of User 4 can be input.

[0084] In addition, history of various logs and the like may be used as the behavioral data. For example, search history of a search engine, history of various apps, call history, meal history, browser access history, and the like may be input as the behavioral data.

[0085] Furthermore, through pre-processing, i.e., learning, using big data D4, the trained model A2 has learned what kind of behavior is human-like, event information, and specific examples of desirable behavior, based on information on the Internet. The learning may also utilize the results of image analysis of video. Therefore, the trained model A2 is in a state where it can evaluate what kind of behavior is desirable.

[0086] 9 is a diagram showing an example of input to the second trained model. For example, the second trained model A22 is input with the same unknown data D2 as the first trained model A21. However, auxiliary data (hereinafter referred to as "auxiliary data") for determining the authenticity of the behavior of user 4 may be further added to the second trained model A22.

[0087] The auxiliary data is, for example, data other than the unknown data D2 that is not copied and generated to determine whether or not the exercise behavior was performed, or other supporting data (for example, if the unknown data D2 is a self-reported record, it is sensor data from the time period when the exercise was reported to have been performed that supports the exercise), etc.

[0088] With such auxiliary data, the second trained model A22 can more accurately determine whether or not an exercise behavior has been performed.

[0089] [Execution processing example] 10 is a diagram showing an example of the execution process. Hereinafter, an example will be described in which the user 4 self-reports (for example, inputs text) "muscle training" and "talk to others" as unknown data D2.

[0090] The first trained model A21 generates first output data D31 indicating an evaluation result of "human-like" behavior. For example, the first output data D31 indicates a quantitative evaluation result indicating the probability of "human-like" behavior. However, if there is no probability and "human-like" behavior is performed with a high probability (for example, a probability of 50% or more), the first output data D31 may simply be an evaluation result such as "the behavior occurred."

[0091] The first output data D31 may include a qualitative evaluation. For example, the first output data D31 indicates details such as the type of exercise (type of exercise) for an exercise behavior. The first output data D31 may also include an evaluation result such as whether the behavior is desirable.

[0092] The second trained model A22 generates second output data D32 that indicates a judgment result indicating the truth or falsity of the actions taken by the user 4. In the following example, the second trained model A22 judges the truth or falsity of whether the actions "muscle training" and "talking to others" indicated in the unknown data D2 were actually performed.

[0093] The second output data D32 determines whether the behavior of "muscle training" was performed, for example. For example, the second trained model A22 acquires auxiliary data such as vibration data, heart rate data, and check-in data for the time period when the self-reported behavior was performed, and determines whether the behavior was performed based on the auxiliary data.

[0094] Similarly, the second output data D32 determines whether the behavior "talking to others" was performed. For example, the second trained model A22 obtains auxiliary data of voice data picked up by a microphone on the date and time of the self-report and determines whether the behavior is true or false based on the auxiliary data.

[0095] The auxiliary data used for the judgment may be used as evidential data. For example, for the action of "talking to others," audio data or the like may be left as evidential data and stored in the cloud, etc. Furthermore, if the AI ​​can identify the people who are talking, it may further generate information such as who is talking to whom or a summary of the conversation content.

[0096] As described above, the second trained model A22 has already learned what type of data should be acquired for a certain behavior, what judgment criteria should be set, etc. Therefore, the judgment method, the data to be used, the judgment criteria, etc. differ depending on the type of behavior, etc.

[0097] Note that advice may be further output for these actions. For example, if there is an action such as exercise or eating, and if there is a more health-appropriate exercise or dietary content, advice indicating the recommended exercise or dietary content may be output in text or the like.

[0098] [Example of conversion to in-game rewards] 11 is a diagram showing an example of conversion to an in-game reward. An input of a behavior may be converted into a reward in the game.

[0099] For example, when an exercise action is performed, the first output data D31 is converted into points that can be used in a game. The points become, for example, currency in the game.

[0100] The reward is not limited to points, but may be an item, a score, or an enhancement of a character's parameters.

[0101] The actions to be converted, the type, the number of points, the timing of receiving the reward, etc. can be set as appropriate. These may be set in advance, or may be selected arbitrarily by the user 4.

[0102] [Overall processing example] FIG. 12 is a diagram showing an example of the overall process. In the following example, the overall process is a series of pre-processing and execution processing. Specifically, step S01 is the learning process. Steps S02 to S05 are the execution processes. However, the overall process may include other processes.

[0103] In step S01, the server 11 inputs learning data D1 including behavioral data and correct answer data, and trains a first learning model A11 and a second learning model A12. Then, when step S01 is performed, a first trained model A21 and a second trained model A22 are generated. In this way, the learning model A1 is trained to become the trained model A2, and then the execution process is performed using the trained model A2.

[0104] In step S02, the server 11 inputs unknown data D2. For example, in step S02, the unknown data D2 is input by acquiring sensor data or by input from the user 4. Hereinafter, the time point at which the unknown data D2 is input is taken as a trigger for starting execution of subsequent processing (evaluation, etc.), but the start of subsequent processing, i.e., the timing of output, etc., may be a timing designated by the user 4.

[0105] In step S03, the server 11 evaluates the behavior, and then generates first output data D31 based on the evaluation result.

[0106] In step S04, the server 11 makes a determination as to whether the user's action is true or false, and then generates second output data D32 based on the determination result.

[0107] In step S05, the server 11 performs output based on the first output data D31 and the second output data D32. For example, the timing of output or the screen to be output, i.e., the timing and target for executing step S05, are set in advance.

[0108] [Example of functional configuration] 13 is a diagram illustrating an example of a functional configuration. For example, the system 1 is a personal AI system including a learning device 31 and an execution device 32.

[0109] The learning device 31 includes a first learning data input means 1F1, a first learning means 1F2, a second learning data input means 1F3, and a second learning means 1F4.

[0110] The first learning data input means 1F1 performs a first learning data input procedure for inputting learning data D1 including behavioral data and first supervised answer data D51. For example, the first learning data input means 1F1 is realized by the communication interface 115 or the like.

[0111] The first learning means 1F2 performs a first learning procedure in which the first learning model A11 is trained using the behavioral data and the first supervised data D51 to generate the first trained model A21. For example, the first learning means 1F2 is realized by the processor 111 or the like.

[0112] The second learning data input means 1F3 performs a second learning data input procedure for inputting the learning data D1 including the behavioral data and the second supervised answer data D52. For example, the second learning data input means 1F3 is realized by the communication interface 115 or the like.

[0113] The second learning means 1F4 performs a second learning procedure in which the second learning model A12 is trained using the behavioral data and the second supervised data D52 to generate the second trained model A22. For example, the second learning means 1F4 is realized by the processor 111 or the like.

[0114] The execution device 32 comprises a first unknown data input means 1F5, a first generation means 1F6, a second unknown data input means 1F7, a second generation means 1F8, and an output means 1F9. The execution device 32 preferably further comprises a conversion means 1F10.

[0115] The first unknown data input means 1F5 performs a first unknown data input procedure for inputting unknown data D2 including unknown behavioral data to the first trained model A21 generated by the first learning means 1F2. For example, the first unknown data input means 1F5 is realized by the communication interface 115 or the like.

[0116] When the unknown data D2 is input, the first generation means 1F6 performs a first generation procedure to generate first output data D31 based on the unknown data D2. For example, the first generation means 1F6 is realized by the processor 111 or the like.

[0117] The second unknown data input means 1F7 performs a second unknown data input procedure for inputting unknown data D2 including unknown behavioral data to the second trained model A22 generated by the second learning means 1F4. For example, the second unknown data input means 1F7 is realized by the communication interface 115 or the like.

[0118] When the unknown data D2 is input, the second generation means 1F8 performs a second generation procedure to generate second output data D32 based on the unknown data D2. For example, the second generation means 1F8 is realized by the processor 111 or the like.

[0119] The output means 1F9 performs an output procedure based on the first output data D31 and the second output data D32 to output to the user terminal 20. For example, the output means 1F9 is realized by the communication interface 115 or the like.

[0120] The conversion means 1F10 performs a conversion procedure for converting the evaluation result into a reward in the game. For example, the conversion means 1F10 is realized by the processor 111 or the like.

[0121] The learning device 31 and the execution device 32 are, for example, the server 11. However, the learning device 31 and the execution device 32 may be different information processing devices.

[0122] With the above configuration, it is possible to objectively evaluate whether the behavior of user 4 is "human-like." First, when user 4 takes an action, unknown data D2 is input. Next, the first trained model A21 evaluates the behavior of user 4 based on the unknown data D2. Therefore, the AI ​​evaluates whether the behavior is difficult for anyone other than user 4, i.e., a human, to perform.

[0123] Next, based on the unknown data D2, the second trained model A22 determines the authenticity of the behavior of the user 4. For example, the second trained model A22 determines whether there is any fraudulent activity, such as the behavior data being fabricated or the actual behavior being different from the declared behavior.

[0124] For example, the determination is "true" if none of the multiple auxiliary data indicates any wrongdoing, whereas the determination is "false" if the behavior indicated by the auxiliary data differs from the declared behavior.

[0125] Fraudulent acts include having AI, software, or robots act on behalf of users 4, or inputting data that makes it appear as if an action has been taken when in fact no action has been taken.

[0126] Therefore, AI is used to evaluate whether there is any behavior that is difficult to replace with software, i.e., "human-like" behavior, and to determine whether the behavior being evaluated is free of misconduct. Once such evaluation and determination are made, it becomes possible to objectively evaluate whether the behavior is "human-like."

[0127] The type of game is not limited as long as the game provides reward points, items, etc. For example, the type of game may include RPG, sports, action, puzzle, or fighting games.

[0128] [Other embodiments] In the above example, the information processing device performs both pre-processing for the learning model and execution processing using the learned model. However, the pre-processing and execution processing do not have to be performed by the same information processing device. Furthermore, the pre-processing and execution processing do not have to be consistently performed by a single information processing device. In other words, each process and data storage may be performed by an information system or the like composed of multiple information processing devices.

[0129] The learning process may be additionally performed after the execution process or before the execution process.

[0130] The above-mentioned processing may be performed by an information processing device other than the server 11 and the user terminal 20 in an auxiliary manner.

[0131] Fig. 14 is a diagram showing an example of a configuration using an auxiliary device. Compared to the example shown in Fig. 1, the configuration shown in Fig. 14 differs in that an auxiliary device 60 is added. Note that the auxiliary device 60 may be a configuration that is used temporarily.

[0132] The auxiliary device 60 is an information processing device installed near the user terminal 20 (in this example, it is installed near the user terminal 20A, but it may be installed near another device.) The auxiliary device 60 executes a part or all of a specific process on behalf of the user terminal 20 or the server 11.

[0133] For example, the auxiliary device 60 may be equipped with a device specialized for graphic processing and perform graphic processing at high speed. In this way, so-called edge computing may be performed by installing the auxiliary device 60 or the like. In this way, the above-described processing may be performed by utilizing the hardware resources of various information processing devices. Therefore, the above-described processing may be performed by an information processing device different from the above-described one.

[0134] The above-described processes and data used in the processes executed in this embodiment may be executed and stored by an information processing system. For example, the information processing system may execute or store data on multiple information processing devices to achieve redundant, distributed, parallel, or a combination thereof. Therefore, the present invention may be realized in devices with hardware configurations other than those described above and in systems other than those described above.

[0135] Furthermore, the program according to the present invention is not limited to a single program, but may be a collection of multiple programs. Furthermore, the program according to the present invention is not limited to being executed by a single device, but may be executed by multiple information processing devices in a shared manner. Furthermore, the allocation of roles among the information processing devices is not limited to the above-mentioned example. In other words, some or all of the above-mentioned processes may be executed by information processing devices different from the above-mentioned information processing device.

[0136] Furthermore, some or all of the means implemented by the program can be realized by hardware such as an integrated circuit. Furthermore, the program may be provided by being recorded on a non-transitory recording medium readable by a computer. Examples of the recording medium include a hard disk, an SD card (registered trademark), an optical disk such as a DVD, or a server on the Internet. Therefore, the program may be distributed via a telecommunications line such as the Internet.

[0137] Furthermore, the information processing devices that make up the information processing system may be located overseas. That is, an information processing device that executes some of the processes executed by the information processing system may be located overseas.

[0138] The present invention is not limited to the above-described exemplary embodiments. Therefore, the present invention may be modified or added to within the scope of the technical gist. Therefore, all technical matters included in the technical concept described in the claims are subject to the present invention. The above-described exemplary embodiments are preferred examples. Those skilled in the art will be able to realize various modifications from the disclosed content, and such modifications are included in the technical scope described in the claims. [Explanation of symbols]

[0139] 1: System 1F1: First learning data input means 1F2: First learning method 1F3: Second learning data input means 1F4: Second learning method 1F5: First unknown data input means 1F6: 1st generation means 1F7: Second unknown data input means 1F8:Second generation means 1F9: Output method 1F10: Conversion method 3:Administrator 4: User 5:Administrator 11: Server 20: User terminal 31: Learning device 32: Execution device 60: Auxiliary equipment D1: Training data D2: Unknown data D20: Correct data D4: Big Data A1: Learning model A2: Pre-trained model A1: Learning model A11: First learning model A12: Second learning model A21: First trained model A22: Second trained model D1: Training data D11: 11th data D12: 12th data D13: 13th data D2: Unknown data D20: Correct data D201: Behavioral data D21: 21st data D22: 22nd data D23: 23rd data D3: Output data D31: First output data D32: Second output data D4: Big Data D51: First correct data D52: Second correct data

Claims

1. Computer, a generation means for generating output data indicating at least a determination result of the truth or falsity of the user's behavior when unknown data relating to the user's behavior involving physicality is input to a trained model that has learned the user's behavior data; an output means for outputting the output data to the trained model; and make it work, The trained model is The system learns a correlation between the behavior and the truth or falsity of the behavior based on sensor data recorded by a sensor or an input of the behavior, When the unknown data is input, the truth or falsity of the behavior is output based on the correlation. program.

2. Computer, a generation means for generating output data indicating at least an evaluation result of the user's behavior when unknown data relating to the user's physical behavior is input to a trained model that has learned the user's behavior data; and an output means for outputting the output data to the trained model; and make it work, The trained model is learning a correlation between the behavior and an evaluation of human-likeness based on sensor data recorded by a sensor or an input of the behavior; When the unknown data is input, an evaluation of human-likeness is output based on the correlation. program.

3. The computer further comprises: a reward generating unit configured to generate a reward corresponding to the output data and to grant the generated reward to the user; 3. The program according to claim 1 or 2, which functions as

Images