Information processing device, information processing method, and information processing program

Abstract

Supporting user use of generative AI. [Solution] The information processing device 100 comprises a learning means for having an AI agent learn how the user will use the generated AI, and a recording means for recording data for using the AI ​​agent learned by the learning means.The information processing device 100 also comprises a prompt input means for inputting a prompt entered by the user to the trained AI agent, a converted prompt acquisition means for acquiring the prompt converted by the AI ​​agent, and a converted prompt input means for inputting the converted prompt acquired by the converted prompt acquisition means to the generated AI.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] The following system is known: In this system, a question input by a user from a user terminal is received, the question is sent to a generation AI, an answer to the question is received from the generation AI, and an explanation for the user is generated based on the answer (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2025-074556 Summary of the Invention [Problem to be solved by the invention]

[0004] In conventional systems, where a user inputs a question (prompt) to a generating AI and the generating AI outputs an answer, if an AI agent that has learned the user's usage of the generating AI is prepared, the convenience of the user's use of the generating AI can be improved by using that AI agent. However, no technology for this purpose has been considered up to now. [Means for solving the problem]

[0005] The information processing device according to the present invention includes a prompt input means for inputting a prompt input by a user to a trained AI agent that has been made to learn how the generated AI will be used by the user, a converted prompt acquisition means for acquiring a prompt converted by the AI ​​agent, and a converted prompt input means for inputting the converted prompt acquired by the converted prompt acquisition means to the generated AI. an inference result acquisition means for acquiring an inference result output from the generation AI in response to input of the converted prompt; an inference result input means for inputting the inference result acquired by the inference result acquisition means to the AI ​​agent; a converted inference result acquisition means for acquiring the inference result converted by the AI ​​agent; and a presentation means for presenting the converted inference result acquired by the converted inference result acquisition means to the user. The present invention is characterized by comprising: The information processing device according to the present invention comprises an inference result acquisition means for acquiring the inference result output from the generation AI, an inference result input means for inputting the inference result acquired by the inference result acquisition means to a trained AI agent that has been made to learn how the user will use the generation AI, a converted inference result acquisition means for acquiring the inference result converted by the AI ​​agent, and a presentation means for presenting the converted inference result acquired by the converted inference result acquisition means to the user, wherein the presentation means explains the intention behind the generation by the generation AI when presenting the inference result to the user. The information processing method according to the present invention includes the steps of: a prompt input means inputting a prompt input by a user to a trained AI agent that has been made to learn how the generated AI will be used by the user; a converted prompt acquisition means acquiring a prompt converted by the AI ​​agent; and a converted prompt input means inputting the converted prompt acquired by the converted prompt acquisition means to the generated AI. a step in which an inference result acquisition means acquires an inference result output from the generation AI in response to input of the converted prompt; a step in which an inference result input means inputs the inference result acquired by the inference result acquisition means to the AI ​​agent; a step in which a converted inference result acquisition means acquires the inference result converted by the AI ​​agent; and a step in which a presentation means presents the converted inference result acquired by the converted inference result acquisition means to a user. The method is a computer-implemented method comprising: The information processing method according to the present invention is an information processing method carried out on a computer, comprising the steps of: an inference result acquisition means acquiring an inference result output from an AI; an inference result input means inputting the inference result acquired by the inference result acquisition means to a trained AI agent that has been made to learn how the user will use the generating AI; a converted inference result acquisition means acquiring the inference result converted by the AI ​​agent; and a presentation means presenting the converted inference result acquired by the converted inference result acquisition means to the user, wherein the presentation means explains the intention behind the generation by the generating AI when presenting the inference result to the user. An information processing program according to the present invention is a program for causing a computer to execute the above-described information processing method. [Effects of the Invention]

[0006] According to the present invention, it is possible to create a trained AI agent that has learned how a user will use the generated AI, and it is also possible to provide a system that utilizes the trained AI agent, thereby improving the convenience for users when using the generated AI. [Brief explanation of the drawings]

[0007] [Figure 1] 1 is a block diagram showing a configuration of an embodiment of an information processing device 100. FIG. [Figure 2] FIG. 1 is a flowchart showing the process flow in the learning stage of an AI agent. [Figure 3] FIG. 10 is a flowchart showing the flow of input utilization processing of an AI agent. [Figure 4] FIG. 10 is a flowchart showing the flow of the output utilization process of the AI ​​agent. [Figure 5] FIG. 1 is a diagram showing a schematic flow of processing in the learning stage of an AI agent. [Figure 6] FIG. 1 is a diagram showing a schematic diagram of the processing flow at the stage of using an AI agent. DETAILED DESCRIPTION OF THE INVENTION

[0008] 1 is a block diagram showing the configuration of an embodiment of an information processing device 100 constituting an information processing system according to the present embodiment. The information processing device 100 is, for example, a personal computer or a server, and includes a communication module 101, a control device 102, and a recording device 103.

[0009] The communication module 101 includes a module for connecting the information processing device 100 to a communication line such as a LAN or the Internet, wirelessly or by wire. In this embodiment, the information processing device 100 can communicate with other devices by connecting to the communication line via the communication module 101.

[0010] The control device 102 is configured with a CPU, memory, and other peripheral circuits, and controls the entire information processing device 100. The memory that configures the control device 102 is, for example, a volatile memory such as SDRAM. This memory is used as a work memory for the CPU to expand programs when the programs are executed, and as a buffer memory for temporarily recording data. For example, data received via the communication module 101 is temporarily recorded in the buffer memory.

[0011] The recording device 103 is a recording device for recording various data stored in the information processing device 100, program data to be executed by the control device 102, etc., and may be, for example, a hard disk drive (HDD) or a solid state drive (SSD). The program data recorded in the recording device 103 is provided by being recorded on a recording medium such as a CD-ROM or DVD-ROM, or provided via a network, and the user can install the acquired program data in the recording device 103, thereby enabling the control device 102 to execute the program.

[0012] The information processing device 100 of this embodiment provides a mechanism for creating an AI agent that supports the use of generative AI in an organization, such as a company, department, or team, and a mechanism for using this AI agent to support the use of generative AI in the organization. To this end, the information processing device 100 of this embodiment provides an AI agent that supports the use of generative AI in the organization by learning in collaboration with the use of generative AI by individuals (users) belonging to the organization, thereby sharing individual knowledge and rules with the entire organization.

[0013] In this embodiment, when a user uses the generated AI, the AI ​​agent learns by collaborating with the user to train or cowork with the user. For this purpose, the information processing device 100 in this embodiment executes a process for creating an AI agent that supports the user's use of the generated AI, and a process for using this AI agent to support the user's use of the generated AI.

[0014] Here, an AI agent is a system that acts as a human agent, has the ability to learn and make decisions on its own, and integrates various AI technologies to solve complex problems. For example, when a user gives an AI agent a question (prompt), the AI ​​agent searches for an answer to the question, generates an answer, and outputs the answer. Also, when a user gives an AI agent a task (prompt), the AI ​​agent searches for a solution to the task, generates an answer, and outputs the solution.

[0015] In this embodiment, an LLM (Large Scale Language Model) is assumed as an example of an AI agent, but the AI ​​agent according to the present invention is not limited to an LLM (Large Scale Language Model).

[0016] The AI ​​agent in this embodiment learns how users belonging to an organization use the generated AI, and supports the use of the generated AI in the organization. To this end, the AI ​​agent in this embodiment learns the content of the prompts that users input to the generated AI and the actions that users take in response to the inference results output by the generated AI. The trained AI agent is then shared throughout the organization. This supports the input of prompts to the generated AI by users belonging to the organization. The trained AI agent also supports the output of inference results from the generated AI. Below, we will explain the learning stage processing in this embodiment, in which the AI ​​agent is trained, and the usage stage processing in which the AI ​​agent is used.

[0017] First, we will explain the learning stage processing executed by the control device 102. In the AI ​​agent's learning stage, as described above, the control device 102 causes the AI ​​agent to learn how the generated AI is used by the user. In this embodiment, the AI ​​agent learns the content of the prompts the user inputs to the generated AI and the user's actions in response to the inference results output by the generated AI as the content of how the generated AI is used by the user.

[0018] The control device 102 inputs the contents of the prompts input by the user when using the generation AI to the AI ​​agent as learning data, and causes the AI ​​agent to learn the contents of the prompts input by the user. This allows the AI ​​agent to learn what expressions, such as terminology, wording, and phrasing, the user uses to input prompts when using the generation AI on a regular basis. In this embodiment, when using the generation AI, the user inputs prompts based on organizational rules, tacit knowledge, or personal knowledge and rules based on judgment criteria. This allows the AI ​​agent to learn the contents of prompts that match organizational rules, tacit knowledge, or judgment criteria.

[0019] The control device 102 also learns the user's behavior in response to the inference results output from the generation AI, which indicates the judgments and actions the user took in response to the output from the generation AI. This is intended to verbalize judgment criteria that are difficult to verbalize, such as the user's experiences and feelings.

[0020] For example, if a user asks a question to the generating AI and receives an answer, the AI ​​agent will learn the question (prompt) that the user entered into the generating AI. Also, if a user asks a question to the generating AI, the generating AI outputs an answer, and the user corrects the answer, the AI ​​agent will learn the corrections made by the user to the output from the generating AI.

[0021] For example, if a user determines that the inference result (answer) output from the generation AI differs from the terminology or language commonly used within the organization, the user corrects the output from the generation AI to the commonly used terminology or language. Also, for example, if a user determines that the inference result (answer) output from the generation AI uses expressions that differ from the organization's conventions, the user corrects the output from the generation AI to expressions that conform to the organization's conventions. In these cases, the AI ​​agent can learn the answers that the user corrected in response to the output from the generation AI. This allows the AI ​​agent to learn what expressions, such as terminology, language, and phrasing, the user desires in answers that conform to the rules, tacit knowledge, or judgment criteria of the organization to which the user belongs.

[0022] Furthermore, for example, if a user determines that the inference result (answer) output by the generation AI is not what is required within the organization, the user may re-enter a question (prompt) to the generation AI with a different expression. In this case, the AI ​​agent can learn the content of the user's re-question in response to the output from the generation AI. This allows the AI ​​agent to learn what kind of answer the user was looking for when entering the prompt, and what their intention was when entering the prompt, so that it conforms to the rules, tacit knowledge, or judgment criteria of the organization to which the user belongs.

[0023] When the learning of the AI ​​agent is complete, the control device 102 records data for using the learned AI agent in a preset storage location. In this embodiment, the data for using the AI ​​agent created by the above-mentioned process may be recorded in the recording device 103, or in another device connectable via the communication module 101. Furthermore, if the learning of the above-mentioned AI agent is performed on multiple individuals (users) belonging to an organization, it is possible to more accurately reflect the organization's rules, tacit knowledge, judgment criteria, etc.

[0024] Next, we will explain the utilization stage processing executed by the control device 102. In this embodiment, there are two methods for utilizing the AI ​​agent: input utilization, in which the AI ​​agent supports the input of prompts when the user inputs them to the generation AI, and output utilization, in which the AI ​​agent supports the output from the generation AI when the generation AI outputs an inference result.

[0025] In input utilization, when a user inputs a prompt, the control device 102 inputs the prompt input by the user to the AI ​​agent. As described above, the AI ​​agent has learned the input content of the prompt that conforms to the rules, tacit knowledge, or judgment criteria of the organization to which the user belongs, and has also learned what kind of answer the user was looking for when inputting the prompt, or what intention the user had when inputting the prompt, so that it conforms to the rules, tacit knowledge, or judgment criteria of the organization to which the user belongs. Therefore, the control device 102 converts the prompt input by the user into content that accurately expresses the organization's intentions, so that the generation AI can obtain the inference result desired by the organization, and generates and outputs the prompt. The control device 102 inputs the prompt output from the AI ​​agent to the generation AI. This allows the generation AI to input a prompt that accurately expresses the organization's intentions, for example, a prompt that conforms to the organization's rules, tacit knowledge, or judgment criteria, thereby supporting the generation AI in outputting the inference result desired by the organization.

[0026] In output utilization, when an inference result is output from the generation AI, the control device 102 inputs the inference result output from the generation AI to the AI ​​agent. As described above, the AI ​​agent has learned what kind of expressions, such as terminology, wording, and phrasing, the user desires in an answer so as to conform to the rules, tacit knowledge, or judgment criteria of the organization to which the user belongs. Therefore, the AI ​​agent converts the inference result output from the generation AI into an expression that conforms to the organization's intentions, such as an expression that conforms to the organization's rules, tacit knowledge, or judgment criteria, and outputs it. The control device 102 presents the inference result output from the AI ​​agent to the user. This allows the output content from the generation AI to be converted into an expression that conforms to the organization and presented to the user. Furthermore, when presenting the inference result to the user, an explanation of the intention behind the generation by the generation AI may be provided. Note that the method of presenting the inference result to the user is not particularly limited. For example, the inference result may be displayed on a monitor (not shown) or transmitted to a user terminal connected via a communication line.

[0027] 2 is a flowchart showing the flow of processing in the learning stage of an AI agent executed by information processing device 100 in this embodiment. The processing shown in FIG. 2 is executed by control device 102 as a program that is activated when it is time to train the AI ​​agent. Note that there are no particular limitations on the timing at which the AI ​​agent is trained. For example, in order to train the AI ​​agent in cooperation with the user's use of the generated AI, control device 102 may detect the timing at which the user starts using the generated AI during the AI ​​agent's learning stage as the timing to train the AI ​​agent.

[0028] In step S10, the control device 102 inputs the contents of the prompt entered by the user when using the generated AI as learning data to the AI ​​agent, as described above, and causes the AI ​​agent to learn the contents of the prompt entered by the user. Then, the process proceeds to step S20.

[0029] In step S20, the control device 102 inputs the user's judgment and the action taken in response to the output from the generation AI as learning data to the AI ​​agent, as described above, and causes the AI ​​agent to learn the user's actions in response to the inference result. Then, the process proceeds to step S30.

[0030] In step S30, the control device 102 determines whether the learning of the AI ​​agent has ended. The control device 102 may determine that the learning of the AI ​​agent has ended, for example, when the user has finished using the generated AI or when the user has instructed the AI ​​agent to end learning. If the determination in step S30 is negative, the process returns to step S10. On the other hand, if the determination in step S30 is positive, the process proceeds to step S40.

[0031] In step S40, the control device 102 records the data for using the trained AI agent in a recording destination, as described above, and then ends the process.

[0032] 3 is a flowchart showing the flow of the input utilization process of the AI ​​agent executed by the information processing device 100 in this embodiment. The process shown in FIG. 3 is executed by the control device 102 as a program that is activated when the user inputs a prompt to the generated AI during the AI ​​agent usage stage.

[0033] In step S110, the control device 102 inputs the prompt entered by the user to the AI ​​agent, and then proceeds to step S120.

[0034] In step S120, the control device 102 determines whether or not a converted prompt has been output from the AI ​​agent. If the determination in step S120 is affirmative, the process proceeds to step S130.

[0035] In step S130, the control device 102 acquires the prompt output from the AI ​​agent and inputs it to the generated AI, after which the process ends.

[0036] 4 is a flowchart showing the flow of the output utilization process of the AI ​​agent executed by the information processing device 100 in this embodiment. The process shown in FIG. 4 is executed by the control device 102 as a program that is activated when an inference result is output from the generation AI during the AI ​​agent utilization stage.

[0037] In step S210, the control device 102 inputs the inference result output from the generating AI to the AI ​​agent, and then proceeds to step S220.

[0038] In step S220, the control device 102 determines whether or not the converted inference result has been output from the AI ​​agent. If the determination in step S220 is affirmative, the process proceeds to step S230.

[0039] In step S230, the control device 102 acquires the inference result output from the AI ​​agent and presents it to the user, after which the process ends.

[0040] FIG. 5 is a diagram showing a typical flow of processing in the learning stage of an AI agent according to this embodiment.

[0041] In step S310, the generated AI learns the question (prompt) input by the user, and then proceeds to step S320.

[0042] In step S320, the AI ​​agent learns the content of the action, i.e., the behavior content, in response to the inference result output by the user-generated AI, and then proceeds to step S330.

[0043] In step S330, if the AI ​​agent has completed learning, the process ends, and if learning is to continue, the process returns to step S310.

[0044] FIG. 6 is a diagram showing a typical flow of processing at the stage of using an AI agent in this embodiment.

[0045] In step S410, the question (prompt) input by the user to the generation AI is converted, and then the process proceeds to step S420.

[0046] In step S420, the AI ​​agent converts the output content from the generating AI, that is, the content of the inference result output from the generating AI, and then proceeds to step S430.

[0047] In step S430, if the user has finished using the generated AI, the process ends, and if the user intends to continue using the generated AI, the process returns to step S410.

[0048] According to the embodiment described above, the following advantageous effects can be obtained. (1) The control device 102 allows the AI ​​agent to learn how the user will use the generated AI, and records data for using the trained AI agent. This makes it possible to create a trained AI agent that has learned how the user will use the generated AI.

[0049] (2) The control device 102 makes the AI ​​agent learn the contents of the prompts the user inputs to the generating AI and the actions of the user in response to the inference results output from the generating AI as the user's usage of the generating AI. This allows the AI ​​agent to learn the user's usage of the generating AI in cooperation with the user who uses the generating AI.

[0050] (3) The control device 102 makes the AI ​​agent learn the content indicating what decisions the user made and what actions they took in response to the output from the generating AI as the content of the user's actions in response to the inference results output from the generating AI. This allows the AI ​​agent to learn what decisions the user made and what actions they took in response to the output from the generating AI in cooperation with the user who uses the generating AI.

[0051] (4) When a user corrects an inference result output from the generation AI, the control device 102 causes the AI ​​agent to learn the corrections made by the user as the user's action in response to the inference result output from the generation AI. This allows the AI ​​agent to learn what kind of expressions, such as terminology, wording, and phrasing, the user desires in the response.

[0052] (5) When the user asks a follow-up question about the inference result output from the generation AI, the control device 102 causes the AI ​​agent to learn the content of the follow-up question as the user's behavior in response to the inference result output from the generation AI. This allows the AI ​​agent to learn what expressions, such as terms, wording, and phrasing, the user tends to use when entering prompts.

[0053] (6) When using generative AI, users input prompts in accordance with organizational rules, tacit knowledge, or personal knowledge and rules based on judgment criteria. By sharing the AI ​​agent that has learned the prompts entered by the user within the organization, the use of generative AI within an organization, such as a company, department, or team, can be supported.

[0054] (7) The control device 102 acquires the inference result output from the generating AI in response to the input of the converted prompt, inputs the acquired inference result to the AI ​​agent, acquires the inference result converted by the AI ​​agent, and presents the acquired converted inference result to the user. This allows the inference result output from the generating AI to be converted into an expression tailored to the user that the AI ​​agent has learned and presented to the user.

[0055] -Variation- The information processing device according to the above-described embodiment can be modified as follows. (1) In the above-described embodiment, an example was described in which an AI agent was created to support the use of generative AI in an organization by learning in collaboration with the use of generative AI by individuals (users) belonging to an organization, thereby sharing individual knowledge and rules throughout the organization. However, the learning method of the AI ​​agent is not limited to this, as long as an AI agent capable of sharing individual knowledge and rules throughout the organization can be provided. For example, an AI agent may be created to support the use of generative AI in an organization by learning in collaboration with the use of generative AI by individuals (users) belonging to an organization in the development of AI and LLMs (large-scale language models), thereby sharing individual knowledge and rules throughout the organization. In this way, an AI agent can be provided that learns how the generative AI is used by users belonging to the organization in the development of AI and LLMs (large-scale language models) to support the use of generative AI in an organization.

[0056] It should be noted that the present invention is not limited to the configurations in the above-described embodiments, as long as the characteristic functions of the present invention are not impaired. [Explanation of symbols]

[0057] 100 Information processing device 101 Communication Module 102 Control device 103 Recording Device

Claims

1. A prompt input means for inputting a prompt input by a user to a trained AI agent that has learned the usage content of the generated AI by the user; a converted prompt acquisition means for acquiring a prompt converted by the AI ​​agent; a converted prompt input means for inputting the converted prompt acquired by the converted prompt acquisition means to the generation AI; an inference result acquisition means for acquiring an inference result output from the generation AI in response to input of the converted prompt; an inference result input means for inputting the inference result acquired by the inference result acquisition means to the AI ​​agent; A converted inference result acquisition means for acquiring the inference result converted by the AI ​​agent; An information processing apparatus comprising: a presentation unit that presents the converted inference result acquired by the converted inference result acquisition unit to a user.

2. 2. The information processing device according to claim 1, The trained AI agent is an information processing device characterized in that it has learned the content of the prompts that the user inputs to the generated AI and the user's actions in response to the inference results output from the generated AI as the user's use of the generated AI.

3. 3. The information processing device according to claim 2, An information processing device characterized in that the trained AI agent has learned content indicating what actions the user took in response to the output from the generating AI as the user's behavior in response to the inference results output from the generating AI.

4. 4. The information processing device according to claim 3, An information processing device characterized in that, when a user corrects the inference result output from the generation AI, the trained AI agent learns the user's correction as the user's action in response to the inference result output from the generation AI.

5. 4. The information processing device according to claim 3, This information processing device is characterized in that when a user asks a follow-up question about the inference result output from the generation AI, the trained AI agent learns the content of the user's follow-up question as the user's behavior in response to the inference result output from the generation AI.

6. In the information processing device according to claim 1, The information processing device is characterized in that the presentation means, when presenting the inference results to the user, explains the intention behind the generation by the generation AI.

7. An inference result acquisition means for acquiring an inference result output from the generation AI; an inference result input means for inputting the inference result acquired by the inference result acquisition means into a trained AI agent that has learned how the generated AI is used by a user; A converted inference result acquisition means for acquiring the inference result converted by the AI ​​agent; a presentation means for presenting the converted inference result acquired by the converted inference result acquisition means to a user, The information processing device is characterized in that the presentation means, when presenting the inference results to the user, explains the intention behind the generation by the generation AI.

8. A step in which a prompt input means inputs a prompt input by a user to a trained AI agent that has learned the usage content of the generated AI by the user; A step in which a converted prompt acquisition means acquires a prompt converted by the AI ​​agent; a step in which a converted prompt input means inputs the converted prompt acquired by the converted prompt acquisition means to the generation AI; an inference result acquisition means for acquiring an inference result output from the generation AI in response to input of the converted prompt; an inference result input means inputting the inference result acquired by the inference result acquisition means to the AI ​​agent; A step in which a converted inference result acquisition means acquires the inference result converted by the AI ​​agent; and a step in which a presentation means presents the converted inference result acquired by the converted inference result acquisition means to a user.

9. 9. The information processing method according to claim 8, An information processing method characterized in that the trained AI agent has been trained to learn the content of prompts input by the user to the generated AI and the user's actions in response to the inference results output from the generated AI as the user's use of the generated AI.

10. 10. The information processing method according to claim 9, An information processing method characterized in that the trained AI agent has been trained to learn what actions the user took in response to the output from the generating AI as the user's behavior in response to the inference results output from the generating AI.

11. 11. The information processing method according to claim 10, An information processing method characterized in that, when a user corrects the inference result output from the generation AI, the trained AI agent learns the user's correction as the user's action in response to the inference result output from the generation AI.

12. 11. The information processing method according to claim 10, An information processing method characterized in that, when a user asks a follow-up question about the inference result output from the generation AI, the trained AI agent learns the content of the user's follow-up question as the user's behavior in response to the inference result output from the generation AI.

13. The information processing method according to claim 8, The information processing method is characterized in that the presentation means, when presenting the inference results to the user, explains the intention behind the generation by the generation AI.

14. A step in which an inference result acquisition means acquires an inference result output from the generation AI; An inference result inputting means inputs the inference result acquired by the inference result acquiring means to a trained AI agent that has been trained on how the generated AI is used by a user; A step in which a converted inference result acquisition means acquires the inference result converted by the AI ​​agent; a presentation means for presenting the converted inference result acquired by the converted inference result acquisition means to a user, The information processing method is characterized in that the presentation means, when presenting the inference results to the user, explains the intention behind the generation by the generation AI.

15. An information processing program for causing a computer to execute the information processing method according to any one of claims 8 to 14.

Images