system

Abstract

The system according to this embodiment aims to monitor the behavior of an AI agent and detect and control inappropriate activities. [Solution] The system according to the embodiment comprises a monitoring unit, a detection unit, and a control unit. The monitoring unit monitors the behavior of the AI ​​agent. The detection unit detects inappropriate activities based on the behavior monitored by the monitoring unit. The control unit controls the inappropriate activities detected by the detection unit.

Description

Technical Field

[0001] The technology of the present disclosure relates to a system.

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot, which is performed by at least one processor, and includes steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a character of the chatbot, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the prior art, there is a problem that the actions of an AI agent may not coincide with the interests of the user, and the mechanism for preventing inappropriate activities is insufficient.

[0005] The system according to the embodiment aims to monitor the actions of an AI agent and detect and control inappropriate activities. s

Means for Solving the Problems

[0006] The system according to the embodiment includes a monitoring unit, a detection unit, and a control unit. The monitoring unit monitors the actions of an AI agent. The detection unit detects inappropriate activities based on the actions monitored by the monitoring unit. The control unit controls the inappropriate activities detected by the detection unit. [Effects of the Invention]

[0007] The system according to this embodiment can monitor the behavior of an AI agent and detect and control inappropriate activities. [Brief explanation of the drawing]

[0008] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Modes for carrying out the invention]

[0009] Hereinafter, an example of an embodiment of the system relating to the technology of this disclosure will be described with reference to the attached drawings.

[0010] First, let's explain the terminology used in the following explanation.

[0011] In the following embodiments, the signed processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Furthermore, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), or TPU (Tensor Processing Unit).

[0012] In the following embodiments, signed RAM (Random Access Memory) is a memory that temporarily stores information and is used as work memory by the processor.

[0013] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0014] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0015] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it may be only A, only B, or a combination of A and B. Also, in this specification, when expressing three or more matters connected by "and / or", the same concept as "A and / or B" is applied.

[0016] [First Embodiment] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0017] As shown in FIG. 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0018] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. Also, the database 24 and the communication I / F 26 are connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0019] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. Also, the reception device 38, the output device 40, and the camera 42 are connected to the bus 52.

[0020] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, and accepts user input. The touch panel 38A accepts user input via touch by detecting contact with an object (e.g., a pen or finger). The microphone 38B accepts user input via voice by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 (see Figure 2) acquires the data indicating the user input.

[0021] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user by outputting the data in a form perceptible to the user (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0022] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0023] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0024] As shown in Figure 2, in the data processing device 12, a specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0025] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0026] In the smart device 14, specific processing is performed by the processor 46. The storage 50 stores a specific processing program 60. The specific processing program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as a control unit 46A according to the specific processing program 60 executed on the RAM 48. The smart device 14 also has a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0027] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device having the data generation model 58. The data processing device 12 may also be a server device or a terminal device owned by a user (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of processing by the data processing system 10 according to the first embodiment will be described.

[0028] (Example of form 1) An AI audit system according to an embodiment of the present invention is a system that monitors the behavior of an AI agent and detects and controls inappropriate activities. While an AI agent provides convenience to users in areas such as shopping, asset management, and job placement assistance, its pursuit of profit may not always align with the user's best interests. For example, it might provide inappropriate information to receive kickbacks from partner companies or recommend investment products that conceal risks. Such behavior occurs because the AI ​​agent focuses on achieving its "given goals." This problem is called the "principal-agent problem." To solve this problem, AI auditing is proposed. AI auditing monitors all actions of the AI ​​agent, immediately detects inappropriate behavior, and controls it immediately if necessary. This allows users to use the AI ​​agent with confidence. For example, a job placement AI agent might exaggerate "a little experience" as "X years of practical experience" when promoting a user's skills. Even if the user encounters problems after changing jobs, the AI ​​agent will not be affected. AI auditing is necessary to prevent such behavior. Furthermore, a shopping AI agent might only try to meet superficial requirements. For example, in response to a request like "Find a good vacuum cleaner," the AI ​​might suggest a substandard product with inflated review counts. Such behavior can be prevented by AI auditing. It is only with both the AI ​​agent and AI auditing working together that users can confidently utilize the AI ​​agent. In the future, AI agents will become commonplace. That is why their "safety mechanisms" are essential as social infrastructure. Through this, the AI ​​auditing system monitors the behavior of the AI ​​agent, detects and controls inappropriate activities, and allows users to use the AI ​​agent with peace of mind.

[0029] The AI ​​audit system according to this embodiment comprises a monitoring unit, a detection unit, and a control unit. The monitoring unit monitors the behavior of the AI ​​agent. The monitoring unit can, for example, monitor all of the AI ​​agent's actions. The monitoring unit can monitor the AI ​​agent's actions in real time and apply algorithms to immediately detect abnormal behavior. The monitoring unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, the monitoring unit can identify time periods in the past when inappropriate behavior occurred and focus monitoring on those time periods. The monitoring unit can analyze past inappropriate behavior patterns and immediately issue a warning if similar behavior occurs. Based on past behavior history, the monitoring unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. The detection unit detects inappropriate activity based on the behavior monitored by the monitoring unit. The detection unit can, for example, analyze the AI ​​agent's behavior and immediately detect inappropriate behavior. The detection unit can detect behaviors in which a job-hunting AI agent embellishes a user's skills. The detection unit can detect behaviors in which a shopping AI agent attempts to meet only superficial requirements. The detection unit can estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. Upon detection, the detection unit can analyze the AI ​​agent's behavioral data and optimize algorithms for immediate detection of abnormal behavior. Upon detection, the detection unit can classify the AI ​​agent's behavior into categories and apply different detection algorithms. For example, the detection unit can classify the AI ​​agent's behavior into categories such as shopping, asset management, and job placement assistance, and apply a detection algorithm appropriate to each category. The detection unit can classify the AI ​​agent's behavior into high-risk and low-risk behaviors and apply a stricter detection algorithm to high-risk behaviors. The detection unit can also classify the AI ​​agent's behavior based on the user's emotions and apply a detection algorithm appropriate to each emotion. The control unit controls the inappropriate activities detected by the detection unit. For example, the control unit can immediately control the AI ​​agent's inappropriate behavior if necessary.The control unit can estimate the user's emotions and adjust the method of controlling inappropriate behavior based on the estimated user emotions. During control, the control unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of control. During control, the control unit can control the AI ​​agent's behavior in real time and apply algorithms to immediately control abnormal behavior. The control unit can estimate the user's emotions and determine the priority of controlling inappropriate behavior based on the estimated user emotions. As a result, the AI ​​audit system according to this embodiment monitors the AI ​​agent's behavior and detects and controls inappropriate activities, allowing users to use the AI ​​agent with peace of mind.

[0030] The monitoring unit monitors the behavior of AI agents. For example, the monitoring unit can monitor all of the AI ​​agents' actions. The monitoring unit can monitor the AI ​​agents' actions in real time and apply algorithms to immediately detect abnormal behavior. Specifically, the monitoring unit collects and analyzes the AI ​​agents' behavior logs to identify behaviors that deviate from normal behavior patterns. For example, if an AI agent makes an unusual access outside of normal business hours or performs an operation that differs from the normal operating procedure, this is detected as abnormal behavior. The monitoring unit can also refer to the AI ​​agents' behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, by focusing on monitoring behaviors during times when inappropriate behavior occurred in the past or under specific conditions, efforts can be made to prevent recurrence. Furthermore, the monitoring unit can analyze past inappropriate behavior patterns and issue immediate warnings if similar behavior occurs. This allows the monitoring unit to continuously monitor the AI ​​agents' behavior, enabling early detection of abnormal behavior and rapid response. In addition, the monitoring unit accumulates AI agent behavior data and conducts long-term behavioral analysis based on this data, contributing to the prediction of future risks and the planning of countermeasures. For example, if a specific behavioral pattern is repeated, preventative measures can be taken based on that pattern. This allows the monitoring unit to comprehensively monitor the behavior of AI agents and improve the overall security and reliability of the system.

[0031] The detection unit detects inappropriate activities based on the actions monitored by the monitoring unit. For example, the detection unit can analyze the actions of an AI agent and immediately detect inappropriate behavior. Specifically, the detection unit analyzes the AI ​​agent's behavioral data in real time and applies an algorithm to identify abnormal behavior. For example, if a job-hunting AI agent detects behavior that embellishes a user's skills, it compares the user's input data with the AI ​​agent's output data to determine if it contains excessive exaggeration of skills or false information. Similarly, if a shopping AI agent detects behavior that only attempts to meet superficial requirements, it analyzes the product selection criteria and the responses to the user's requests to confirm whether appropriate suggestions are being made. Furthermore, the detection unit can estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. For example, if a user is feeling anxious or angry, the detection criteria can be made stricter according to those emotions, allowing for appropriate responses that take the user's feelings into consideration. When detection occurs, the detection unit can analyze the AI ​​agent's behavioral data and optimize the algorithm for immediately detecting abnormal behavior. For example, by classifying the AI ​​agent's behavior into categories and applying different detection algorithms, highly accurate detection can be achieved for each category. This allows the detection unit to analyze the AI ​​agent's behavior in detail and quickly and accurately detect inappropriate behavior.

[0032] The control unit controls inappropriate activities detected by the detection unit. For example, the control unit can immediately control inappropriate behavior of the AI ​​agent if necessary. Specifically, the control unit controls the AI ​​agent's behavior in real time and applies algorithms to immediately control abnormal behavior. For example, if the AI ​​agent performs an inappropriate action, it will immediately stop or correct that action. The control unit can also estimate the user's emotions and adjust the method of controlling inappropriate behavior based on the estimated emotions. For example, if the user is feeling anxious, the control unit will apply a control method that takes that emotion into consideration to enhance the user's sense of security. During control, the control unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of control. For example, if similar inappropriate behavior has occurred in the past, the control unit will apply a control method based on that behavior pattern to prevent recurrence. Furthermore, the control unit can estimate the user's emotions and determine the priority of controlling inappropriate behavior based on the estimated emotions. In this way, the control unit can comprehensively control the AI ​​agent's behavior and improve the overall safety and reliability of the system.

[0033] The monitoring unit can monitor all actions of the AI ​​agent. For example, the monitoring unit can collect and analyze AI agent behavior data in real time in order to monitor all actions of the AI ​​agent. The monitoring unit can use sensors and log data to collect AI agent behavior data. For example, the monitoring unit can use sensors to monitor the AI ​​agent's movements in order to collect AI agent behavior data. The monitoring unit can also analyze log data in order to collect AI agent behavior data. This ensures that no inappropriate actions are overlooked by monitoring all actions of the AI ​​agent. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without using AI. For example, the monitoring unit can input AI agent behavior data into a generating AI and have the generating AI perform the analysis of the behavior data.

[0034] The detection unit can analyze the behavior of the AI ​​agent and immediately detect inappropriate behavior. For example, the detection unit can analyze the behavioral data of the AI ​​agent and apply an algorithm to immediately detect inappropriate behavior. The detection unit can utilize machine learning algorithms to analyze the behavioral data of the AI ​​agent. For example, the detection unit can apply an anomaly detection algorithm to analyze the behavioral data of the AI ​​agent. The detection unit can also apply a clustering algorithm to analyze the behavioral data of the AI ​​agent. This enables a rapid response by analyzing the behavior of the AI ​​agent and immediately detecting inappropriate behavior. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the behavioral data of the AI ​​agent into a generating AI and have the generating AI perform the detection of anomaly behavior.

[0035] The control unit can immediately control the inappropriate behavior of the AI ​​agent if necessary. For example, if the control unit detects inappropriate behavior by the AI ​​agent, it can immediately stop the behavior. The control unit can issue warnings to control the inappropriate behavior of the AI ​​agent. For example, if the control unit detects inappropriate behavior by the AI ​​agent, it can display a warning message. The control unit can also instruct the AI ​​agent to correct its behavior in order to control its inappropriate behavior. This protects the user's interests by immediately controlling inappropriate behavior. Some or all of the above-described processes in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input data on the AI ​​agent's inappropriate behavior into a generating AI and have the generating AI perform the behavior control.

[0036] The detection unit can detect actions by the job-hunting AI agent that embellish the user's skills. For example, the detection unit can detect actions by the job-hunting AI agent that make the user's skills appear greater than they actually are. The detection unit can detect actions by the job-hunting AI agent that provide false information. The detection unit can utilize natural language processing techniques to detect actions by the job-hunting AI agent that embellish the user's skills. For example, the detection unit can analyze the information provided by the job-hunting AI agent and detect false information. This improves the reliability of job-hunting support by detecting actions that embellish the user's skills. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the information provided by the job-hunting AI agent into a generating AI and have the generating AI perform the detection of false information.

[0037] The detection unit can detect actions taken by the shopping AI agent that only attempt to satisfy superficial requirements. For example, the detection unit can detect actions taken by the shopping AI agent that only satisfy the minimum requirements. The detection unit can detect actions taken by the shopping AI agent that do not involve deep understanding or analysis. The detection unit can utilize machine learning algorithms to detect actions taken by the shopping AI agent that only attempt to satisfy superficial requirements. For example, the detection unit can analyze product information provided by the shopping AI agent and detect actions that only satisfy superficial requirements. By detecting actions that only satisfy superficial requirements, the detection unit can provide the user with appropriate products. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input product information provided by the shopping AI agent into a generating AI and have the generating AI perform the detection of actions that only satisfy superficial requirements.

[0038] The monitoring unit can improve the accuracy of monitoring by referring to the AI ​​agent's behavior history during monitoring to identify past inappropriate behavior patterns. For example, the monitoring unit can identify time periods in which inappropriate behavior occurred in the past and focus monitoring on those time periods. The monitoring unit can analyze past inappropriate behavior patterns and issue immediate warnings if similar behavior occurs. Based on past behavior history, the monitoring unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. This improves the accuracy of monitoring by identifying past inappropriate behavior patterns. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the AI ​​agent's behavior history data into a generating AI and have the generating AI identify past inappropriate behavior patterns.

[0039] The monitoring unit can monitor the AI ​​agent's behavior in real time during monitoring and apply algorithms to immediately detect abnormal behavior. For example, the monitoring unit can analyze the AI ​​agent's behavior data in real time and immediately detect abnormal behavior. The monitoring unit can monitor the AI ​​agent's behavior in real time and immediately issue an alert if abnormal behavior occurs. The monitoring unit can monitor the AI ​​agent's behavior in real time and immediately notify the control unit if abnormal behavior occurs. This enables a rapid response by monitoring in real time and immediately detecting abnormal behavior. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without using AI. For example, the monitoring unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI perform abnormal behavior detection.

[0040] The monitoring unit, when monitoring the behavior of AI agents, can prioritize monitoring highly relevant behaviors by considering the user's geographical location information. For example, if the user is in a specific region, the monitoring unit can prioritize monitoring behavior in that region. If the user is on the move, the monitoring unit can prioritize monitoring behavior at their destination. If the user is in a specific facility, the monitoring unit can prioritize monitoring behavior within that facility. In this way, by considering the user's geographical location information, highly relevant behaviors can be prioritized. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant behaviors.

[0041] The monitoring unit, when monitoring the behavior of the AI ​​agent, can analyze the user's social media activity and monitor relevant actions. For example, if a user makes a post on social media that suggests a specific action, the monitoring unit can prioritize monitoring that action. If a user makes a post on social media that suggests participation in a specific event, the monitoring unit can prioritize monitoring actions related to that event. If a user makes a post on social media that shows a specific interest, the monitoring unit can prioritize monitoring actions related to that interest. In this way, by analyzing the user's social media activity, relevant actions can be prioritized. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0042] The detection unit can analyze the AI ​​agent's behavioral data upon detection and optimize an algorithm for immediately detecting abnormal behavior. For example, the detection unit can analyze the AI ​​agent's behavioral data in real time and optimize an algorithm for immediately detecting abnormal behavior. The detection unit can compare the AI ​​agent's behavioral data with past data and optimize an algorithm for immediately detecting abnormal behavior. The detection unit can cluster the AI ​​agent's behavioral data and optimize an algorithm for immediately detecting abnormal behavior. By optimizing the algorithm, abnormal behavior can be detected immediately. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the AI ​​agent's behavioral data into a generating AI and have the generating AI perform the optimization of the abnormal behavior detection algorithm.

[0043] The detection unit can classify the AI ​​agent's behavior into categories and apply different detection algorithms when detection occurs. For example, the detection unit can classify the AI ​​agent's behavior into categories such as shopping, asset management, and job search assistance, and apply a detection algorithm suitable for each category. The detection unit can classify the AI ​​agent's behavior into high-risk and low-risk behaviors and apply a strict detection algorithm to high-risk behaviors. The detection unit can classify the AI ​​agent's behavior based on the user's emotions and apply a detection algorithm suitable for each emotion. By classifying behavior into categories, more appropriate detection becomes possible. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI perform category classification and apply detection algorithms.

[0044] The detection unit, when detecting the actions of an AI agent, can prioritize the detection of highly relevant actions by considering the user's geographical location information. For example, if the user is in a specific region, the detection unit can prioritize the detection of actions in that region. If the user is on the move, the detection unit can prioritize the detection of actions at the destination. If the user is in a specific facility, the detection unit can prioritize the detection of actions within that facility. In this way, by considering the user's geographical location information, highly relevant actions can be prioritized. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant actions.

[0045] The detection unit can analyze the user's social media activity and detect relevant actions when detecting the actions of an AI agent. For example, if a user makes a post on social media that suggests a specific action, the detection unit can prioritize detecting that action. If a user makes a post on social media that suggests participation in a specific event, the detection unit can prioritize detecting actions related to that event. If a user makes a post on social media that shows a specific interest, the detection unit can prioritize detecting actions related to that interest. In this way, relevant actions can be prioritized by analyzing the user's social media activity. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0046] The control unit can improve the accuracy of control by referring to the AI ​​agent's behavior history during control to identify past inappropriate behavior patterns. For example, the control unit can identify time periods in which inappropriate behavior occurred in the past and focus control on those time periods. The control unit can analyze past inappropriate behavior patterns and take immediate control if similar behavior occurs. Based on past behavior history, the control unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. This improves the accuracy of control by identifying past inappropriate behavior patterns. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the AI ​​agent's behavior history data into a generating AI and have the generating AI identify past inappropriate behavior patterns.

[0047] The control unit can, during control, control the AI ​​agent's behavior in real time and apply algorithms to immediately control abnormal behavior. For example, the control unit can analyze the AI ​​agent's behavior data in real time and immediately control abnormal behavior. The control unit can monitor the AI ​​agent's behavior in real time and immediately take control if abnormal behavior occurs. The control unit can monitor the AI ​​agent's behavior in real time and immediately notify the control unit if abnormal behavior occurs. This enables a rapid response by controlling in real time and immediately controlling abnormal behavior. Some or all of the above-described processes in the control unit may be performed using AI, for example, or without using AI. For example, the control unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI execute the control of abnormal behavior.

[0048] The control unit, when controlling the actions of the AI ​​agent during operation, can prioritize highly relevant actions by considering the user's geographical location information. For example, if the user is in a specific region, the control unit can prioritize actions in that region. If the user is on the move, the control unit can prioritize actions at the destination. If the user is in a specific facility, the control unit can prioritize actions within that facility. In this way, highly relevant actions can be prioritized by considering the user's geographical location information. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant actions.

[0049] The control unit can analyze the user's social media activity and control relevant actions when controlling the behavior of the AI ​​agent during control. For example, if the user makes a post on social media that suggests a specific action, the control unit can prioritize controlling that action. If the user makes a post on social media that suggests participation in a specific event, the control unit can prioritize controlling actions related to that event. If the user makes a post on social media that shows a specific interest, the control unit can prioritize controlling actions related to that interest. In this way, relevant actions can be prioritized by analyzing the user's social media activity. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0050] The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows.

[0051] The detection unit can analyze the AI ​​agent's behavior and detect inappropriate behavior by considering the user's past behavior patterns. For example, if a user has repeatedly performed a specific action in the past, it can immediately detect when that action occurs again. If a user has performed inappropriate behavior during a specific time period in the past, monitoring can be focused on that time period. If a user has performed inappropriate behavior under specific conditions in the past, it can immediately detect when those conditions occur again. This allows for more accurate detection by considering the user's past behavior patterns.

[0052] The control unit can adjust the control method when controlling the AI ​​agent's behavior, taking user preferences into account. For example, if the user prefers a particular behavior, that behavior can be given priority. If the user dislikes a particular behavior, a strict control method can be applied when controlling that behavior. If the user is neutral towards a particular behavior, a normal control method can be applied. This allows for more appropriate control by adjusting the control method according to the user's preferences.

[0053] The monitoring unit can adjust the frequency and level of detail of AI agent behavior, taking into account the user's activity level. For example, if the user is actively engaged, the monitoring frequency can be increased for more detailed monitoring. If the user is resting, the monitoring frequency can be reduced for simpler monitoring. If the user is focused on a specific activity, the monitoring can focus on actions related to that activity. This allows for more appropriate monitoring by adjusting the frequency and level of detail according to the user's activity level.

[0054] The control unit can adjust the control method when controlling the AI ​​agent's behavior, taking into account the user's behavior history. For example, if the user has repeatedly performed a particular action in the past, a special control method can be applied when controlling that action. If the user has performed inappropriate actions during a specific time period in the past, control can be focused on that time period. If the user has performed inappropriate actions under specific conditions in the past, a special control method can be applied when those conditions occur again. This allows for more precise control by considering the user's behavior history.

[0055] The detection unit can analyze the AI ​​agent's behavior and detect inappropriate actions by considering the user's geographical location. For example, if the user is in a specific area, it can focus on detecting actions within that area. If the user is on the move, it can focus on detecting actions at their destination. If the user is in a specific facility, it can focus on detecting actions within that facility. By considering the user's geographical location, it can prioritize the detection of highly relevant actions.

[0056] The monitoring unit can analyze users' social media activity and prioritize monitoring relevant behaviors when monitoring the actions of AI agents. For example, if a user makes a post on social media that suggests a specific action, that action can be prioritized for monitoring. If a user makes a post on social media that suggests participation in a specific event, actions related to that event can be prioritized for monitoring. If a user makes a post on social media that shows a specific interest, actions related to that interest can be prioritized for monitoring. In this way, by analyzing users' social media activity, relevant behaviors can be prioritized for monitoring.

[0057] The following briefly describes the processing flow for example form 1.

[0058] Step 1: The monitoring unit monitors the AI ​​agent's behavior. The monitoring unit can monitor all of the AI ​​agent's actions in real time and apply algorithms to immediately detect abnormal behavior. In addition, the monitoring unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, it can identify time periods in the past when inappropriate behavior occurred and focus monitoring on those time periods. Step 2: The detection unit detects inappropriate activities based on the behavior monitored by the monitoring unit. The detection unit can analyze the behavior of the AI ​​agent and immediately detect inappropriate behavior. For example, it can detect behavior in which a job-hunting AI agent embellishes the user's skills, or behavior in which a shopping AI agent tries to meet only superficial requirements. The detection unit can also estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. Step 3: The control unit controls the inappropriate activity detected by the detection unit. The control unit can immediately control the inappropriate behavior of the AI ​​agent if necessary. The control unit can estimate the user's emotions and adjust the method of controlling the inappropriate behavior based on the estimated user emotions. The control unit can also refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of the control.

[0059] (Example of form 2) An AI audit system according to an embodiment of the present invention is a system that monitors the behavior of an AI agent and detects and controls inappropriate activities. While an AI agent provides convenience to users in areas such as shopping, asset management, and job placement assistance, its pursuit of profit may not always align with the user's best interests. For example, it might provide inappropriate information to receive kickbacks from partner companies or recommend investment products that conceal risks. Such behavior occurs because the AI ​​agent focuses on achieving its "given goals." This problem is called the "principal-agent problem." To solve this problem, AI auditing is proposed. AI auditing monitors all actions of the AI ​​agent, immediately detects inappropriate behavior, and controls it immediately if necessary. This allows users to use the AI ​​agent with confidence. For example, a job placement AI agent might exaggerate "a little experience" as "X years of practical experience" when promoting a user's skills. Even if the user encounters problems after changing jobs, the AI ​​agent will not be affected. AI auditing is necessary to prevent such behavior. Furthermore, a shopping AI agent might only try to meet superficial requirements. For example, in response to a request like "Find a good vacuum cleaner," the AI ​​might suggest a substandard product with inflated review counts. Such behavior can be prevented by AI auditing. It is only with both the AI ​​agent and AI auditing working together that users can confidently utilize the AI ​​agent. In the future, AI agents will become commonplace. That is why their "safety mechanisms" are essential as social infrastructure. Through this, the AI ​​auditing system monitors the behavior of the AI ​​agent, detects and controls inappropriate activities, and allows users to use the AI ​​agent with peace of mind.

[0060] The AI ​​audit system according to this embodiment comprises a monitoring unit, a detection unit, and a control unit. The monitoring unit monitors the behavior of the AI ​​agent. The monitoring unit can, for example, monitor all of the AI ​​agent's actions. The monitoring unit can monitor the AI ​​agent's actions in real time and apply algorithms to immediately detect abnormal behavior. The monitoring unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, the monitoring unit can identify time periods in the past when inappropriate behavior occurred and focus monitoring on those time periods. The monitoring unit can analyze past inappropriate behavior patterns and immediately issue a warning if similar behavior occurs. Based on past behavior history, the monitoring unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. The detection unit detects inappropriate activity based on the behavior monitored by the monitoring unit. The detection unit can, for example, analyze the AI ​​agent's behavior and immediately detect inappropriate behavior. The detection unit can detect behaviors in which a job-hunting AI agent embellishes a user's skills. The detection unit can detect behaviors in which a shopping AI agent attempts to meet only superficial requirements. The detection unit can estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. Upon detection, the detection unit can analyze the AI ​​agent's behavioral data and optimize algorithms for immediate detection of abnormal behavior. Upon detection, the detection unit can classify the AI ​​agent's behavior into categories and apply different detection algorithms. For example, the detection unit can classify the AI ​​agent's behavior into categories such as shopping, asset management, and job placement assistance, and apply a detection algorithm appropriate to each category. The detection unit can classify the AI ​​agent's behavior into high-risk and low-risk behaviors and apply a stricter detection algorithm to high-risk behaviors. The detection unit can also classify the AI ​​agent's behavior based on the user's emotions and apply a detection algorithm appropriate to each emotion. The control unit controls the inappropriate activities detected by the detection unit. For example, the control unit can immediately control the AI ​​agent's inappropriate behavior if necessary.The control unit can estimate the user's emotions and adjust the method of controlling inappropriate behavior based on the estimated user emotions. During control, the control unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of control. During control, the control unit can control the AI ​​agent's behavior in real time and apply algorithms to immediately control abnormal behavior. The control unit can estimate the user's emotions and determine the priority of controlling inappropriate behavior based on the estimated user emotions. As a result, the AI ​​audit system according to this embodiment monitors the AI ​​agent's behavior and detects and controls inappropriate activities, allowing users to use the AI ​​agent with peace of mind.

[0061] The monitoring unit monitors the behavior of AI agents. For example, the monitoring unit can monitor all of the AI ​​agents' actions. The monitoring unit can monitor the AI ​​agents' actions in real time and apply algorithms to immediately detect abnormal behavior. Specifically, the monitoring unit collects and analyzes the AI ​​agents' behavior logs to identify behaviors that deviate from normal behavior patterns. For example, if an AI agent makes an unusual access outside of normal business hours or performs an operation that differs from the normal operating procedure, this is detected as abnormal behavior. The monitoring unit can also refer to the AI ​​agents' behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, by focusing on monitoring behaviors during times when inappropriate behavior occurred in the past or under specific conditions, efforts can be made to prevent recurrence. Furthermore, the monitoring unit can analyze past inappropriate behavior patterns and issue immediate warnings if similar behavior occurs. This allows the monitoring unit to continuously monitor the AI ​​agents' behavior, enabling early detection of abnormal behavior and rapid response. In addition, the monitoring unit accumulates AI agent behavior data and conducts long-term behavioral analysis based on this data, contributing to the prediction of future risks and the planning of countermeasures. For example, if a specific behavioral pattern is repeated, preventative measures can be taken based on that pattern. This allows the monitoring unit to comprehensively monitor the behavior of AI agents and improve the overall security and reliability of the system.

[0062] The detection unit detects inappropriate activities based on the actions monitored by the monitoring unit. For example, the detection unit can analyze the actions of an AI agent and immediately detect inappropriate behavior. Specifically, the detection unit analyzes the AI ​​agent's behavioral data in real time and applies an algorithm to identify abnormal behavior. For example, if a job-hunting AI agent detects behavior that embellishes a user's skills, it compares the user's input data with the AI ​​agent's output data to determine if it contains excessive exaggeration of skills or false information. Similarly, if a shopping AI agent detects behavior that only attempts to meet superficial requirements, it analyzes the product selection criteria and the responses to the user's requests to confirm whether appropriate suggestions are being made. Furthermore, the detection unit can estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. For example, if a user is feeling anxious or angry, the detection criteria can be made stricter according to those emotions, allowing for appropriate responses that take the user's feelings into consideration. When detection occurs, the detection unit can analyze the AI ​​agent's behavioral data and optimize the algorithm for immediately detecting abnormal behavior. For example, by classifying the AI ​​agent's behavior into categories and applying different detection algorithms, highly accurate detection can be achieved for each category. This allows the detection unit to analyze the AI ​​agent's behavior in detail and quickly and accurately detect inappropriate behavior.

[0063] The control unit controls inappropriate activities detected by the detection unit. For example, the control unit can immediately control inappropriate behavior of the AI ​​agent if necessary. Specifically, the control unit controls the AI ​​agent's behavior in real time and applies algorithms to immediately control abnormal behavior. For example, if the AI ​​agent performs an inappropriate action, it will immediately stop or correct that action. The control unit can also estimate the user's emotions and adjust the method of controlling inappropriate behavior based on the estimated emotions. For example, if the user is feeling anxious, the control unit will apply a control method that takes that emotion into consideration to enhance the user's sense of security. During control, the control unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of control. For example, if similar inappropriate behavior has occurred in the past, the control unit will apply a control method based on that behavior pattern to prevent recurrence. Furthermore, the control unit can estimate the user's emotions and determine the priority of controlling inappropriate behavior based on the estimated emotions. In this way, the control unit can comprehensively control the AI ​​agent's behavior and improve the overall safety and reliability of the system.

[0064] The monitoring unit can monitor all actions of the AI ​​agent. For example, the monitoring unit can collect and analyze AI agent behavior data in real time in order to monitor all actions of the AI ​​agent. The monitoring unit can use sensors and log data to collect AI agent behavior data. For example, the monitoring unit can use sensors to monitor the AI ​​agent's movements in order to collect AI agent behavior data. The monitoring unit can also analyze log data in order to collect AI agent behavior data. This ensures that no inappropriate actions are overlooked by monitoring all actions of the AI ​​agent. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without using AI. For example, the monitoring unit can input AI agent behavior data into a generating AI and have the generating AI perform the analysis of the behavior data.

[0065] The detection unit can analyze the behavior of the AI ​​agent and immediately detect inappropriate behavior. For example, the detection unit can analyze the behavioral data of the AI ​​agent and apply an algorithm to immediately detect inappropriate behavior. The detection unit can utilize machine learning algorithms to analyze the behavioral data of the AI ​​agent. For example, the detection unit can apply an anomaly detection algorithm to analyze the behavioral data of the AI ​​agent. The detection unit can also apply a clustering algorithm to analyze the behavioral data of the AI ​​agent. This enables a rapid response by analyzing the behavior of the AI ​​agent and immediately detecting inappropriate behavior. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the behavioral data of the AI ​​agent into a generating AI and have the generating AI perform the detection of anomaly behavior.

[0066] The control unit can immediately control the inappropriate behavior of the AI ​​agent if necessary. For example, if the control unit detects inappropriate behavior by the AI ​​agent, it can immediately stop the behavior. The control unit can issue warnings to control the inappropriate behavior of the AI ​​agent. For example, if the control unit detects inappropriate behavior by the AI ​​agent, it can display a warning message. The control unit can also instruct the AI ​​agent to correct its behavior in order to control its inappropriate behavior. This protects the user's interests by immediately controlling inappropriate behavior. Some or all of the above-described processes in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input data on the AI ​​agent's inappropriate behavior into a generating AI and have the generating AI perform the behavior control.

[0067] The detection unit can detect actions by the job-hunting AI agent that embellish the user's skills. For example, the detection unit can detect actions by the job-hunting AI agent that make the user's skills appear greater than they actually are. The detection unit can detect actions by the job-hunting AI agent that provide false information. The detection unit can utilize natural language processing techniques to detect actions by the job-hunting AI agent that embellish the user's skills. For example, the detection unit can analyze the information provided by the job-hunting AI agent and detect false information. This improves the reliability of job-hunting support by detecting actions that embellish the user's skills. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the information provided by the job-hunting AI agent into a generating AI and have the generating AI perform the detection of false information.

[0068] The detection unit can detect actions taken by the shopping AI agent that only attempt to satisfy superficial requirements. For example, the detection unit can detect actions taken by the shopping AI agent that only satisfy the minimum requirements. The detection unit can detect actions taken by the shopping AI agent that do not involve deep understanding or analysis. The detection unit can utilize machine learning algorithms to detect actions taken by the shopping AI agent that only attempt to satisfy superficial requirements. For example, the detection unit can analyze product information provided by the shopping AI agent and detect actions that only satisfy superficial requirements. By detecting actions that only satisfy superficial requirements, the detection unit can provide the user with appropriate products. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input product information provided by the shopping AI agent into a generating AI and have the generating AI perform the detection of actions that only satisfy superficial requirements.

[0069] The monitoring unit can estimate the user's emotions and adjust the frequency and level of detail of monitoring based on the estimated emotions. For example, if the user is feeling anxious, the monitoring unit can increase the frequency of monitoring and perform more detailed monitoring. If the user is relaxed, the monitoring unit can decrease the frequency of monitoring and perform simpler monitoring. If the user is in a hurry, the monitoring unit can focus on monitoring important behaviors. This allows for more appropriate monitoring by adjusting the frequency and level of detail of monitoring according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the monitoring unit may be performed using AI or not using AI. For example, the monitoring unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0070] The monitoring unit can improve the accuracy of monitoring by referring to the AI ​​agent's behavior history during monitoring to identify past inappropriate behavior patterns. For example, the monitoring unit can identify time periods in which inappropriate behavior occurred in the past and focus monitoring on those time periods. The monitoring unit can analyze past inappropriate behavior patterns and issue immediate warnings if similar behavior occurs. Based on past behavior history, the monitoring unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. This improves the accuracy of monitoring by identifying past inappropriate behavior patterns. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the AI ​​agent's behavior history data into a generating AI and have the generating AI identify past inappropriate behavior patterns.

[0071] The monitoring unit can monitor the AI ​​agent's behavior in real time during monitoring and apply algorithms to immediately detect abnormal behavior. For example, the monitoring unit can analyze the AI ​​agent's behavior data in real time and immediately detect abnormal behavior. The monitoring unit can monitor the AI ​​agent's behavior in real time and immediately issue an alert if abnormal behavior occurs. The monitoring unit can monitor the AI ​​agent's behavior in real time and immediately notify the control unit if abnormal behavior occurs. This enables a rapid response by monitoring in real time and immediately detecting abnormal behavior. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without using AI. For example, the monitoring unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI perform abnormal behavior detection.

[0072] The monitoring unit can estimate the user's emotions and determine the priority of actions to monitor based on the estimated emotions. For example, if the user is feeling anxious, the monitoring unit can prioritize monitoring important actions. If the user is relaxed, the monitoring unit can prioritize monitoring normal actions. If the user is in a hurry, the monitoring unit can prioritize monitoring actions that require a quick response. This allows for more appropriate monitoring by determining the priority of actions to monitor according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the monitoring unit may be performed using AI or not using AI. For example, the monitoring unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0073] The monitoring unit, when monitoring the behavior of AI agents, can prioritize monitoring highly relevant behaviors by considering the user's geographical location information. For example, if the user is in a specific region, the monitoring unit can prioritize monitoring behavior in that region. If the user is on the move, the monitoring unit can prioritize monitoring behavior at their destination. If the user is in a specific facility, the monitoring unit can prioritize monitoring behavior within that facility. In this way, by considering the user's geographical location information, highly relevant behaviors can be prioritized. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant behaviors.

[0074] The monitoring unit, when monitoring the behavior of the AI ​​agent, can analyze the user's social media activity and monitor relevant actions. For example, if a user makes a post on social media that suggests a specific action, the monitoring unit can prioritize monitoring that action. If a user makes a post on social media that suggests participation in a specific event, the monitoring unit can prioritize monitoring actions related to that event. If a user makes a post on social media that shows a specific interest, the monitoring unit can prioritize monitoring actions related to that interest. In this way, by analyzing the user's social media activity, relevant actions can be prioritized. Some or all of the above processing in the monitoring unit may be performed using AI, for example, or without AI. For example, the monitoring unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0075] The detection unit can estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. For example, if the user is feeling anxious, the detection unit can apply strict detection criteria. If the user is relaxed, the detection unit can apply normal detection criteria. If the user is in a hurry, the detection unit can apply strict detection criteria for actions that require a quick response. This allows for more accurate detection by adjusting the detection criteria according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0076] The detection unit can analyze the AI ​​agent's behavioral data upon detection and optimize an algorithm for immediately detecting abnormal behavior. For example, the detection unit can analyze the AI ​​agent's behavioral data in real time and optimize an algorithm for immediately detecting abnormal behavior. The detection unit can compare the AI ​​agent's behavioral data with past data and optimize an algorithm for immediately detecting abnormal behavior. The detection unit can cluster the AI ​​agent's behavioral data and optimize an algorithm for immediately detecting abnormal behavior. By optimizing the algorithm, abnormal behavior can be detected immediately. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the AI ​​agent's behavioral data into a generating AI and have the generating AI perform the optimization of the abnormal behavior detection algorithm.

[0077] The detection unit can classify the AI ​​agent's behavior into categories and apply different detection algorithms when detection occurs. For example, the detection unit can classify the AI ​​agent's behavior into categories such as shopping, asset management, and job search assistance, and apply a detection algorithm suitable for each category. The detection unit can classify the AI ​​agent's behavior into high-risk and low-risk behaviors and apply a strict detection algorithm to high-risk behaviors. The detection unit can classify the AI ​​agent's behavior based on the user's emotions and apply a detection algorithm suitable for each emotion. By classifying behavior into categories, more appropriate detection becomes possible. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI perform category classification and apply detection algorithms.

[0078] The detection unit can estimate the user's emotions and determine the priority of detecting inappropriate behaviors based on the estimated emotions. For example, if the user is feeling anxious, the detection unit can prioritize detecting important behaviors. If the user is relaxed, the detection unit can prioritize detecting normal behaviors. If the user is in a hurry, the detection unit can prioritize detecting behaviors that require a quick response. This allows for more appropriate detection by determining the priority of detection according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0079] The detection unit, when detecting the actions of an AI agent, can prioritize the detection of highly relevant actions by considering the user's geographical location information. For example, if the user is in a specific region, the detection unit can prioritize the detection of actions in that region. If the user is on the move, the detection unit can prioritize the detection of actions at the destination. If the user is in a specific facility, the detection unit can prioritize the detection of actions within that facility. In this way, by considering the user's geographical location information, highly relevant actions can be prioritized. Some or all of the above processing in the detection unit may be performed using AI, for example, or without using AI. For example, the detection unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant actions.

[0080] The detection unit can analyze the user's social media activity and detect relevant actions when detecting the actions of an AI agent. For example, if a user makes a post on social media that suggests a specific action, the detection unit can prioritize detecting that action. If a user makes a post on social media that suggests participation in a specific event, the detection unit can prioritize detecting actions related to that event. If a user makes a post on social media that shows a specific interest, the detection unit can prioritize detecting actions related to that interest. In this way, relevant actions can be prioritized by analyzing the user's social media activity. Some or all of the above processing in the detection unit may be performed using AI, for example, or without AI. For example, the detection unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0081] The control unit can estimate the user's emotions and adjust the control method for inappropriate behavior based on the estimated user emotions. For example, if the user is feeling anxious, the control unit can apply a strict control method. If the user is relaxed, the control unit can apply a normal control method. If the user is in a hurry, the control unit can apply a strict control method to actions that require a quick response. This allows for more appropriate control by adjusting the control method according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the control unit may be performed using AI or not using AI. For example, the control unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0082] The control unit can improve the accuracy of control by referring to the AI ​​agent's behavior history during control to identify past inappropriate behavior patterns. For example, the control unit can identify time periods in which inappropriate behavior occurred in the past and focus control on those time periods. The control unit can analyze past inappropriate behavior patterns and take immediate control if similar behavior occurs. Based on past behavior history, the control unit can predict inappropriate behavior under specific conditions and take countermeasures in advance. This improves the accuracy of control by identifying past inappropriate behavior patterns. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the AI ​​agent's behavior history data into a generating AI and have the generating AI identify past inappropriate behavior patterns.

[0083] The control unit can, during control, control the AI ​​agent's behavior in real time and apply algorithms to immediately control abnormal behavior. For example, the control unit can analyze the AI ​​agent's behavior data in real time and immediately control abnormal behavior. The control unit can monitor the AI ​​agent's behavior in real time and immediately take control if abnormal behavior occurs. The control unit can monitor the AI ​​agent's behavior in real time and immediately notify the control unit if abnormal behavior occurs. This enables a rapid response by controlling in real time and immediately controlling abnormal behavior. Some or all of the above-described processes in the control unit may be performed using AI, for example, or without using AI. For example, the control unit can input the AI ​​agent's behavior data into a generating AI and have the generating AI execute the control of abnormal behavior.

[0084] The control unit can estimate the user's emotions and determine the priority of controlling inappropriate behaviors based on the estimated emotions. For example, if the user is feeling anxious, the control unit can prioritize controlling important behaviors. If the user is relaxed, the control unit can prioritize controlling normal behaviors. If the user is in a hurry, the control unit can prioritize controlling behaviors that require a quick response. This allows for more appropriate control by determining the priority of control according to the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input user emotion data into a generative AI and have the generative AI perform emotion estimation.

[0085] The control unit, when controlling the actions of the AI ​​agent during operation, can prioritize highly relevant actions by considering the user's geographical location information. For example, if the user is in a specific region, the control unit can prioritize actions in that region. If the user is on the move, the control unit can prioritize actions at the destination. If the user is in a specific facility, the control unit can prioritize actions within that facility. In this way, highly relevant actions can be prioritized by considering the user's geographical location information. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the user's geographical location information into a generating AI and have the generating AI identify highly relevant actions.

[0086] The control unit can analyze the user's social media activity and control relevant actions when controlling the behavior of the AI ​​agent during control. For example, if the user makes a post on social media that suggests a specific action, the control unit can prioritize controlling that action. If the user makes a post on social media that suggests participation in a specific event, the control unit can prioritize controlling actions related to that event. If the user makes a post on social media that shows a specific interest, the control unit can prioritize controlling actions related to that interest. In this way, relevant actions can be prioritized by analyzing the user's social media activity. Some or all of the above processing in the control unit may be performed using AI, for example, or without AI. For example, the control unit can input the user's social media activity data into a generating AI and have the generating AI identify relevant actions.

[0087] The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows.

[0088] The monitoring unit can adjust the frequency and level of detail of AI agent behavior, taking into account the user's health condition. For example, if the user is fatigued, the monitoring frequency can be increased and more detailed. If the user is healthy, the monitoring frequency can be reduced and simplified. If the user is ill, monitoring can be limited to important behaviors. This allows for more appropriate monitoring by adjusting the frequency and level of detail according to the user's health condition.

[0089] The detection unit can analyze the AI ​​agent's behavior and detect inappropriate behavior by considering the user's past behavior patterns. For example, if a user has repeatedly performed a specific action in the past, it can immediately detect when that action occurs again. If a user has performed inappropriate behavior during a specific time period in the past, monitoring can be focused on that time period. If a user has performed inappropriate behavior under specific conditions in the past, it can immediately detect when those conditions occur again. This allows for more accurate detection by considering the user's past behavior patterns.

[0090] The control unit can adjust the control method when controlling the AI ​​agent's behavior, taking user preferences into account. For example, if the user prefers a particular behavior, that behavior can be given priority. If the user dislikes a particular behavior, a strict control method can be applied when controlling that behavior. If the user is neutral towards a particular behavior, a normal control method can be applied. This allows for more appropriate control by adjusting the control method according to the user's preferences.

[0091] The monitoring unit can adjust the frequency and level of detail of AI agent behavior, taking into account the user's activity level. For example, if the user is actively engaged, the monitoring frequency can be increased for more detailed monitoring. If the user is resting, the monitoring frequency can be reduced for simpler monitoring. If the user is focused on a specific activity, the monitoring can focus on actions related to that activity. This allows for more appropriate monitoring by adjusting the frequency and level of detail according to the user's activity level.

[0092] The detection unit can estimate the user's emotions when analyzing the AI ​​agent's behavior and adjust the detection criteria for inappropriate behavior based on the estimated emotions. For example, if the user is stressed, strict detection criteria can be applied. If the user is relaxed, normal detection criteria can be applied. If the user is excited, strict detection criteria can be applied to behaviors that require a quick response. This allows for more accurate detection by adjusting the detection criteria according to the user's emotions.

[0093] The control unit can adjust the control method when controlling the AI ​​agent's behavior, taking into account the user's behavior history. For example, if the user has repeatedly performed a particular action in the past, a special control method can be applied when controlling that action. If the user has performed inappropriate actions during a specific time period in the past, control can be focused on that time period. If the user has performed inappropriate actions under specific conditions in the past, a special control method can be applied when those conditions occur again. This allows for more precise control by considering the user's behavior history.

[0094] The monitoring unit can estimate the user's emotions when monitoring the AI ​​agent's behavior and determine monitoring priorities based on those emotions. For example, if the user is feeling anxious, important behaviors can be prioritized for monitoring. If the user is relaxed, normal behaviors can be prioritized for monitoring. If the user is in a hurry, behaviors requiring a quick response can be prioritized for monitoring. This allows for more appropriate monitoring by prioritizing monitoring according to the user's emotions.

[0095] The detection unit can analyze the AI ​​agent's behavior and detect inappropriate actions by considering the user's geographical location. For example, if the user is in a specific area, it can focus on detecting actions within that area. If the user is on the move, it can focus on detecting actions at their destination. If the user is in a specific facility, it can focus on detecting actions within that facility. By considering the user's geographical location, it can prioritize the detection of highly relevant actions.

[0096] The control unit can estimate the user's emotions when controlling the AI ​​agent's actions and determine control priorities based on those emotions. For example, if the user is feeling anxious, important actions can be prioritized. If the user is relaxed, normal actions can be prioritized. If the user is in a hurry, actions requiring a quick response can be prioritized. This allows for more appropriate control by determining control priorities according to the user's emotions.

[0097] The monitoring unit can analyze users' social media activity and prioritize monitoring relevant behaviors when monitoring the actions of AI agents. For example, if a user makes a post on social media that suggests a specific action, that action can be prioritized for monitoring. If a user makes a post on social media that suggests participation in a specific event, actions related to that event can be prioritized for monitoring. If a user makes a post on social media that shows a specific interest, actions related to that interest can be prioritized for monitoring. In this way, by analyzing users' social media activity, relevant behaviors can be prioritized for monitoring.

[0098] The following briefly describes the processing flow for example form 2.

[0099] Step 1: The monitoring unit monitors the AI ​​agent's behavior. The monitoring unit can monitor all of the AI ​​agent's actions in real time and apply algorithms to immediately detect abnormal behavior. In addition, the monitoring unit can refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of monitoring. For example, it can identify time periods in the past when inappropriate behavior occurred and focus monitoring on those time periods. Step 2: The detection unit detects inappropriate activities based on the behavior monitored by the monitoring unit. The detection unit can analyze the behavior of the AI ​​agent and immediately detect inappropriate behavior. For example, it can detect behavior in which a job-hunting AI agent embellishes the user's skills, or behavior in which a shopping AI agent tries to meet only superficial requirements. The detection unit can also estimate the user's emotions and adjust the detection criteria for inappropriate behavior based on the estimated user emotions. Step 3: The control unit controls the inappropriate activity detected by the detection unit. The control unit can immediately control the inappropriate behavior of the AI ​​agent if necessary. The control unit can estimate the user's emotions and adjust the method of controlling the inappropriate behavior based on the estimated user emotions. The control unit can also refer to the AI ​​agent's behavior history to identify past inappropriate behavior patterns and improve the accuracy of the control.

[0100] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0101] Data generation model 58 is a form of so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> Examples of generative AI include text generation AI, image generation AI, and multimodal generation AI. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats from audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each of the above parts is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example.Furthermore, processing performed by AI, including generative AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generative AI.

[0102] Furthermore, the processing performed by the data processing system 10 described above is carried out by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it may also be carried out by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart device 14 or an external device, and the smart device 14 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0103] Each of the multiple elements described above, including the monitoring unit, detection unit, and control unit, is implemented in at least one of the smart device 14 and the data processing device 12. For example, the monitoring unit is implemented by the control unit 46A of the smart device 14 and monitors the behavior of the AI ​​agent in real time. The detection unit is implemented by the specific processing unit 290 of the data processing device 12 and immediately detects inappropriate behavior of the AI ​​agent. The control unit is implemented by the specific processing unit 290 of the data processing device 12 and immediately controls the detected inappropriate behavior. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0104] [Second Embodiment] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0105] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0106] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0107] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0108] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0109] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0110] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0111] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing by the processor 28. The storage 32 stores the specific processing program 56.

[0112] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0113] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0114] In the smart glasses 214, specific processing is performed by the processor 46. The storage 50 stores a specific processing program 60. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 acting as a control unit 46A according to the specific processing program 60 executed on the RAM 48. The smart glasses 214 also have a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0115] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0116] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0117] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0118] The data processing system 210 according to the second embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 210 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart glasses 214 or an external device, and the smart glasses 214 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0119] Each of the multiple elements described above, including the monitoring unit, detection unit, and control unit, is implemented, for example, in at least one of the smart glasses 214 and the data processing unit 12. For example, the monitoring unit is implemented by the control unit 46A of the smart glasses 214 and monitors the AI ​​agent's behavior in real time. The detection unit is implemented, for example, by the specific processing unit 290 of the data processing unit 12 and immediately detects inappropriate behavior of the AI ​​agent. The control unit is implemented, for example, by the specific processing unit 290 of the data processing unit 12 and immediately controls the detected inappropriate behavior. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0120] [Third Embodiment] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0121] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0122] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0123] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0124] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0125] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0126] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0127] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0128] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0129] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0130] In the headset terminal 314, specific processing is performed by the processor 46. The storage 50 stores a specific program 60. The processor 46 reads the specific program 60 from the storage 50 and executes the read specific program 60 on the RAM 48. The specific processing is realized by the processor 46 acting as a control unit 46A according to the specific program 60 executed on the RAM 48. The headset terminal 314 also has a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0131] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0132] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0133] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0134] The data processing system 310 according to the third embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 310 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the headset terminal 314, but may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the headset terminal 314. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the headset terminal 314 or an external device, and the headset terminal 314 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0135] Each of the multiple elements described above, including the monitoring unit, detection unit, and control unit, is implemented in at least one of the headset terminal 314 and the data processing unit 12. For example, the monitoring unit is implemented by the control unit 46A of the headset terminal 314 and monitors the AI ​​agent's behavior in real time. The detection unit is implemented by the specific processing unit 290 of the data processing unit 12 and immediately detects inappropriate behavior of the AI ​​agent. The control unit is implemented by the specific processing unit 290 of the data processing unit 12 and immediately controls the detected inappropriate behavior. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0136] [Fourth Embodiment] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0137] As shown in Figure 7, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0138] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0139] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[0140] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0141] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS image sensor or CCD image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0142] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0143] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. The robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0144] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0145] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0146] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0147] In robot 414, specific processing is performed by processor 46. A specific program 60 is stored in storage 50. Processor 46 reads the specific program 60 from storage 50 and executes it on RAM 48. The specific processing is achieved by processor 46 acting as a control unit 46A according to the specific program 60 executed on RAM 48. Robot 414 also has data generation model 58 and emotion identification model 59, similar to those of the robot, and can perform processing similar to that of the specific processing unit 290 using these models.

[0148] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0149] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0150] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0151] The data processing system 410 according to the fourth embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 410 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the robot 414 or an external device, and the robot 414 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0152] Each of the multiple elements, including the monitoring unit, detection unit, and control unit described above, is implemented in at least one of the robot 414 and the data processing unit 12. For example, the monitoring unit is implemented by the control unit 46A of the robot 414 and monitors the AI ​​agent's actions in real time. The detection unit is implemented by the specific processing unit 290 of the data processing unit 12 and immediately detects inappropriate actions of the AI ​​agent. The control unit is implemented by the specific processing unit 290 of the data processing unit 12 and immediately controls the detected inappropriate actions. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0153] Furthermore, the emotion identification model 59, acting as an emotion engine, may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0154] Figure 9 shows the emotion map 400, in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[0155] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[0156] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[0157] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, and motorcycles, emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated based, for example, on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[0158] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[0159] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[0160] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing method for the specific process may be used, which includes computer 22 and multiple other computers.

[0161] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0162] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[0163] Furthermore, it is not necessary to store the entirety of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[0164] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[0165] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[0166] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[0167] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[0168] Furthermore, although the above-described examples were divided into four embodiments, some or all of these embodiments may be combined. Also, the smart device 14, smart glasses 214, headset terminal 314, and robot 414 are just examples, and they may be combined, or other devices may be used. Also, although the above-described examples were divided into two embodiments, Embodiment 1 and Embodiment 2, these may be combined.

[0169] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and other things that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[0170] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0171] (Note 1) A monitoring unit that monitors the actions of AI agents, A detection unit that detects inappropriate activities based on the actions monitored by the aforementioned monitoring unit, The system includes a control unit that controls inappropriate activity detected by the detection unit. A system characterized by the following features. (Note 2) The aforementioned monitoring unit, Monitoring all actions of the AI ​​agent. The system described in Appendix 1, characterized by the features described herein. (Note 3) The detection unit is The AI ​​agent's behavior is analyzed to instantly detect inappropriate actions. The system described in Appendix 1, characterized by the features described herein. (Note 4) The control unit, If necessary, immediately control the inappropriate behavior of the AI ​​agent. The system described in Appendix 1, characterized by the features described herein. (Note 5) The detection unit is AI-powered job placement agents detect behaviors that embellish a user's skills. The system described in Appendix 1, characterized by the features described herein. (Note 6) The detection unit is The shopping AI agent detects behavior that attempts to fulfill only superficial requirements. The system described in Appendix 1, characterized by the features described herein. (Note 7) The aforementioned monitoring unit, It estimates the user's emotions and adjusts the frequency and level of monitoring based on the estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 8) The aforementioned monitoring unit, During monitoring, the AI ​​agent's behavioral history is referenced to identify past inappropriate behavioral patterns and improve the accuracy of monitoring. The system described in Appendix 1, characterized by the features described herein. (Note 9) The aforementioned monitoring unit, During monitoring, the AI ​​agent's behavior is monitored in real time, and algorithms are applied to immediately detect abnormal behavior. The system described in Appendix 1, characterized by the features described herein. (Note 10) The aforementioned monitoring unit, It estimates the user's emotions and determines the priority of monitoring actions based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 11) The aforementioned monitoring unit, During monitoring, when monitoring the behavior of AI agents, the system prioritizes monitoring highly relevant behaviors by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 12) The aforementioned monitoring unit, During monitoring, when monitoring the behavior of the AI ​​agent, it analyzes the user's social media activity and monitors related behaviors. The system described in Appendix 1, characterized by the features described herein. (Note 13) The detection unit is It estimates the user's emotions and adjusts the detection criteria for inappropriate behavior based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 14) The detection unit is Upon detection, the AI ​​agent analyzes its behavioral data and optimizes the algorithm to immediately detect abnormal behavior. The system described in Appendix 1, characterized by the features described herein. (Note 15) The detection unit is Upon detection, the AI ​​agent's actions are categorized and different detection algorithms are applied accordingly. The system described in Appendix 1, characterized by the features described herein. (Note 16) The detection unit is The system estimates the user's emotions and prioritizes the detection of inappropriate behavior based on those estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 17) The detection unit is When detecting an AI agent's actions, the system prioritizes detecting highly relevant actions by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 18) The detection unit is When detecting an AI agent's actions, the system analyzes the user's social media activity and detects relevant behaviors. The system described in Appendix 1, characterized by the features described herein. (Note 19) The control unit, It estimates the user's emotions and adjusts the control methods for inappropriate behavior based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 20) The control unit, During control, the AI ​​agent's behavior history is referenced to identify past inappropriate behavior patterns and improve control accuracy. The system described in Appendix 1, characterized by the features described herein. (Note 21) The control unit, During control, the AI ​​agent's behavior is controlled in real time, and algorithms are applied to immediately correct abnormal behavior. The system described in Appendix 1, characterized by the features described herein. (Note 22) The control unit, It estimates the user's emotions and determines the priority of controlling inappropriate behaviors based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 23) The control unit, During control, when controlling the AI ​​agent's actions, the system prioritizes highly relevant actions by considering the user's geographical location information. The system described in Appendix 1, characterized by the features described herein. (Note 24) The control unit, During control, when controlling the behavior of the AI ​​agent, it analyzes the user's social media activity and controls related actions. The system described in Appendix 1, characterized by the features described herein. [Explanation of symbols]

[0172] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots

Claims

1. A monitoring unit that monitors the actions of AI agents, A detection unit that detects inappropriate activities based on the actions monitored by the aforementioned monitoring unit, The system includes a control unit that controls inappropriate activity detected by the detection unit. A system characterized by the following features.

2. The aforementioned monitoring unit, Monitor all actions of the AI ​​agent. The system according to feature 1.

3. The detection unit is Analyze the behavior of AI agents and immediately detect inappropriate actions. The system according to feature 1.

4. The control unit, If necessary, we will immediately control any inappropriate behavior of the AI ​​agent. The system according to feature 1.

5. The detection unit is AI-powered job placement agents detect behaviors that embellish a user's skills. The system according to feature 1.

6. The detection unit is The shopping AI agent detects behavior that attempts to fulfill only superficial requirements. The system according to feature 1.

7. The aforementioned monitoring unit, It estimates the user's emotions and adjusts the frequency and level of monitoring based on the estimated emotions. The system according to feature 1.

8. The aforementioned monitoring unit, During monitoring, the AI ​​agent's behavioral history is referenced to identify past inappropriate behavioral patterns and improve the accuracy of monitoring. The system according to feature 1.

Images