Information processing systems, computers, information processing methods, and programs

Abstract

To achieve a balance between cost and safety. [Solution] An information processing system that utilizes a generative AI to generate responses to prompts acquires a prompt, determines whether to connect to an on-premise or cloud-based generative AI for the prompt, inputs the prompt to the connected generative AI based on the determination result, obtains a response to the prompt from the connected generative AI that received the prompt, and outputs the response.

Description

[[Technical Field]]

[0001] The present invention relates to a technology effective for utilizing generative AI (Artificial Intelligence). [[Background Art]]

[0002] In recent years, generative AI has become increasingly widespread. For example, Patent Document 1 discloses a technology in which a script including target data and a command is input to generative AI, output data from the generative AI is stored in a database and presented to a user. In addition, Patent Document 2 discloses a technology that uses text-generating AI to evaluate the financial risk of a target company. [[Prior Art Documents]] [[Patent Documents]]

[0003] [[Patent Document 1]] Japanese Unexamined Patent Application Publication No. 2025-13259 [[Patent Document 2]] Japanese Patent No. 7626411 [[Summary of the Invention]] [[Problem to be Solved by the Invention]]

[0004] Even when a user has a desire to use generative AI, there are a certain number of users who also desire to avoid situations where data is uploaded to the cloud even temporarily. For such users, a conceivable solution is to always provide generative AI that can be used in a closed network environment (on-premises, in-house LAN (Local Area Network), VPN (Virtual private network), etc.), but this solution requires enormous resources and has not been practical. Therefore, there is a need for technology that balances cost and security by using a closed network environment only when generating AI with information that users wish to keep confidential, and an open network environment (Internet, Wi-Fi, SaaS (Software as a Service) services, etc.) when generating AI with other information. However, the technologies described in Patent Documents 1 and 2 did not include a mechanism for using different network environments based on whether or not the information was to be kept confidential.

[0005] Therefore, the inventor focused on a mechanism that appropriately determines whether the network environment is closed or open in response to a prompt, and then uses the network environment based on the determination result to obtain an answer from the generating AI.

[0006] In view of these challenges, the present invention aims to provide an information processing system, computer, information processing method, and program that can achieve both cost and safety. [Means for solving the problem]

[0007] The present invention is an information processing system that utilizes generative AI to generate responses to prompts, A prompt acquisition unit that acquires the prompt, The aforementioned prompt is configured to determine whether to connect to an on-premise generation AI or a cloud-based generation AI, and A prompt input unit that inputs the aforementioned prompt to the AI ​​that generates the connection destination based on the judgment result, A response acquisition unit obtains a response to the prompt from the connection destination generation AI that has received the prompt, A response output unit that outputs the aforementioned response, We provide an information processing system equipped with the following features.

[0008] According to the present invention, the system determines whether to use an on-premise generation AI or a cloud-based generation AI as the destination generation AI in response to a prompt, and obtains a response to the prompt from the destination generation AI based on the determination result. As a result, the system will determine whether the network environment is closed or open based on the prompt, allowing security to be maintained without requiring massive resources, thus enabling a balance between cost and security.

[0009] Although this invention falls under the category of systems, it will produce similar functions and effects in other categories such as computers, methods, and programs. [Effects of the Invention]

[0010] According to the present invention, it is possible to achieve both cost and safety. [Brief explanation of the drawing]

[0011] [Figure 1] This is a diagram illustrating the overview of Information Processing System 1. [Figure 2] This diagram shows the functional configuration of information processing system 1. [Figure 3] This diagram shows a flowchart of the response processing performed by computer 10. [Figure 4] This diagram schematically shows the prompts that computer 10 receives. [Figure 5] This diagram shows a flowchart of the first connection destination determination process performed by computer 10. [Figure 6] This diagram schematically shows the classification prompts generated by computer 10. [Figure 7] This diagram shows a flowchart of the second connection destination determination process performed by computer 10. [Figure 8] This diagram schematically shows the decision prompts generated by computer 10. [Modes for carrying out the invention]

[0012] Hereinafter, modes for carrying out the present invention (hereinafter, embodiments) will be described in detail with reference to the accompanying drawings. In the following figures, the same elements are denoted by the same numbers or reference signs throughout the entire description of the embodiments.

[0013] Note that "generative AI" as used herein generally refers to generative artificial intelligence technology that generates new data and content (data including but not limited to text characters, such as images, videos, programs, etc.) based on given input data, and may be implemented by a chatbot utilizing a large language model.

[0014] [Outline of Information Processing System 1] FIG. 1 is a schematic diagram for explaining the outline of the information processing system 1. Components of the information processing system 1 will be described based on FIG. 1. The information processing system 1 is a system comprising at least a computer 10 that has a server function and utilizes generative AI for generating a response to a prompt. In the present embodiment, in addition to the computer 10, the information processing system 1 includes a user terminal 3 used by a user who asks questions to generative AI and obtains answers, an on-premises computer 4 that provides on-premises generative AI (generative AI provided as a closed network environment), and a cloud computer 5 that provides cloud generative AI (generative AI provided as an open network environment).

[0015] The user terminal 3 is a terminal device used by a user, for example, terminal devices such as mobile phones, smart phones, tablet terminals, personal computers and laptop computers, or wearable terminals such as smart watches, smart glasses and HMD (Head Mounted Display).

[0016] On-premise computer 4 is a computer with server functionality located in a closed network environment. On-premise computer 4 stores vector data and file summaries of confidential information for companies, organizations, and individuals, including important information (customer information, personnel information, financial information, business plans, product technology information, system information, etc.), confidential information (meeting minutes, customer lists, proposals, contracts, personnel information, accounting information, etc.), and personal information (name, gender, age, address, telephone number, job title, wages, family structure, etc.). On-premise computer 4 also has the functionality of an on-premise type generative AI. There is only one on-premise computer 4 within the system.

[0017] Cloud Computer 5 is a computer that exists in an open network environment and has server functionality. Cloud Computer 5 has the functionality of cloud-based generative AI, such as SaaS-type generative AI.

[0018] Computer 10 has server functionality and may be implemented using a single computer, or it may be implemented using multiple computers, like a cloud computer. In this specification, a cloud computer may be either a system that uses any computer in a scalable manner to perform a specific function, or a system that includes multiple functional modules to realize a certain system, and uses those functions in a freely combined manner. Furthermore, the information processing system 1 may include other terminals and devices in addition to the user terminals 3, on-premise computers 4, cloud computers 5, and computers 10 mentioned above. The number, types, and functions of these devices are not particularly limited and can be designed as appropriate.

[0019] This invention utilizes generative AI (such as a large-scale language model that performs natural language processing). The generative AI only needs to be capable of outputting a response to a predetermined prompt that is input. In the information processing system 1, computer 10 has the function of generating AI, and on-premise computer 4 and cloud computer 5 also have the function of generating AI.

[0020] When computer 10 utilizes the functions of its own generative AI, it inputs a predetermined prompt into the generative AI and obtains the response output by the generative AI. Furthermore, when computer 10 utilizes the generative AI functions of on-premise computer 4 or cloud computer 5, it outputs a prompt to on-premise computer 4 or cloud computer 5, obtains the response obtained by on-premise computer 4 or cloud computer 5 as a result of on-premise computer 4 or cloud computer 5 inputting this prompt into the generative AI, and obtains the response to this prompt.

[0021] The details concerning the generative AI itself, such as the data it learns from, its learning method, and the method for constructing the learning model, are not particularly limited. Furthermore, the generative AI may be general-purpose or it may be tuned according to its intended use and method of use, and is not particularly limited. For the reasons stated above, and because this invention is for utilizing generative AI, a detailed explanation of the generative AI is omitted in order to simplify the description in the specification.

[0022] This section outlines the processing steps involved when information processing system 1 utilizes generative AI to generate responses to prompts.

[0023] Computer 10 receives a prompt (step S1). Computer 10 obtains the prompt entered by the user from user terminal 3.

[0024] Computer 10 determines whether to connect to an on-premises generative AI or a cloud-based generative AI as the target generative AI (Step S2). Computer 10 determines the destination AI based, for example, whether the wording of the acquired prompt matches or approximates a set of keywords that correspond to confidential information. Furthermore, computer 10 inputs the acquired prompts into a classification generation AI that classifies prompts based on whether or not they contain confidential information, and determines the destination generation AI based on the response. Furthermore, the computer 10 inputs the acquired prompt into a judgment generation AI that determines the appropriate connection destination generation AI, and determines the connection destination generation AI based on the response.

[0025] This explains one specific example of how computer 10 determines the AI ​​that will generate the connection destination. Computer 10 performs various language analysis processes on the acquired prompt, such as summarizing confidential information and generating keywords. Computer 10 vectorizes the summary of confidential information, keywords, etc., compares these vectors with pre-registered confidential information vector data, identifies the registered file actual summary corresponding to the acquired prompt, and analyzes and identifies the content of the confidential information contained in the acquired prompt. Computer 10 generates a prompt using this analysis and specific result, and inputs this generated prompt into the classification generation AI and the decision generation AI. Based on the responses from the classification generation AI and the decision generation AI, Computer 10 determines which generation AI to connect to.

[0026] Computer 10 inputs a prompt to the AI ​​that generates the connection destination based on the judgment result (step S3). If computer 10 determines that the target generation AI is an on-premises generation AI, it outputs a prompt to on-premises computer 4. On-premises computer 4 receives this prompt and inputs the received prompt to the on-premises generation AI. On the other hand, if computer 10 determines that the destination generation AI is a cloud-based generation AI, it outputs a prompt to cloud computer 5. Cloud computer 5 receives this prompt and inputs the received prompt to the cloud-based generation AI.

[0027] Computer 10 obtains the response to the prompt from the connection generation AI that entered the prompt (step S4). Upon receiving a prompt, the on-premises computer 4 and the cloud computer 5 generate an answer to this prompt using the functions of the on-premises generative AI and the cloud-based generative AI, and output the generated answer to computer 10. Computer 10 receives this response and obtains the answer to the prompt from the connection generation AI that entered the prompt.

[0028] Computer 10 outputs the answer (step S5). Computer 10 outputs the acquired response to user terminal 3. User terminal 3 receives this response and displays it on its own display unit, etc. Computer 10 outputs the answer by displaying it on user terminal 3.

[0029] The above is an overview of Information Processing System 1. According to this information processing system 1, a prompt is input to the appropriate generating AI, either an on-premise generating AI or a cloud-based generating AI, and a response is obtained. For example, if the prompt contains confidential information, the prompt is input to the on-premise generating AI; if the prompt does not contain confidential information, the prompt is input to the cloud-based generating AI and a response is obtained. As a result, safety can be maintained without requiring enormous resources, making it possible to balance cost and safety.

[0030] [Device configuration] Figure 2 is a block diagram showing the configuration of Information Processing System 1. The device configuration of Information Processing System 1 will be described based on Figure 2. The information processing system 1 consists of at least a computer 10 that utilizes generative AI to generate responses to prompts. In this embodiment, the information processing system 1 consists of the computer 10, a user terminal 3, an on-premise computer 4, and a cloud computer 5.

[0031] Information processing system 1 is a system in which computer 10 is connected to user terminals 3, on-premise computers 4, and cloud computers 5 via a network 8 such as the Internet, internal LAN, Wi-Fi, or VPN, enabling data communication. Here, the on-premises computer 4 is connected to computer 10 via a closed network environment such as a company LAN or VPN, enabling data communication. The cloud computer 5 is also connected to computer 10 via an open network environment such as the internet or Wi-Fi, enabling data communication. The user terminal 3 may be connected to computer 10 via a closed network environment or via an open network environment, enabling data communication. Furthermore, the information processing system 1 may include other terminals and devices in addition to the user terminal 3, on-premise computer 4, cloud computer 5, and computer 10, and the number, types, and functions of these other terminals and devices can be designed as appropriate.

[0032] User terminal 3 is the terminal device used by the user as described above. User terminal 3 includes a terminal control unit with a CPU (Central Processing Unit), GPU (Graphics Processing Unit), RAM (Random Access Memory), ROM (Read Only Memory), etc., and a communication unit with devices that enable communication with other terminals and devices. Furthermore, the user terminal 3 is equipped with various devices as an input / output unit that receive predetermined inputs, input and output various types of data, etc.

[0033] The on-premise computer 4 has server functionality and is implemented as a single computer that is connected to computer 10 for data communication within a closed network environment (such as a company LAN or VPN). The on-premise computer 4 includes a control unit with a CPU, GPU, RAM, ROM, etc., and a communication unit with devices that enable communication with other terminals and equipment. On-premise computer 4 is equipped with a data storage unit consisting of a hard disk, semiconductor memory, recording media, memory card, etc., and the on-premise generation AI is registered there. The on-premise computer 4, as a processing unit, is equipped with various devices for executing various processes, as well as functions such as outputting responses to prompts by an on-premise generation AI.

[0034] The cloud computer 5 is implemented using one or more computers that have server functionality and are connected to computer 10 for data communication in an open network environment (Internet, Wi-Fi, etc.). Cloud Computer 5 includes a control unit comprising a CPU, GPU, RAM, ROM, etc., and a communication unit comprising devices that enable communication with other terminals and equipment. Cloud Computer 5 is equipped with a data storage unit consisting of a hard disk, semiconductor memory, recording media, memory card, etc., and a cloud-based generative AI is registered there. Cloud Computer 5, as a processing unit, is equipped with various devices for executing various processes, as well as functions such as outputting responses to prompts using cloud-based generative AI.

[0035] Computer 10 has server functionality and may be implemented using a single computer, or it may be implemented using multiple computers, like a cloud computer. Computer 10 includes a CPU, GPU, RAM, ROM, etc. as a control unit, and a communication unit that includes a device for communicating with other terminals and devices, a prompt acquisition unit for acquiring prompts, a prompt input unit for inputting prompts to a connection destination generation AI based on the judgment result, a response acquisition unit for acquiring a response to the prompt from the connection destination generation AI that received the prompt, and a response output unit for outputting the response. Computer 10 includes a data storage unit, such as a hard disk, semiconductor memory, recording media, or memory card, as its storage unit. The computer 10 includes, as a processing unit, various devices for executing various processes, a connection destination determination unit that determines whether to connect to an on-premise generation AI or a cloud-based generation AI as the destination generation AI, and so on.

[0036] In computer 10, the control unit loads a predetermined program and, in cooperation with the communication unit, realizes a prompt acquisition module, a prompt input module, an answer acquisition module, and an answer output module. Furthermore, in the computer 10, the control unit loads a predetermined program and, in cooperation with the processing unit, realizes a connection destination determination module, a prompt generation module, a prompt input module, a classification result acquisition module, a classification module, and a determination result acquisition module.

[0037] The following describes each process performed by Information Processing System 1, along with the processes performed by each module mentioned above. In this specification, each module may perform its processing as a function of its own, or it may perform it through a predetermined application.

[0038] [Answer processing performed by computer 10] Based on Figure 3, the response processing performed by computer 10 will be described. The diagram shows a flowchart of the response processing performed by computer 10. The response processing is detailed as follows: prompt acquisition processing (step S1) to obtain a prompt; connection destination determination processing (step S2) to determine whether to connect to an on-premise generation AI or a cloud-based generation AI; prompt input processing (step S3) to input the prompt to the connection destination generation AI based on the determination result; response acquisition processing (step S4) to obtain a response to the prompt from the connection destination generation AI to which the prompt was input; and response output processing (step S5) to output the response.

[0039] The prompt acquisition module acquires the prompt (step S10). A prompt is a command given by the user to use the generated AI, and can include questions, explanations, instructions, requests, summaries, etc. Furthermore, there are no particular restrictions on the type of prompt; any prompt entered by the user as appropriate is acceptable. The prompt acquisition module acquires the prompt from user terminal 3 (see Figure 4). The prompt acquisition module acquires prompts from the user through a predetermined UI (User Interface), such as a chatbot, which communicates with the user in an interactive manner. The user terminal 3 receives prompt input from the user via a predetermined UI, etc., and transmits the received prompt to the computer 10. The prompt acquisition module receives this prompt and acquires the prompt.

[0040] Based on Figure 4, the prompts acquired by the prompt acquisition module will be explained. This figure schematically shows the prompts acquired by the prompt acquisition module. In this figure, prompt 20 is shown.

[0041] Prompt 20 uses the example: "You are an employee of the sales department. Please provide a summary of this quarter's operating profit and a forecast for next quarter's operating profit." The prompt acquisition module uses this prompt 20 as the original prompt (the prompt entered by the user) and executes the connection destination determination process described later.

[0042] Returning to Figure 3, let's continue explaining the response processing. Computer 10 performs a connection destination determination process to determine whether to connect to an on-premises generation AI or a cloud-based generation AI (step S11). In step S11, the computer 10 executes the first connection destination determination process and / or the second connection destination determination process described later, and determines the connection destination generation AI that will be the input destination for the prompt.

[0043] <First connection destination determination process> Based on Figure 5, the first connection destination determination process performed by computer 10 will be explained. The same figure is a flowchart of the first connection destination determination process.

[0044] The prompt generation module generates a prompt for classification (step S100). The classification prompt is a prompt to be input to the classification generation AI of the computer 10, and is a prompt for classifying the prompt (original prompt) obtained by the processing in step S10. The classification prompt is, for example, a prompt to cause the classification generation AI to answer whether or not the original prompt contains confidential information, and consists of the original prompt and an instruction statement that instructs the classification generation AI to classify this original prompt (see Figure 6).

[0045] Let's look at another specific example of what the prompt generation module does. The prompt generation module performs various language analysis processes on the acquired prompts, such as summarizing confidential information and generating keywords. The prompt generation module vectorizes the processing results and compares the vectorized processing results with pre-registered vector data of confidential information. The prompt generation module, based on the comparison results, identifies a summary of the registered file corresponding to the acquired prompt, and analyzes and identifies the content of the confidential information contained in the acquired prompt. The prompt generation module generates classification prompts based on these analysis and identification results.

[0046] The classification generation AI is a function of computer 10, and in response to a classification prompt, it is a generation AI that classifies whether or not the original prompt contains confidential information as a response. The specific process by which the classification generation AI classifies the original prompt may be the same as a general generation AI, which outputs a response to the input prompt, or it may be based on the similarity between the original prompt contained in the classification prompt and a pre-set set of keywords corresponding to confidential information, classifying it as containing confidential information if there is a match or approximation, and classifying it as not containing confidential information if there is a mismatch or approximation, and outputting each as a response.

[0047] Furthermore, the classification prompts are not limited to the examples described above and can be designed as appropriate. Furthermore, the classification-based generation AI is not limited to the examples described above and can be designed as needed. Furthermore, the classification generation AI may be configured not only to have its function on computer 10, but also on an on-premise computer 4.

[0048] Based on Figure 6, the classification prompts generated by the prompt generation module will be explained. This figure schematically shows the classification prompts generated by the prompt generation module. In this figure, classification prompt 30 is shown.

[0049] The classification prompt 30 consists of the original prompt 31 and the instruction statement 32. Original prompt 31 is a prompt entered by the user, and consists of a prompt indicating that the user entered it and the prompt that was actually entered. Original prompt 31 is "The prompt entered by the user is 'You are an employee of the sales department. Please provide a summary of this quarter's operating profit and a forecast for next quarter's operating profit.'" Of these, the prompt indicating that the user entered it is the phrase "The prompt entered by the user is ~~". The prompt that was actually entered is the phrase sandwiched between the prompt indicating that the user entered it ("You are an employee of the sales department. Please provide a summary of this quarter's operating profit and a forecast for next quarter's operating profit."). Command 32 is a prompt for the classification generation AI to classify the original prompt 31, and consists of commands for classifying the original prompt 31. The command in question is, "Classify this prompt based on whether or not it contains confidential information and provide a response."

[0050] Returning to Figure 5, let's continue explaining the first connection destination determination process. The prompt input module inputs a classification prompt to the classification generation AI (step S101), and the classification result acquisition module acquires the classification result (step S102). The prompt input module inputs the generated classification prompts to the classification generation AI. The classification generation AI classifies whether the original prompt contains confidential information based on the classification prompt, and outputs the classification result as the answer. The classification result acquisition module retrieves the answer and obtains the classification result answer for the original prompt.

[0051] The classification module classifies the original prompt based on the acquired classification results (step S103). The classification module determines whether the original prompt contains confidential information based on the response output by the classification generation AI.

[0052] For example, if operating profit is confidential information, the classification generation AI will output a response indicating that the original prompt contains confidential information. Therefore, the classification module will classify the original prompt 31 as a prompt containing confidential information. On the other hand, if operating profit is not confidential information, the classification generation AI outputs a response indicating that the original prompt does not contain confidential information. Therefore, the classification module classifies the original prompt 31 as a prompt that does not contain confidential information.

[0053] The connection destination determination module determines the AI ​​for generating the connection destination based on the classification result (step S104). The connection destination determination module determines the connection destination generation AI to be an on-premises generation AI if the original prompt is classified as containing confidential information, and determines the connection destination generation AI to be a cloud-based generation AI if the original prompt is classified as not containing confidential information.

[0054] Computer 10 uses the determined connection destination generation AI to execute the processes from step S12 onward.

[0055] The above is the first connection destination determination process.

[0056] <Second connection destination determination process> Based on Figure 7, the second connection destination determination process performed by computer 10 will be explained. This figure is a flowchart of the second connection destination determination process.

[0057] The prompt generation module generates a decision prompt (step S200). The decision prompt is a prompt to be input to the decision generation AI of computer 10, and is a prompt for determining the appropriate destination generation AI as the input destination for the prompt (original prompt) obtained by the processing in step S10. The decision prompt is, for example, a prompt for determining whether an on-premise generation AI or a cloud-based generation AI is more appropriate as the appropriate destination generation AI to be input to, and consists of an original prompt and a command statement that instructs the decision generation AI to determine the appropriate destination generation AI as the input destination for this original prompt (see Figure 8).

[0058] Let's look at another specific example of what the prompt generation module does. The prompt generation module performs various language analysis processes on the acquired prompts, such as summarizing confidential information and generating keywords. The prompt generation module vectorizes the processing results and compares the vectorized processing results with pre-registered vector data of confidential information. The prompt generation module, based on the comparison results, identifies a summary of the registered file corresponding to the acquired prompt, and analyzes and identifies the content of the confidential information contained in the acquired prompt. The prompt generation module generates decision prompts based on these analysis and identification results.

[0059] The decision-making generation AI is a function of computer 10, and is a generation AI that determines a destination generation AI capable of generating an appropriate response to a decision prompt. The specific process by which the decision-making generation AI determines a suitable destination generation AI as the input destination for the original prompt may be to output a response to the input prompt, as is the case with general generation AIs, or it may be to determine an on-premise generation AI as a suitable destination generation AI based on the similarity between the original prompt contained in the decision prompt and a set of keywords corresponding to pre-configured confidential information, and output each as a response if there is a match or approximation, and if there is a mismatch or approximation, determine a cloud-based generation AI as a suitable destination generation AI.

[0060] Furthermore, the decision prompts are not limited to the examples described above and can be designed as appropriate. Furthermore, the AI ​​used for decision-making is not limited to the examples described above and can be designed as appropriate. Furthermore, the decision-making generation AI may be configured not only to have its function on computer 10, but also on an on-premise computer 4.

[0061] Based on Figure 8, the decision prompts generated by the prompt generation module will be explained. This figure schematically shows the decision prompts generated by the prompt generation module. In this figure, decision prompt 40 is shown.

[0062] The decision prompt 40 consists of the original prompt 41 and the instruction statement 42. Original prompt 41 is a prompt entered by the user, and consists of a prompt indicating that the user entered it and the prompt that was actually entered. Original prompt 41 is "The prompt entered by the user is 'You are an employee of the sales department. Please provide a summary of this quarter's operating profit and a forecast for next quarter's operating profit.'" Of these, the prompt indicating that the user entered it is the phrase "The prompt entered by the user is ~~". The prompt that was actually entered is the phrase sandwiched between the prompt indicating that the user entered it ("You are an employee of the sales department. Please provide a summary of this quarter's operating profit and a forecast for next quarter's operating profit."). Command 42 is a prompt that causes the decision-making generation AI to determine which connection destination generation AI is appropriate as the input destination for the original prompt 41, and consists of a command that causes the AI ​​to determine which connection destination generation AI is appropriate as the input destination for the original prompt 41. The command in question is: "This prompt asks you to answer which of the generation AIs, one on OpenAI or one on the company's internal server, will provide higher response accuracy and response density."

[0063] Returning to Figure 7, let's continue explaining the second connection destination determination process. The prompt input module inputs a decision prompt to the decision generation AI (step S201), and the decision result acquisition module acquires the decision result (step S202). The prompt input module inputs the generated decision prompts to the decision-making AI. The decision-making generation AI determines the appropriate destination generation AI based on the decision prompt, specifically whether the original prompt contains confidential information, and outputs the decision result as an answer. If the original prompt contains confidential information, the decision-making generation AI determines the appropriate destination generation AI to be an on-premises generation AI; if the original prompt does not contain confidential information, it determines the appropriate destination generation AI to be a cloud-based generation AI.

[0064] For example, if operating profit is confidential information, the decision-making generation AI will determine that the appropriate destination generation AI for the original prompt 41 is the on-premise generation AI. On the other hand, if operating profit is not confidential information, the judgment generation AI will determine that the appropriate destination generation AI for input to the original prompt 41 is the cloud-based generation AI. The AI ​​for generating decisions outputs the decision result as the answer. The judgment result acquisition module retrieves the answer and obtains the judgment result of the original prompt.

[0065] The connection destination determination module determines the AI ​​for generating the connection destination based on the determination result (step S203). The connection destination determination module determines the destination generation AI to be an on-premises generation AI if the original prompt's determination result is an on-premises generation AI, and determines the destination generation AI to be a cloud-based generation AI if the original prompt's determination result is a cloud-based generation AI.

[0066] Computer 10 uses the determined connection destination generation AI to execute the processes from step S12 onward.

[0067] The above is the second connection destination determination process.

[0068] Computer 10 may execute either the first connection destination determination process or the second connection destination determination process, or it may execute both processes simultaneously. When both are performed simultaneously, computer 10 can determine the destination generation AI based on both the classification result and the judgment result.

[0069] If computer 10 determines that the classification result and the judgment result are the same destination generation AI, it should determine this destination generation AI as the input destination for the original prompt.

[0070] On the other hand, if the system determines that the destination generation AI differs from the classification result, it may prioritize the on-premise generation AI as the destination generation AI. Furthermore, in this case, the system may also be configured to ask the user which destination generation AI they would like to use.

[0071] Furthermore, the computer 10 can also be configured to determine the destination generation AI in ways other than performing the determination of the destination generation AI using the generation AI, such as the first destination determination process and the second destination determination process. Let me explain this case.

[0072] First, one case is determining whether or not the prompt obtained by the process in step S10 (the original prompt) contains confidential information. In this case, the connection destination determination module determines, based on the similarity between the original prompt and the pre-configured set of keywords corresponding to confidential information, that confidential information is included if they match or are similar, and determines the connection destination generation AI to be an on-premise generation AI. If there is no match or similarity, it determines that confidential information is not included, and determines the connection destination generation AI to be a cloud-based generation AI.

[0073] In addition, the connection destination determination module can also be configured to determine the connection destination generation AI based on wording that the user has included in the prompt beforehand, such as "use on-premise generation AI," "do not use on-premise generation AI," "use cloud-based generation AI," or "do not use cloud-based generation AI." However, in this case, even if the user has specified a connection destination generation AI, it may be inappropriate (for example, if the wording specifies the use of a cloud-based generation AI despite containing confidential information). Therefore, to exclude such cases, it is also possible to configure the system to perform the first connection destination determination process and the second connection destination determination process in parallel.

[0074] Even in such cases, it goes without saying that a configuration combining the first connection destination determination process and / or the second connection destination determination process is also possible.

[0075] Returning to Figure 3, let's continue explaining the response processing. The prompt input module inputs a prompt to the connection destination generation AI (step S12), and the response acquisition module acquires the response from the connection destination generation AI (step S13). The prompt input module inputs the acquired prompt (original prompt) to the connection destination generation AI, which was determined by the connection destination determination process. The prompt input module sends a prompt to a computer with the functionality of the connection destination generation AI (on-premise computer 4 if the connection destination generation AI is an on-premise generation AI, or cloud computer 5 if the connection destination generation AI is a cloud-based generation AI), and inputs the prompt to the connection destination generation AI. A computer with connection destination generation AI functionality receives a prompt, and the connection destination generation AI generates a response to this prompt. The computer with connection destination generation AI functionality sends the generated response to computer 10. The response acquisition module receives this response and retrieves the response from the AI ​​that generates the connection destination.

[0076] The response output module outputs the response (step S14). The response output module outputs the acquired responses via a predetermined UI, such as a chatbot, which communicates with the user through a dialogue format. The response output module sends the response to user terminal 3. User terminal 3 receives this response and displays it on its own display unit via a designated UI or the like. The response output module outputs the response by displaying it on the user terminal 3.

[0077] The above is the response process.

[0078] The means and functions described above are realized by a computer (including the CPU, information processing unit, and various terminals) reading and executing a predetermined program. The program may be provided, for example, via a network from the computer (SaaS: Software as a Service) or as a cloud service. Alternatively, the program may be provided in a form recorded on a computer-readable recording medium. In this case, the computer reads the program from the recording medium, transfers it to an internal or external recording device, records it, and executes it. Alternatively, the program may be pre-recorded on a recording device (recording medium) and provided to the computer from that recording device via a communication line.

[0079] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above. Furthermore, the effects described in the embodiments of the present invention are merely a list of the most preferred effects arising from the present invention, and the effects of the present invention are not limited to those described in the embodiments of the present invention.

[0080] A first aspect disclosed in this embodiment is an information processing system that utilizes generative AI to generate responses to prompts, A prompt acquisition unit that acquires the prompt, The aforementioned prompt is configured to determine whether to connect to an on-premise generation AI or a cloud-based generation AI, and A prompt input unit that inputs the aforementioned prompt to the AI ​​that generates the connection destination based on the judgment result, A response acquisition unit obtains a response to the prompt from the connection destination generation AI that has received the prompt, A response output unit that outputs the aforementioned response, We provide an information processing system equipped with the following features.

[0081] A second aspect disclosed in this embodiment is a classification unit for classifying the prompts, Furthermore, The connection destination determination unit determines the connection destination generation AI based on the classification result. The present invention provides an information processing system as described in the first embodiment.

[0082] A third aspect disclosed in this embodiment includes a classification input unit that inputs the prompts to a classification generation AI that classifies the prompts, A classification result acquisition unit obtains the classification result answer for the prompt from a classification generation AI that has been input with the prompt, Furthermore, The classification unit classifies the prompts based on the acquired classification results. The present invention provides an information processing system as described in the second embodiment.

[0083] A fourth aspect disclosed in this embodiment is that the connection destination determination unit determines, based on the classification result, that the on-premise generation AI is the connection destination generation AI if the prompt contains confidential information, and that the cloud-based generation AI is the connection destination generation AI if it does not contain confidential information. The present invention provides an information processing system as described in the second embodiment.

[0084] A fifth aspect disclosed in this embodiment is that the connection destination determination unit determines that the generation AI that generates an appropriate response to the prompt is the connection destination generation AI. The present invention provides an information processing system as described in the first embodiment.

[0085] A sixth aspect disclosed in this embodiment includes a decision input unit that inputs the prompt to a decision generation AI that determines a generation AI that generates an appropriate answer, A judgment result acquisition unit that acquires the judgment result of the generation AI that generates an appropriate answer from the judgment generation AI that has received the aforementioned prompt, Furthermore, The connection destination determination unit determines the connection destination generation AI based on the response of the determination result. The present invention provides an information processing system as described in the fifth aspect.

[0086] A seventh aspect disclosed in this embodiment is that the connection destination determination unit determines, based on the determination result, that the on-premise generation AI is the connection destination generation AI if the prompt contains confidential information, and that the cloud-based generation AI is the connection destination generation AI if it does not contain confidential information. The present invention provides an information processing system as described in the fifth aspect. [Explanation of symbols]

[0087] 1. Information Processing System 3. User terminals 4. On-Premise Computers 5. Cloud Computing 8 Networks 10 Computers 20 Prompts 30 Classification Prompts 31 Original prompt 32 Imperative sentence 40 Decision prompt 41 Original prompt 42 Imperative sentence

Claims

1. An information processing system that utilizes generative AI to generate responses to prompts, A prompt acquisition unit that acquires the prompt, The aforementioned prompt is configured to determine whether to use an on-premise generation AI or a cloud-based generation AI as the destination generation AI, and A prompt input unit inputs the aforementioned prompt to the AI ​​that generates the connection destination based on the judgment result, A response acquisition unit obtains a response to the prompt from the connection destination generation AI that has received the prompt, A response output unit that outputs the aforementioned response, An information processing system equipped with the following features.

2. A classification unit that classifies the aforementioned prompts, Furthermore, The connection destination determination unit determines the connection destination generation AI based on the classification result. The information processing system according to claim 1.

3. A classification generation AI that classifies the aforementioned prompts includes a classification input unit that inputs the aforementioned prompts, A classification result acquisition unit obtains the classification result answer for the prompt from the classification generation AI that has been inputted the prompt, Furthermore, The classification unit classifies the prompts based on the acquired classification results. The information processing system according to claim 2.

4. Based on the classification result, the connection destination determination unit determines the on-premise generation AI to be the connection destination generation AI if the prompt contains confidential information, and determines the cloud-based generation AI to be the connection destination generation AI if the prompt does not contain confidential information. The information processing system according to claim 2.

5. The connection destination determination unit determines that the generating AI that generates an appropriate response to the prompt is the connection destination generating AI. The information processing system according to claim 1.

6. A judgment input unit inputs the aforementioned prompt to a judgment generation AI that determines which generation AI will generate an appropriate answer, A judgment result acquisition unit obtains the judgment result of the generation AI that generates an appropriate answer from the judgment generation AI that has received the aforementioned prompt, Furthermore, The connection destination determination unit determines the connection destination generation AI based on the response of the determination result. The information processing system according to claim 5.

7. Based on the determination result, the connection destination determination unit determines the on-premise generation AI to be the connection destination generation AI if the prompt contains confidential information, and determines the cloud-based generation AI to be the connection destination generation AI if the prompt does not contain confidential information. The information processing system according to claim 5.

8. A computer that utilizes generative AI to generate responses to prompts, A prompt acquisition unit that acquires the prompt, The aforementioned prompt is configured to determine whether to use an on-premise generation AI or a cloud-based generation AI as the destination generation AI, and A prompt input unit inputs the aforementioned prompt to the AI ​​that generates the connection destination based on the judgment result, A response acquisition unit obtains a response to the prompt from the connection destination generation AI that has received the prompt, A response output unit that outputs the aforementioned response, A computer equipped with [a specific feature / equipment].

9. An information processing method performed by a computer that utilizes generative AI to generate responses to prompts, Steps to obtain a prompt, The aforementioned prompt includes the step of determining whether to connect to an on-premises generation AI or a cloud-based generation AI as the destination generation AI, The steps include inputting the prompt to the AI ​​that generates the connection destination based on the judgment result, The steps include obtaining a response to the prompt from the connection destination generation AI that has received the prompt, The steps include outputting the aforementioned answer, An information processing method comprising the following:

10. Computers that utilize generative AI to generate responses to prompts, Steps to obtain a prompt, The step of determining whether the prompt should be an on-premise generation AI or a cloud-based generation AI, The step of inputting the prompt to the AI ​​that generates the connection destination based on the judgment result, A step of obtaining a response to the prompt from the connection destination generation AI that has entered the aforementioned prompt, Steps to output the above answer, A computer-readable program for executing [something].

Images