System

Abstract

To provide a system.SOLUTION: A system is provided, comprising means of inputting an identical question to multiple generative AI models, means of displaying a list of answers to the question generated by the multiple generative AI models, means of regenerating a new answer by combining the answers with respectively specified weights, and means of outputting the regenerated new answer.SELECTED DRAWING: Figure 1

Description

[Technical field]

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

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including a description and related instruction sentence regarding the chatbot character, 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] JP 2022-180282 A Summary of the Invention [Problem to be solved by the invention]

[0004] When the same question is posed to multiple generation AIs, it is difficult to efficiently recognize the differences in the answers from each generation AI and combine them to generate the optimal answer. [Means for solving the problem]

[0005] The present invention provides a means for asking the same question to multiple AI generators, displaying a list of the answers from each AI generator, recognizing the differences between each answer, and combining each answer in a ratio specified by the user to regenerate a hybrid answer. This makes it possible to efficiently recognize the differences between answers from each AI generator and generate an optimal answer. [Brief description of the drawings]

[0006] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Diagram 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. FIG. [Diagram 3] FIG. 11 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Diagram 5] FIG. 13 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 13 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 13 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 2 is a sequence diagram showing a process flow of the data processing system in the first embodiment of the first form example. [Figure 12] FIG. 11 is a sequence diagram showing a process flow of the data processing system in the application example 1 of the first embodiment. [Figure 13] FIG. 11 is a sequence diagram showing a process flow of the data processing system in the second embodiment of the second embodiment. [Figure 14] FIG. 13 is a sequence diagram showing a process flow of the data processing system in application example 2 of embodiment 2. [Figure 15] FIG. 13 is a sequence diagram showing a process flow of a data processing system in a third embodiment of the present invention. [Figure 16]FIG. 13 is a sequence diagram showing a process flow of the data processing system in the third application example of the third embodiment. [Figure 17] FIG. 13 is a sequence diagram showing the flow of processing in the data processing system in the first embodiment of the first form example when an emotion engine is combined. [Figure 18] FIG. 13 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Form Example 1 when combined with an emotion engine. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0008] First, the terms used in the following description will be explained.

[0009] In the following embodiments, a signed processor (hereinafter simply referred to as a "processor") may be one arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be one type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (TENSOR PROCESSING UNIT (registered trademark)).

[0010] In the following embodiments, a signed RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0011] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0012] In the following embodiments, a communication I / F (Interface) with a code is an interface including a communication processor and an antenna. The communication I / F controls communication between multiple computers. An example of a communication standard applied to the communication I / F is a wireless communication standard including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0013] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. In addition, in this specification, the same idea as "A and / or B" is also applied when three or more things are expressed by connecting them with "and / or."

[0014] [First embodiment]

[0015] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0016] 1, a 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.

[0017] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[0018] 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. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0019] The reception device 38 includes a touch panel 38A and a microphone 38B, and receives user input. The touch panel 38A detects contact with a pointer (e.g., a pen or a finger) to receive user input by the touch of the pointer. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0020] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (e.g., voice 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 voice according to instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a Complementary Metal-Oxide-Semiconductor (CMOS) image sensor or a Charge Coupled Device (CCD) image sensor.

[0021] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54.

[0022] FIG. 2 shows an example of main functions of the data processing device 12 and the smart device 14.

[0023] As shown in Fig. 2, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32. The specific process program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific process program 56 from the storage 32, and executes the read specific process program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific process program 56 executed on the RAM 30.

[0024] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0025] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores a reception output program 60. The reception output program 60 is used together with the specific processing program 56 by the data processing system 10. The processor 46 reads out the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0026] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0027] "Example 1"

[0028] In one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[0029] "Example 2"

[0030] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[0031] The process flow of each embodiment will be described below.

[0032] "Example 1"

[0033] Step 1: The user inputs a question through the question input interface. This question is then posed to multiple generation AIs.

[0034] Step 2: The system sends a question to each Generator AI, which answers the question using its own algorithm.

[0035] Step 3: The system receives answers from each AI generator. These answers are displayed in a list format. Specifically, the name of each AI generator and its answer are displayed side by side on the screen.

[0036] Step 4: The system compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using color coding and graphs based on the similarity.

[0037] Step 5: The user specifies the percentage of each answer to be used. For example, "Answer A will be 70%, Answer B will be 20%, and Answer C will be 10%."

[0038] Step 6: The system combines the answers based on the percentages specified by the user to generate a hybrid answer, which becomes the final output.

[0039] Example 1

[0040] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal".

[0041] Conventional generative AI systems rely on answers from a single generative AI, which limits the variety and accuracy of answers. In addition, there is no way to compare answers from multiple generative AIs, making it difficult for users to select the most appropriate answer. Furthermore, there is a lack of functionality to combine multiple answers to generate hybrid answers, which means that it is not possible to flexibly generate answers that meet the needs of users.

[0042] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0043] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually showing the differences, and a means for regenerating a hybrid answer by combining the answers from each generation AI based on a ratio specified by the user. This makes it possible to compare various answers from multiple generation AIs and select the optimal answer. It also makes it possible to flexibly generate hybrid answers according to the needs of the user.

[0044] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[0045] "Generative AI" is an artificial intelligence model that generates natural language responses based on a given prompt.

[0046] An "API endpoint" is an interface for sending questions to the generation AI and receiving answers.

[0047] "List format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen.

[0048] "Visually showing" means showing the differences in responses using visual means such as color coding or graphs.

[0049] "Similarity" is an index showing the degree of agreement between the contents of multiple answers.

[0050] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user.

[0051] The "proportion" is the ratio that indicates the extent to which the user will use the answers from each generation AI.

[0052] This invention is a system that asks the same question to multiple generative AI models and compares and integrates their answers. A specific embodiment of this system will be described below.

[0053] Question input interface

[0054] A user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript (registered trademark) and provides a form where the user can input a question in text format. When the user enters a question and clicks the "Submit" button, the question is sent to the server.

[0055] Submit a question

[0056] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt. The generative AI models used include, for example, natural language processing models and machine learning models.

[0057] Receiving and listing responses

[0058] The server receives the answers from each generative AI model and temporarily stores them. Next, the server generates a web page to display the received answers in a list format. This web page displays the name of each generative AI model and its answer side by side. The user can view this page in the web browser of the terminal.

[0059] Recognizing differences in answers

[0060] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity. This process is carried out using text analysis algorithms and data visualization tools.

[0061] Regenerating a hybrid answer

[0062] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[0063] Examples

[0064] Example of question input

[0065] The user enters the following question into the question entry interface:

[0066] "What do you think the future of AI is?"

[0067] Examples of prompt statements

[0068] Here is an example of a prompt to be input to the generative AI model:

[0069] "What do you think the future of AI is?"

[0070] Example of answer list display

[0071] Generative AI Model 1: “AI will play an important role in many fields in the future, including medicine and education.”

[0072] Generative AI Model 2: "AI will help make autonomous driving and smart cities a reality."

[0073] Generative AI Model 3: “AI will continue to evolve as a tool to make human life more convenient.”

[0074] Examples of visual display of response differences

[0075] The parts of the answers from Generative AI Model 1 and Generative AI Model 2 that are highly similar are displayed in green.

[0076] The areas where the answers from Generative AI Model 1 and Generative AI Model 3 have low similarity are displayed in red.

[0077] A concrete example of hybrid answer regeneration

[0078] The user specifies "70% of the answers from Generative AI Model 1, 20% from Generative AI Model 2, and 10% from Generative AI Model 3." The server combines each answer based on that ratio and generates a hybrid answer like this:

[0079] "Going forward, AI will play an important role in many fields, including healthcare and education, and will also contribute to the realization of autonomous driving and smart cities, continuing to evolve as a tool to make human life more convenient."

[0080] In this way, users can compare various answers from multiple generative AI models and select the optimal answer. It also enables flexible generation of hybrid answers according to the needs of users.

[0081] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0082] Step 1: Enter your question

[0083] The user opens a web browser on the terminal and accesses the question input interface. The user enters a question in the text box and clicks the "Send" button. This operation sends the question to the server.

[0084] Input: A question typed in by the user (e.g., "Tell me about the future of AI.")

[0085] Output: The question data sent to the server

[0086] Step 2: Submit your question

[0087] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt.

[0088] Input: The query data sent to the server

[0089] Output: A prompt sent to each generative AI model (e.g., "Tell me about the future of AI.")

[0090] Step 3: Receiving a response

[0091] The server receives answers from each generative AI model. Each generative AI model generates an answer based on the prompt sentence and returns it to the server. The server temporarily stores these answers.

[0092] Input: Answer data from each generative AI model

[0093] Output: Answer data stored on the server (e.g., "AI will play an important role in many fields, such as medicine and education, in the future.")

[0094] Step 4: List the responses

[0095] The server generates a web page to display the answers of each generative AI model in a list format. The user views this page in a web browser on their device.

[0096] Input: Answer data saved on the server

[0097] Output: A web page displayed in list format (Example: Generative AI model 1: "AI will play an important role in many fields in the future, such as medicine and education.")

[0098] Step 5: Recognize differences in responses

[0099] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity.

[0100] Input: Answer data saved on the server

[0101] Output: A web page with visual indication of the differences (e.g. high similarity is shown in green, low similarity in red)

[0102] Step 6: Regenerate the hybrid answer

[0103] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[0104] Input: The percentage of each answer specified by the user (e.g. 70% of the answers from AI model 1, 20% of the answers from AI model 2, and 10% of the answers from AI model 3)

[0105] Output: A web page with a hybrid answer (e.g., "AI will play an important role in many fields, such as medicine and education, and will continue to evolve as a tool to make human life more convenient, contributing to the realization of autonomous driving and smart cities.")

[0106] (Application example 1)

[0107] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0108] In systems using conventional generative AI models, there is a lack of a means to compare answers from multiple generative AI models and obtain the optimal answer. In addition, it is difficult to visually recognize the differences between answers from each generative AI model, and there is insufficient support for users to select the optimal answer. Furthermore, when combining answers from multiple generative AI models to generate a hybrid answer, a means is required that can flexibly respond based on the ratio specified by the user.

[0109] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0110] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for recognizing the difference between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for calculating and visually displaying the similarity between each answer, and a means for the user to specify the ratio for generating the hybrid answer. This makes it possible to compare answers from a plurality of generative AI models and obtain an optimal answer. In addition, by visually recognizing the difference between each answer, support for the user to select the optimal answer is enhanced. Furthermore, it becomes possible to flexibly generate a hybrid answer based on the ratio specified by the user.

[0111] A "generative AI model" is a type of artificial intelligence that uses natural language processing to generate answers to user questions or prompts.

[0112] The "means of asking questions" is a function that allows users to send questions to multiple generative AI models.

[0113] "Means of displaying answers in a list" is a function for displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[0114] A "means of recognizing differences" is a function that can compare answers from multiple generative AI models and identify the differences between them.

[0115] A "means for regenerating hybrid answers" is a function for generating new answers by combining answers from multiple generative AI models in a ratio specified by the user.

[0116] "Means for calculating similarity" is a function for numerically evaluating the similarity between answers from multiple generative AI models.

[0117] "Means for visual display" refers to a function for displaying the calculated similarity or difference in answers in a visual format such as a graph or color coding.

[0118] "Means for specifying proportions" refers to a function that enables users to specify the proportions in which answers from each generative AI model should be combined.

[0119] The system for implementing the present invention operates in cooperation with a server, a terminal, and a user. A specific embodiment of the system will be described below.

[0120] System Configuration

[0121] server

[0122] The server includes means for asking the same question to multiple generative AI models, means for displaying a list of answers from each generative AI model, means for recognizing differences between each answer, means for combining each answer in a ratio specified by a user to regenerate a hybrid answer, means for calculating and visually displaying the similarity of each answer, and means for a user to specify the ratio for generating a hybrid answer.

[0123] The server uses the following software and hardware:

[0124] Generative AI model: OpenAI(R) API

[0125] Data processing: Python program

[0126] Visualization: Matplotlib, Scikit-learn

[0127] Terminal

[0128] The terminal provides an interface for the user to enter questions and view generated answers. The terminal may be a smartphone, smart glasses, a head-mounted display, or the like.

[0129] User

[0130] Users input questions through their devices, review answers from generative AI models, and specify the proportions for generating hybrid answers.

[0131] Processing flow

[0132] 1. Entering a question: The user enters a question through the terminal, for example, a prompt such as "Please tell me some ideas for a new blog post."

[0133] 2. Question submission: The server submits the input question to multiple generative AI models.

[0134] 3. Obtaining answers: The server receives answers from each generative AI model.

[0135] 4. Displaying a list of answers: The server displays the answers from each generative AI model in a list format on the terminal.

[0136] 5. Difference Recognition: The server calculates the similarity of each answer and displays it visually, for example using a heatmap to show the similarity between answers.

[0137] 6. Specify Proportion: The user specifies what proportion of each answer should be combined.

[0138] 7. Generate hybrid answer: The server generates a hybrid answer based on the specified percentages and displays it on the terminal.

[0139] Examples

[0140] When a user types a question like "What are some ideas for a new blog post?", the server sends the question to multiple generative AI models, which can produce answers like these:

[0141] Generative AI Model A: "Latest trends in AI technology"

[0142] Generative AI Model B: "Advantages and disadvantages of remote work"

[0143] Generative AI Model C: "Sustainable Energy Future"

[0144] The server lists these answers and visually displays the similarity between the answers in a heat map. For example, the user can specify that "Latest trends in AI technology" should be combined at 70%, and the other two at 20% and 10%, respectively. The server generates a hybrid answer based on this specification and provides it to the user.

[0145] In this way, users can compare answers from multiple generative AI models to obtain the best answer. Also, visual recognition of the differences between each answer provides enhanced support for selecting the more appropriate answer. Furthermore, it is possible to flexibly generate hybrid answers based on the ratio specified by the user.

[0146] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0147] Step 1:

[0148] The user inputs a question through a terminal.

[0149] Input: The prompt the user enters (e.g. "Can you give me some ideas for new blog posts?")

[0150] Output: The prompt text entered

[0151] Specific operations: The user inputs a question using the input interface of the terminal and presses the send button.

[0152] Step 2:

[0153] The server sends the input question to multiple generative AI models.

[0154] Input: The prompt text entered by the user.

[0155] Output: The prompt sent to each generative AI model

[0156] Specific operation: The server uses the API to send prompt sentences to each generative AI model.

[0157] Step 3:

[0158] The server receives the answers from each generative AI model.

[0159] Input: The prompt sent to each generative AI model

[0160] Output: The answer from each generative AI model

[0161] Specific operation: The server receives the API response from each generative AI model and obtains the answer.

[0162] Step 4:

[0163] The server displays the answers from each generative AI model in a list format on the terminal.

[0164] Input: The answer from each generative AI model

[0165] Output: A list of answers displayed on the terminal

[0166] Specific operation: The server compiles the answers from each generative AI model into a single list and sends it to the terminal's display interface.

[0167] Step 5:

[0168] The server calculates the similarity of each answer and displays it visually.

[0169] Input: The answer from each generative AI model

[0170] Output: A heatmap showing similarity

[0171] Specific operation: The server calculates the vector of each answer, calculates the similarity using cosine similarity, and generates a heat map of the result and displays it on the terminal.

[0172] Step 6:

[0173] The user specifies how much of each answer should be combined.

[0174] Input: The percentage of each answer that you specify (e.g. 70%, 20%, 10%)

[0175] Output: Specified percentage

[0176] Specific behavior: The user uses the device interface to specify the percentage of each answer using sliders and input fields.

[0177] Step 7:

[0178] The server generates a hybrid answer based on the specified percentage and displays it on the terminal.

[0179] Input: Answers from each generative AI model, in user-specified proportions

[0180] Output: Hybrid answer

[0181] Specific operation: The server combines each answer based on the specified ratio to generate a new hybrid answer, and sends the result to the device for display.

[0182] Example 2

[0183] Next, a second embodiment of the second embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0184] Conventional generative AI systems lacked the means to compare answers from multiple generative AIs and effectively integrate them. As a result, users had to check the answers from each generative AI individually and compare and integrate them manually, which was inefficient. In addition, there was no way to visually recognize the differences in the answers from each generative AI, making it difficult to intuitively understand the differences in the quality and content of the answers.

[0185] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0186] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually indicating the differences, and a means for combining the answers from each generation AI in a ratio designated by the user to regenerate a hybrid answer. This enables the user to efficiently compare and integrate answers from multiple generation AIs and obtain an optimal hybrid answer while visually recognizing the differences.

[0187] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[0188] "Generative AI" is an artificial intelligence model that generates natural language answers based on input prompts.

[0189] A "list format" is a format in which multiple items are displayed side by side on one screen or page.

[0190] "Visually showing" means displaying information using visual means such as color coding or graphs.

[0191] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[0192] The "proportion specified by the user" is a number that specifies the proportion in which the user will combine answers from each generation AI.

[0193] A "server" is a computer system that performs processes such as receiving questions, sending questions to the generation AI, receiving answers, displaying the answers, comparing answers, and generating hybrid answers.

[0194] This invention is a system that asks the same question to multiple generation AIs and compares and integrates their answers. A specific embodiment of this system will be described below.

[0195] Question input interface

[0196] A user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and includes a text box for the user to enter a question and a submit button. When the user enters a question and clicks the submit button, the question is sent to the server.

[0197] Submit a question

[0198] The server sends the questions received from the user to multiple generation AIs. The server uses a Python HTTP request library (e.g., requests) to send questions to the API endpoints of each generation AI. Specifically, it sends the following prompt text to the generation AI:

[0199] "What do you think the future of AI is?"

[0200] Receiving a response

[0201] The server receives answers from each AI generator. It waits for a response from the API of each AI generator and stores the received answers. For example, it receives an answer from AI generator A saying, "AI will continue to evolve in the future and support people's lives in various fields."

[0202] List of responses

[0203] The server displays the answers received from each AI generator in a list format. Using an HTML template engine (e.g. Jinja2), the name of each AI generator and its answer are displayed side by side on a web page. For example, it is displayed as follows:

[0204] Generative AI A: "AI will continue to evolve and support people's lives in a variety of areas."

[0205] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[0206] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[0207] Recognizing differences in answers

[0208] The server compares the answers from each generation AI and visually displays the differences. It uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using colors and graphs based on the similarity. For example, if the answers of generation AI A and generation AI B show a high similarity, they are displayed in the same color.

[0209] Regenerating a hybrid answer

[0210] The user specifies the percentage of each answer that should be used as input. The server combines the answers based on that percentage to generate a hybrid answer. For example, if the user specifies "70% of the answers should be from answer A, 20% from answer B, and 10% from answer C," the server will generate the following hybrid answer:

[0211] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[0212] In this way, users can efficiently compare and integrate answers from multiple generative AIs, visually recognizing the differences, and arrive at the optimal hybrid answer.

[0213] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0214] Step 1:

[0215] The user opens a web browser on the device and accesses the question input interface. The user enters a question in the text box and clicks the send button. The input is the question entered by the user, and the output is the question data sent to the server. In concrete terms, the user enters "Tell me about the future of AI" and clicks the send button.

[0216] Step 2:

[0217] The server sends questions received from the user to multiple generation AIs. The input is the question data received from the user, and the output is the prompt text sent to each generation AI. The server uses Python's HTTP request library (e.g., requests) to send questions to the API endpoint of each generation AI. Specifically, the server sends the question "Tell me about the future of AI" to generation AI A, generation AI B, and generation AI C.

[0218] Step 3:

[0219] The server receives the answers from each generation AI. The input is the API response from each generation AI, and the output is the received answer data. The server waits for a response from the API of each generation AI and stores the received answers. In concrete terms, the server receives the answer from generation AI A that "AI will continue to evolve in the future and support people's lives in various fields."

[0220] Step 4:

[0221] The server displays the answers received from each generation AI in a list format. The input is the received answer data, and the output is the answers in a list format displayed on a web page. The server uses an HTML template engine (e.g. Jinja2) to display the name of each generation AI and its answer side by side on the web page. Specifically, the server displays the following on the web page:

[0222] Generative AI A: "AI will continue to evolve and support people's lives in a variety of areas."

[0223] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[0224] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[0225] Step 5:

[0226] The server compares the answers from each generation AI and visually displays the differences. The input is the received answer data, and the output is the visually displayed difference information. The server uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using color coding or graphs based on the similarity. Specifically, the server calculates the similarity between the answers of generation AI A and generation AI B, and displays them in the same color to indicate high similarity.

[0227] Step 6:

[0228] The user specifies the usage percentage of each answer as input. The input is the usage percentage specified by the user, and the output is the hybrid answer. The server combines each answer based on the percentage and generates the hybrid answer. In concrete terms, the user specifies "answer A will be used 70%, answer B will be used 20%, and answer C will be used 10%", and the server generates the following hybrid answer:

[0229] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[0230] (Application example 2)

[0231] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0232] In conventional content generation systems using generative AI models, there was a lack of a way to effectively compare and integrate responses from multiple generative AIs, making it difficult for users to generate optimal content. In addition, there was also an issue of efficiency when applying the generated content to content distribution services.

[0233] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0234] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to each generation AI, a means for displaying answers from each generation AI in a list format, a means for visually indicating the difference between each answer, a means for generating a hybrid answer by combining each answer in a ratio specified by the user, and a means for applying the generated hybrid answer to a content distribution service. This enables a user to effectively compare and integrate answers from multiple generation AIs to generate optimal content. In addition, the generated content can be quickly and efficiently applied to a content distribution service.

[0235] The "question input interface" is an interface through which users can input questions to the generation AI.

[0236] "Generative AI" is a system that uses artificial intelligence technology to generate text and content.

[0237] An "answer" is the text or content that a generative AI generates in response to a question.

[0238] The "list format" is a format in which multiple answers are displayed side by side on one screen.

[0239] A "means of visually showing the differences" is a means of displaying the differences in the answers from each generation AI in a visual manner, such as color coding or graphs.

[0240] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[0241] A "content distribution service" is a service that provides generated content to users via the Internet.

[0242] A system for implementing this invention has the following configuration. The server has a means for receiving questions from users through a question input interface and sending the questions to each generation AI. Specifically, the user inputs a question using a smartphone or a head-mounted display. For example, the user inputs a prompt such as "Please tell me some ideas for a new blog post."

[0243] When the server receives a question, it sends the question to multiple generation AIs (e.g., GPT-4 (registered trademark), BERT, T5, etc.). Each generation AI generates an answer to the question and sends it back to the server. The server has a means for displaying the answers from each generation AI in a list format. This allows the user to compare the answers from each generation AI at a glance.

[0244] Furthermore, the server has a means for visually indicating the differences between the answers. Specifically, the server calculates the similarity between the answers and displays the differences using colors and graphs. This allows the user to intuitively understand the differences between the answers.

[0245] The user can choose the answer they like from each answer and specify the percentage. For example, they can specify "70% of the answers from GPT-4, 20% of the answers from BERT, and 10% of the answers from T5". The server has a means to combine each answer based on these percentages and generate a hybrid answer. The generated hybrid answer is provided to the user.

[0246] Finally, the server has means for applying the generated hybrid answer to a content delivery service, thereby allowing users to quickly and efficiently deliver the generated content.

[0247] As a concrete example, when a user enters a question such as "Please tell me some ideas for new blog articles," the answers from each generative AI are displayed in a list, and the differences in the answers are visually shown. A hybrid answer is generated based on the ratio specified by the user, and the answer is applied to the content distribution service.

[0248] Example prompt:

[0249] "Can you give me some ideas for new blog posts?"

[0250] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0251] Step 1:

[0252] The user inputs a question using a terminal. Specifically, the user inputs a prompt sentence through a question input interface on a smartphone or a head-mounted display. The input question is then sent to the server.

[0253] Input: The question entered by the user (prompt)

[0254] Output: The question sent to the server.

[0255] Step 2:

[0256] The server sends the received questions to multiple generative AI models. Specifically, it sends questions to generative AI models such as GPT-4, BERT, and T5 through an API. Each generative AI model generates an answer to the question and sends it back to the server.

[0257] Input: The question sent to the server

[0258] Output: The answer from each generative AI model

[0259] Step 3:

[0260] The server displays the answers from each generative AI model in a list format. Specifically, the name of each generative AI model and its answer are displayed side by side on the screen. This allows the user to compare the answers from each generative AI model at a glance.

[0261] Input: The answer from each generative AI model

[0262] Output: Answers displayed in a list format

[0263] Step 4:

[0264] The server visually indicates the differences between answers. Specifically, it calculates the similarity between answers and displays the differences using colors and graphs. This allows users to intuitively understand the differences between answers.

[0265] Input: The answer from each generative AI model

[0266] Output: Visually indicated difference in answers

[0267] Step 5:

[0268] Users can choose their favorite answer from each answer and specify the percentage. Specifically, users can specify the percentages such as "70% of GPT-4's answers, 20% of BERT's answers, and 10% of T5's answers."

[0269] Input: Percentage of user-specified answers

[0270] Output: Specified percentage

[0271] Step 6:

[0272] The server combines the answers based on the specified percentages to generate a hybrid answer, that is, it combines the text of each answer in the specified percentages to generate a new answer.

[0273] Input: Answers from each generative AI model, in user-specified proportions

[0274] Output: Hybrid answer

[0275] Step 7:

[0276] The server applies the generated hybrid answer to a content delivery service, specifically, providing the generated content to users over the Internet.

[0277] Input: Hybrid answer

[0278] Output: Answers applied to content delivery services

[0279] Furthermore, an emotion engine that estimates the emotion of the user may be combined. That is, the identification processing unit 290 may estimate the emotion of the user using the emotion identification model 59, and perform identification processing using the emotion of the user.

[0280] "Example 1"

[0281] As one embodiment of the present invention, a system including a generation AI incorporating an emotion engine is provided. This system asks a question from a user to a plurality of generation AIs, and displays a list of answers from each generation AI. At least one of the generation AIs includes an emotion engine that recognizes the user's emotion. The emotion engine estimates the emotion from the user's tone of voice, facial expression, choice of words, etc. Then, the answer to the question is adjusted based on the estimation result. For example, if the user feels angry, the emotion engine conveys that information to the generation AI, and the generation AI adjusts to generate a more polite answer. Also, if the user feels happy, the generation AI adjusts to generate a more enjoyable answer. In this way, the emotion engine plays a role in optimizing the answer of the generation AI according to the user's emotion.

[0282] "Example 2"

[0283] Furthermore, as another embodiment of the present invention, a system is provided that generates hybrid answers by taking into account the output of an emotion engine. Specifically, when a user specifies the usage ratio of each generation AI's answer, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine conveys that information to the system, and the system generates hybrid answers that include more fun answers. In this way, by taking into account the output of the emotion engine, it is possible to generate hybrid answers that are more adapted to the user's emotions.

[0284] The process flow of each embodiment will be described below.

[0285] "Example 1"

[0286] Step 1: Receive a question from the user.

[0287] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[0288] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[0289] Step 4: Adjust the responses of each generation AI based on the estimation results. For example, if the user feels angry, the emotion engine will convey that information to the generation AI, and the generation AI will adjust to generate more polite responses.

[0290] Step 5: List the answers from each generation AI.

[0291] "Example 2"

[0292] Step 1: Receive a question from the user.

[0293] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[0294] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[0295] Step 4: When the user specifies the proportion of each AI response to be used, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine will convey that information to the system, and the system will generate hybrid responses that include more happy responses.

[0296] Step 5: Provide the hybrid answer to the user.

[0297] Example 1

[0298] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal".

[0299] In conventional generative AI systems, it is difficult to visually recognize the differences when comparing answers from multiple generative AIs. In addition, the system lacks the functionality to adjust the generative AI's answers according to the user's emotions, making it difficult to increase user satisfaction. Furthermore, there is no means to combine answers from multiple generative AIs to generate the optimal answer, making it difficult to provide users with the most optimal information.

[0300] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0301] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the differences, a means for combining each answer based on a ratio specified by the user to generate a hybrid answer, and a means for recognizing the user's emotions and adjusting the answer of the generation AI based on the emotions. This makes it possible to visually compare answers from multiple generation AIs and provide the optimal answer according to the user's emotions.

[0302] A "question input interface" is a user interface for users to input questions, and is implemented using a web browser or a dedicated application.

[0303] "Generative AI" is an artificial intelligence model that uses natural language processing technology to generate answers to input questions, and examples include GPT-4 and BERT.

[0304] The "list format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen, and the name of each generation AI and its answer are arranged so that they can be visually recognized.

[0305] "Visually showing the differences" means displaying the differences in answers from multiple generation AIs using visual means such as color coding or graphs, and calculating the similarity between answers and visually expressing the result.

[0306] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user, providing information optimized to the user's needs.

[0307] The "emotion engine" is an engine that has the ability to infer emotions from the user's tone of voice, facial expressions, choice of words, etc., and adjust the responses of the generative AI based on these inferred results.

[0308] "Recognizing the user's emotions" means using an emotion engine to estimate the user's emotional state and conveying that information to the generative AI.

[0309] This invention is a system that asks the same question to multiple AI generators, displays the answers, visually recognizes the differences, and generates hybrid answers in the ratio specified by the user. It also includes a function to recognize the user's emotions and adjust the AI ​​generator's answers based on those emotions.

[0310] System Configuration

[0311] Question input interface

[0312] A user inputs a question through a question input interface. This interface is implemented using a web browser or a dedicated application. For example, a user inputs "What's the weather like today?"

[0313] Submit a question

[0314] The server sends the question received from the user to multiple generative AIs. Specifically, the server formats the question as a prompt sentence and sends it to a generative AI model (e.g., GPT-4 or BERT). An example of a prompt sentence is "Please provide the weather forecast for today."

[0315] Receiving and listing responses

[0316] The server receives the answers from each AI generator and displays them in a list. The name of each AI generator and its answer are displayed side by side on the terminal screen. For example, it is displayed as follows:

[0317] Generative AI model 1: "The weather is sunny today."

[0318] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[0319] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[0320] Recognizing differences in answers

[0321] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[0322] Hybrid Answer Generation

[0323] The user specifies the usage ratio of each answer. The server combines each answer based on that ratio to generate a hybrid answer. For example, if the user specifies "50% of the answers from Generative AI Model 1, 30% of the answers from Generative AI Model 2, and 20% of the answers from Generative AI Model 3," the server combines these answers to generate a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[0324] Using the Emotion Engine

[0325] The server uses an emotion engine to recognize the user's emotions. The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc. Based on the inference results, the generation AI adjusts the answer. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with."

[0326] Examples

[0327] The user enters "What's the weather like today?" into the question input interface and clicks the send button. The server sends this question to the generation AI as a prompt sentence, "Please provide the weather forecast for today." The server receives the answer from the generation AI and displays it in a list format on the device screen. The server calculates the similarity of each answer and displays them in different colors. The user specifies the usage ratio of each answer, and the server generates a hybrid answer based on that ratio. In addition, an emotion engine is used to recognize the user's emotions and adjust the answer of the generation AI.

[0328] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0329] Step 1:

[0330] A user inputs a question through the question input interface. Specifically, the user inputs "What's the weather like today?" and clicks the send button. The input question is sent to the server. The input data is the question in text format, and the output is the question data sent to the server.

[0331] Step 2:

[0332] The server sends the questions received from the user to multiple generation AIs. Specifically, the server formats the questions as prompt sentences and sends them to a generation AI model (e.g., GPT-4 or BERT). The input data is the question text from the user, and the output is the prompt sentence sent to the generation AI. For example, the server generates a prompt sentence such as "Please provide the weather forecast for today."

[0333] Step 3:

[0334] The server receives the answers from each generation AI. The responses from the generation AI are returned to the server in a data format such as JSON. The input data is the answer data from the generation AI, and the output is the answer data received by the server. For example, GPT-4 returns the answer "Today's weather is sunny."

[0335] Step 4:

[0336] The server displays the received answers in a list format. The name of each AI generator and its answer are displayed side by side on the terminal screen. The input data are the answer data from each AI generator, and the output is the answer displayed on the terminal in a list format. For example, it is displayed as follows:

[0337] Generative AI model 1: "The weather is sunny today."

[0338] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[0339] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[0340] Step 5:

[0341] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. The input data is the answer data from each generation AI, and the output is the visually displayed difference information. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[0342] Step 6:

[0343] The user specifies the usage ratio of each answer. Specifically, the user specifies "50% of answers from AI model 1, 30% of answers from AI model 2, and 20% of answers from AI model 3." The input data is the usage ratio specified by the user, and the output is the ratio data sent to the server.

[0344] Step 7:

[0345] The server combines each answer based on the ratio and generates a hybrid answer. The input data is the usage ratio specified by the user and the answer data from each generation AI, and the output is the generated hybrid answer. For example, it generates a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[0346] Step 8:

[0347] The server recognizes the user's emotions using an emotion engine. The emotion engine estimates emotions from the user's tone of voice, facial expressions, choice of words, etc. The input data is the user's voice and video data, and the output is estimated emotion information.

[0348] Step 9:

[0349] The server adjusts the answer of the generation AI based on the estimation result. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with." The input data is the estimated emotion information and the answer data from the generation AI, and the output is the adjusted answer data.

[0350] (Application example 1)

[0351] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0352] In conventional generative AI systems, when comparing answers from multiple generative AIs, adjustments were not made to take the user's emotions into account, making it difficult to obtain the optimal answer according to the user's emotions. In addition, when combining answers from generative AIs, it was not possible to reflect the user's emotions, making it difficult to increase user satisfaction.

[0353] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0354] In this invention, the server includes a means for asking the same question to multiple generation AIs, a means for displaying a list of answers from each generation AI, a means for recognizing differences between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for recognizing the user's emotions, and a means for adjusting the answer of the generation AI based on the recognized emotions. This makes it possible to provide an optimal answer according to the user's emotions and increase user satisfaction.

[0355] "Generative AI" is a system that uses artificial intelligence technology to automatically generate answers to questions from users.

[0356] A "means for asking a question" is a device or program that has the function of sending questions from users to multiple generation AIs.

[0357] A "means for displaying a list of answers" is a device or program that has the function of displaying answers received from multiple generation AIs side-by-side on a single screen.

[0358] A "means for recognizing differences" is a device or program that has the function of comparing answers from multiple generation AIs and visually showing the differences.

[0359] A "means for regenerating a hybrid answer" is a device or program that has the function of combining answers from multiple generation AIs based on a ratio specified by the user to generate a new answer.

[0360] A "means for recognizing emotions" is a device or program that has the function of inferring emotions from a user's tone of voice, facial expressions, choice of words, etc.

[0361] A "means for adjusting an answer" is a device or program that has the function of optimizing the answer of the generating AI based on the recognized user's emotions.

[0362] The system for implementing this invention has the functions of asking the same question to multiple AI generators, displaying a list of answers from each AI generator, recognizing the differences between each answer, and regenerating a hybrid answer by combining each answer in a ratio specified by the user. It also includes a function of recognizing the user's emotions and adjusting the AI ​​generator's answer based on those emotions.

[0363] System Configuration

[0364] The system consists of the following main components:

[0365] 1. Question input interface: This is an interface for users to input questions. This is implemented on devices such as smartphones and computers.

[0366] 2. Generative AI module: A module that includes multiple generative AIs. Each generative AI automatically generates answers to questions posed by users.

[0367] 3. Answer display module: This module displays a list of answers from each generation AI. It displays the differences in answers in a format that allows visual recognition.

[0368] 4. Emotion recognition module: This module estimates emotions from the user’s tone of voice, facial expressions, choice of words, etc.

[0369] 5. Answer Adjustment Module: This module adjusts the generative AI’s answers based on the recognized emotions.

[0370] 6. Hybrid answer generation module: This module combines answers from each generation AI based on the ratio specified by the user to generate a hybrid answer.

[0371] Program processing explanation

[0372] The server receives questions from users through a question input interface. These questions are sent to the generation AI module, and multiple generation AIs generate their own answers. The generated answers are displayed in a list by the answer display module. At this time, the differences between each answer are displayed in a format that allows visual recognition.

[0373] Next, the emotion recognition module recognizes the user's emotions. For example, it estimates emotions from the user's tone of voice and facial expressions. This recognized emotion information is sent to the answer adjustment module and used to adjust the answer of the generation AI. The adjusted answer is then displayed again in the answer display module.

[0374] Finally, based on the ratio specified by the user, the hybrid answer generation module combines the answers from each generation AI to generate a hybrid answer, which is finally provided to the user.

[0375] Examples

[0376] For example, if a user asks, "What movie would you recommend to watch tonight?", each generative AI will suggest its own movie recommendation. If the emotion recognition module recognizes the user's emotions and determines that the user is tired, for example, the answer will be adjusted to suggest a relaxing movie. Finally, a hybrid recommendation list is generated that combines answers from multiple generative AIs based on the ratio specified by the user.

[0377] Example prompt:

[0378] What movie would you recommend to watch tonight?

[0379] In this way, it is possible to provide the user with an optimal answer according to his / her feelings, thereby increasing the user's satisfaction.

[0380] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0381] Step 1:

[0382] A user inputs a question into the question input interface using a terminal. The input question is transmitted to the server. The input data is a question about the content that the user wants to view.

[0383] Step 2:

[0384] The server sends the received questions to multiple generation AIs. Each generation AI automatically generates an answer to the question. The generated answer is sent back to the server. The input data is the user's question, and the output data is the answer from each generation AI.

[0385] Step 3:

[0386] The server sends the answers from each AI generator to the answer display module and displays them in a list. The answer display module displays the differences between the answers in a format that allows the differences to be visually recognized. The input data are the answers from each AI generator, and the output data is the visually displayed list of answers.

[0387] Step 4:

[0388] When a user inputs a question into the question input interface using a terminal, the emotion recognition module analyzes the user's tone of voice and facial expression to recognize the emotion. The input data is the user's tone of voice and facial expression, and the output data is the recognized emotion information.

[0389] Step 5:

[0390] The server sends the recognized emotion information to the answer adjustment module. The answer adjustment module adjusts the answer of each generation AI based on the emotion information. For example, if the user is recognized as tired, the answer is adjusted to suggest relaxing content. The input data are the recognized emotion information and the answer from each generation AI, and the output data are the adjusted answer.

[0391] Step 6:

[0392] The server again transmits the adjusted answers to the answer display module for displaying them in a list. The input data is the adjusted answers, and the output data is a visually displayed list of adjusted answers.

[0393] Step 7:

[0394] The user specifies the usage ratio of each answer using the terminal. For example, the user specifies "50% of the answers from AI1, 30% of the answers from AI2, and 20% of the answers from AI3." The input data is the usage ratio specified by the user.

[0395] Step 8:

[0396] The server generates a hybrid answer by combining the answers from each generation AI in the hybrid answer generation module based on the ratio specified by the user. The input data is the usage ratio specified by the user and the answers from each generation AI, and the output data is the hybrid answer.

[0397] Step 9:

[0398] The server provides the generated hybrid answer to the user. The user uses a terminal to confirm the hybrid answer. The input data is the hybrid answer, and the output data is the final answer provided to the user.

[0399] Example 2

[0400] Next, a second embodiment of the second embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0401] In conventional generative AI systems, it was difficult to compare answers from multiple generative AIs and combine them to generate the optimal answer. In addition, the system did not generate answers that took the user's feelings into account, so it was not possible to increase user satisfaction. This meant that users could not efficiently obtain the information they were looking for, and the system's value as a user decreased.

[0402] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0403] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the difference, a means for combining answers from each generation AI in a ratio specified by the user to generate a hybrid answer, and a means for adjusting the hybrid answer by taking into account the output of the emotion analysis engine. This allows the user to compare answers from multiple generation AIs and obtain the optimal answer. Furthermore, the generation of an answer that takes into account the user's emotions can increase the user's satisfaction.

[0404] A "question input interface" is a user interface that allows a user to input a question and send the question to the system.

[0405] "Generative AI" is an artificial intelligence model that generates answers in natural language in response to a given question.

[0406] The "list format" is a format in which a plurality of items are visually organized and displayed, making it easier for the user to compare the items.

[0407] "Visually showing" means displaying information using visual means such as color coding or graphs so that the user can understand it intuitively.

[0408] A "hybrid answer" is a new answer generated by combining answers from multiple generative AIs, incorporating the characteristics of each answer.

[0409] An "emotion analysis engine" is an engine that analyzes emotions from user input and actions and provides the results to the system.

[0410] A "weighted average" is a method of calculating an average by assigning different weights to each piece of data, and is used to emphasize the influence of certain data.

[0411] An "API Request" is a request made to another system or service through an Application Program Interface, which is a means of sending or receiving data.

[0412] The present invention is a system for generating optimal answers to questions from users using multiple generative AI models. A specific embodiment of this system will be described below.

[0413] The server receives questions from the user through a question input interface. This interface is implemented using a web browser or a dedicated application. Specifically, a browser such as GOOGLE CHROME (registered trademark) can be used. The user inputs a question such as, "Tell me about the future of AI technology."

[0414] The server sends the received questions to multiple generation AIs. OpenAI's GPT-3 (registered trademark) and Google's BERT are used as generation AIs. Each generation AI generates an answer to the question and sends it back to the server. Specifically, the question is sent to each generation AI's endpoint using an HTTP POST request.

[0415] The server receives the answers from each AI and displays them in a list format. This display is implemented on a web page using HTML and CSS. For example, the name of each AI and its answer are displayed side by side on the screen in the form of "Answer from AI model A" and "Answer from AI model B."

[0416] The server compares the answers from each generation AI and visually displays the differences. Specifically, it uses the Python library scikit-learn to calculate the similarity between answers, and displays the differences using color coding and graphs based on the similarity. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red.

[0417] Users can specify the proportion of each answer to be used. The server receives input from the user and combines each answer based on that proportion to generate a hybrid answer. Specifically, it uses Python's NumPy library to weight and average each answer by the specified proportion to generate a new answer. For example, if a user specifies "70% of answers from Generative AI Model A and 30% from Generative AI Model B," each answer will be combined according to that proportion.

[0418] Furthermore, by using a sentiment analysis engine, it is possible to generate hybrid answers that are adapted to the user's emotions. As a sentiment analysis engine, IBM's Watson (registered trademark) Tone Analyzer or the like is used. The sentiment analysis engine analyzes the user's input and actions to determine the user's emotions, and sends the information to the server. For example, if the sentiment analysis engine determines that the user is happy, the server adjusts the answer to include more happy content.

[0419] In this way, users can compare answers from multiple AI generators to obtain the best answer. In addition, by generating answers that take the user's emotions into account, user satisfaction can be increased.

[0420] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0421] Step 1:

[0422] The user enters a question.

[0423] A user accesses the question input interface through a web browser or a dedicated application and inputs a question. For example, a user accesses a web page using Google Chrome and inputs "Please tell me about the future of AI technology" in the text box. The input question is sent to the server.

[0424] Step 2:

[0425] The server receives the question and sends it to each generating AI.

[0426] The server sends the question received from the user to multiple generation AIs. Specifically, it uses an HTTP POST request to send the question to the endpoint of each generation AI. OpenAI's GPT-3 and Google's BERT are used as generation AIs. The input is the user's question, and the output is the question sent to each generation AI.

[0427] Step 3:

[0428] Each generation AI generates an answer to the question and sends it back to the server.

[0429] Each generation AI generates an answer to the question it receives. For example, GPT-3 generates the answer "AI technology will continue to evolve in the future and will be used in many fields such as medicine and education," and BERT generates the answer "AI technology will contribute to the realization of autonomous driving and smart cities." These answers are sent back to the server as API responses. The input is the question sent to each generation AI, and the output is the generated answer.

[0430] Step 4:

[0431] The server displays a list of answers from each generation AI.

[0432] The server uses HTML and CSS to display the answers received from each generation AI in a list format on a web page. For example, the name of each generation AI and its answer are displayed side by side on the screen in the form of "Answer from generation AI model A" and "Answer from generation AI model B." The input is the answer from each generation AI, and the output is the answer displayed in a list format.

[0433] Step 5:

[0434] The server visually indicates the difference between each answer.

[0435] The server uses Python's scikit-learn library to calculate the similarity between each answer. Based on the calculation results, the similarity of the answers is visually displayed using color coding or graphs. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red. The input is the answer from each generation AI, and the output is the difference between the visually displayed answers.

[0436] Step 6:

[0437] The user specifies the usage ratio of each answer.

[0438] The user accesses an interface to specify the usage ratio of each answer. For example, the user can use a slider or text box to input "70% of answers from AI model A and 30% of answers from AI model B." The input is the usage ratio specified by the user, and the output is the usage ratio sent to the server.

[0439] Step 7:

[0440] The server generates a hybrid answer based on the specified percentages.

[0441] The server combines each answer based on the ratio specified by the user. Specifically, it uses Python's NumPy library to weight and average each answer to generate a new hybrid answer. For example, it generates an answer such as "AI technology will continue to evolve in the future, contributing to the realization of healthcare, education, autonomous driving, and smart cities." The input is the usage ratio specified by the user and the answer from each generation AI, and the output is the generated hybrid answer.

[0442] Step 8:

[0443] The emotion analysis engine analyzes the user's emotions and transmits them to the server.

[0444] Sentiment analysis engines analyze user input and behavior to determine emotions. For example, IBM's Watson Tone Analyzer can be used to analyze emotions from a user's text input and determine that "the user is happy." This information is sent to a server. The input is the user's input and behavior, and the output is the analyzed emotional information.

[0445] Step 9:

[0446] The server adjusts the hybrid answer taking into account the output of the sentiment analysis engine.

[0447] The server adjusts the hybrid answer taking into account the output from the sentiment analysis engine. For example, if it is determined that the user is happy, the server adjusts the answer to include more fun content. Specifically, it generates an answer that emphasizes the fun element. The input is the emotion information from the sentiment analysis engine and the hybrid answer, and the output is the adjusted hybrid answer.

[0448] (Application example 2)

[0449] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0450] In systems using conventional generative AI models, when comparing answers from multiple generative AI models, it was difficult to visually recognize the differences in answers, and it was not possible to generate hybrid answers that adapted to the user's emotions. As a result, it was necessary for users to make a great deal of effort to obtain the optimal answer, making it difficult to obtain information efficiently. Furthermore, since it was not possible to take the user's emotions into account when combining answers from generative AI models, it was not possible to increase user satisfaction.

[0451] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0452] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for regenerating a hybrid answer by combining each answer in a ratio specified by a user, a means for generating a hybrid answer taking into account the output of the emotion engine, and a means for calculating the similarity between answers and visually displaying the difference. This allows a user to efficiently compare answers from a plurality of generative AI models and visually recognize the difference. In addition, by generating a hybrid answer that takes into account the user's emotions, the user's satisfaction can be increased.

[0453] A "generative AI model" is an algorithm or system that uses artificial intelligence to generate content such as text and images.

[0454] A "means for asking questions" is an interface that has the ability to receive questions from users and send them to a generative AI model.

[0455] A "means for displaying answers in a list" is an interface with the function of displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[0456] A "means of recognizing differences" is a system that has the ability to compare answers from multiple generative AI models and visually show the differences.

[0457] A "means for regenerating hybrid answers" is a system that has the ability to generate new answers by combining answers from multiple generative AI models based on a ratio specified by the user.

[0458] An "emotion engine" is an algorithm or system that analyzes user emotions and adjusts the system's operation based on the results.

[0459] "Means for calculating similarity" is a system capable of numerically evaluating the similarity between answers from multiple generative AI models.

[0460] A "means for visually displaying differences" is a system that has the ability to visually display the differences between responses using color coding, graphs, or other means.

[0461] A system for implementing the present invention includes a server, a user terminal, and a generative AI model. The server has a function of receiving a question from a user and transmitting the question to a plurality of generative AI models. The user terminal provides a question input interface, allowing the user to input a question. The generative AI model generates an answer to the received question and transmits it back to the server.

[0462] The server receives answers from multiple generative AI models and displays them in a list. Specifically, the name of each generative AI model and its answer are displayed side by side. In addition, the server recognizes the differences between each answer and calculates the similarity between the answers. Based on the similarity, the differences between the answers are visually displayed using color coding and graphs.

[0463] The user can check the answers from each generative AI model and specify the proportion of each answer to be used. The server combines each answer based on the proportion specified by the user to generate a hybrid answer. At this time, the emotion engine analyzes the user's emotions and adjusts the hybrid answer based on the results. For example, if the emotion engine determines that the user is happy, it will adjust the hybrid answer to include more cheerful answers.

[0464] As a concrete example, if a user enters the question "What is the best travel destination for our next vacation?", the server sends this question to multiple generative AI models. The answers from each generative AI model are displayed in a list, and the differences between the answers are visually shown. The user specifies the proportion of each answer to be used, and a hybrid answer is generated based on the analysis results of the emotion engine.

[0465] The hardware used includes user devices (e.g., smartphones, smart glasses, head-mounted displays, etc.) and servers. The software used includes generative AI models (e.g., OpenAI APIs), emotion engines (e.g., TextBlob), and data visualization tools (e.g., Matplotlib).

[0466] The following are examples of prompt sentences:

[0467] "What's your recommended destination for our next vacation?"

[0468] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0469] Step 1:

[0470] The user uses the terminal to input a question.

[0471] Input: The prompt text entered by the user (e.g. "What are some recommended travel destinations for our next vacation?")

[0472] Output: The entered question is sent to the server.

[0473] Specific operation: The user inputs a question using the question input interface of the terminal and presses the send button. The terminal sends the input question to the server.

[0474] Step 2:

[0475] The server receives the question and sends it to multiple generative AI models.

[0476] Input: The question submitted by the user

[0477] Output: The question sent to the generative AI model

[0478] How it works: The server sends the received question to multiple generative AI models, each of which receives the question and prepares to generate an answer.

[0479] Step 3:

[0480] The generative AI model generates answers to the questions and sends them back to the server.

[0481] Input: The question sent to the generative AI model

[0482] Output: The answer returned by the generative AI model.

[0483] Specific operation: Each generative AI model generates an answer based on the question and sends the answer back to the server. The server receives the answer from each generative AI model.

[0484] Step 4:

[0485] The server lists the answers from each generative AI model.

[0486] Input: The answer returned by the generative AI model

[0487] Output: List of answers

[0488] Specific operation: The server displays the answers from each generative AI model in a list format on the screen. The name of each generative AI model and its answer are displayed side by side.

[0489] Step 5:

[0490] The server recognizes the difference between each answer and displays it visually.

[0491] Input: Answer from a generative AI model

[0492] Output: Visually displayed difference in answers

[0493] Specific operation: The server calculates the similarity of each answer and visually displays the differences between the answers using color coding and graphs, allowing users to recognize the differences between each answer at a glance.

[0494] Step 6:

[0495] The user specifies the usage ratio of each answer.

[0496] Input: The percentage of each answer that the user specifies to be used (e.g. "70% of the time answer A, 20% of the time answer B, 10% of the time answer C")

[0497] Output: The specified usage percentage is sent to the server.

[0498] Specific operation: The user uses the terminal interface to specify the usage ratio of each answer and presses the send button. The terminal then sends the specified usage ratio to the server.

[0499] Step 7:

[0500] The server generates a hybrid answer based on the specified percentages.

[0501] Input: Usage rate of each answer specified by the user, answers from the generative AI model

[0502] Output: Hybrid answer

[0503] Specific behavior: The server combines each answer based on the user-specified percentage to generate a hybrid answer, which is then displayed to the user.

[0504] Step 8:

[0505] The server uses an emotion engine to analyze the user's emotions and tailor the hybrid response.

[0506] Input: Hybrid answers, user sentiment data

[0507] Output: A tailored hybrid answer

[0508] Specific operation: The server uses the emotion engine to analyze the user's emotions and adjusts the hybrid answers based on the results. For example, if the emotion engine determines that the user is happy, the answers will be adjusted to include more happy answers. The adjusted hybrid answers are finally displayed to the user.

[0509] 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 a voice indicating a user input for the result of the specific processing. The control unit 46A transmits the voice data indicating the 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 voice data.

[0510] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0511] Another example of generative AI is Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) are mentioned.

[0512] In the above embodiment, an example was given in which the specific process was performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0513] [Second embodiment]

[0514] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0515] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0516] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[0517] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, 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. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0518] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[0519] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[0520] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[0521] Fig. 4 shows an example of main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0522] The specific processing program 56 is an example of a "program" according to the technology of the present 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 in accordance with the specific processing program 56 executed on the RAM 30.

[0523] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0524] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0525] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0526] "Example 1"

[0527] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[0528] "Example 2"

[0529] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[0530] The process flow of each embodiment will be described below.

[0531] "Example 1"

[0532] Step 1: The user inputs a question through the question input interface. This question is then posed to multiple generation AIs.

[0533] Step 2: The system sends a question to each Generator AI, which answers the question using its own algorithm.

[0534] Step 3: The system receives answers from each AI generator. These answers are displayed in a list format. Specifically, the name of each AI generator and its answer are displayed side by side on the screen.

[0535] Step 4: The system compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using color coding and graphs based on the similarity.

[0536] Step 5: The user specifies the percentage of each answer to be used. For example, "Answer A will be 70%, Answer B will be 20%, and Answer C will be 10%."

[0537] Step 6: The system combines the answers based on the percentages specified by the user to generate a hybrid answer, which becomes the final output.

[0538] Example 1

[0539] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0540] Conventional generative AI systems rely on answers from a single generative AI, which limits the variety and accuracy of answers. In addition, there is no way to compare answers from multiple generative AIs, making it difficult for users to select the most appropriate answer. Furthermore, there is a lack of functionality to combine multiple answers to generate hybrid answers, which means that it is not possible to flexibly generate answers that meet the needs of users.

[0541] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0542] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually showing the differences, and a means for regenerating a hybrid answer by combining the answers from each generation AI based on a ratio specified by the user. This makes it possible to compare various answers from multiple generation AIs and select the optimal answer. It also makes it possible to flexibly generate hybrid answers according to the needs of the user.

[0543] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[0544] "Generative AI" is an artificial intelligence model that generates natural language responses based on a given prompt.

[0545] An "API endpoint" is an interface for sending questions to the generation AI and receiving answers.

[0546] "List format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen.

[0547] "Visually showing" means showing the differences in responses using visual means such as color coding or graphs.

[0548] "Similarity" is an index showing the degree of agreement between the contents of multiple answers.

[0549] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user.

[0550] The "proportion" is the ratio that indicates the extent to which the user will use the answers from each generation AI.

[0551] This invention is a system that asks the same question to multiple generative AI models and compares and integrates their answers. A specific embodiment of this system will be described below.

[0552] Question input interface

[0553] The user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and provides a form where the user can enter a question in text format. Once the user enters a question and clicks the "Submit" button, the question is sent to the server.

[0554] Submit a question

[0555] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt. The generative AI models used include, for example, natural language processing models and machine learning models.

[0556] Receiving and listing responses

[0557] The server receives the answers from each generative AI model and temporarily stores them. Next, the server generates a web page to display the received answers in a list format. This web page displays the name of each generative AI model and its answer side by side. The user can view this page in the web browser of the terminal.

[0558] Recognizing differences in answers

[0559] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity. This process is carried out using text analysis algorithms and data visualization tools.

[0560] Regenerating a hybrid answer

[0561] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[0562] Examples

[0563] Example of question input

[0564] The user enters the following question into the question entry interface:

[0565] "What do you think the future of AI is?"

[0566] Examples of prompt statements

[0567] Here is an example of a prompt to be input to the generative AI model:

[0568] "What do you think the future of AI is?"

[0569] Example of answer list display

[0570] Generative AI Model 1: “AI will play an important role in many fields in the future, including medicine and education.”

[0571] Generative AI Model 2: "AI will help make autonomous driving and smart cities a reality."

[0572] Generative AI Model 3: “AI will continue to evolve as a tool to make human life more convenient.”

[0573] Examples of visual display of response differences

[0574] The parts of the answers from Generative AI Model 1 and Generative AI Model 2 that are highly similar are displayed in green.

[0575] The areas where the answers from Generative AI Model 1 and Generative AI Model 3 have low similarity are displayed in red.

[0576] A concrete example of hybrid answer regeneration

[0577] The user specifies "70% of the answers from Generative AI Model 1, 20% from Generative AI Model 2, and 10% from Generative AI Model 3." The server combines each answer based on that ratio and generates a hybrid answer like this:

[0578] "Going forward, AI will play an important role in many fields, including healthcare and education, and will also contribute to the realization of autonomous driving and smart cities, continuing to evolve as a tool to make human life more convenient."

[0579] In this way, users can compare various answers from multiple generative AI models and select the optimal answer. It also enables flexible generation of hybrid answers according to the needs of users.

[0580] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0581] Step 1: Enter your question

[0582] The user opens a web browser on the terminal and accesses the question input interface. The user enters a question in the text box and clicks the "Send" button. This operation sends the question to the server.

[0583] Input: A question typed in by the user (e.g., "Tell me about the future of AI.")

[0584] Output: The question data sent to the server

[0585] Step 2: Submit your question

[0586] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt.

[0587] Input: The query data sent to the server

[0588] Output: A prompt sent to each generative AI model (e.g., "Tell me about the future of AI.")

[0589] Step 3: Receiving a response

[0590] The server receives answers from each generative AI model. Each generative AI model generates an answer based on the prompt sentence and returns it to the server. The server temporarily stores these answers.

[0591] Input: Answer data from each generative AI model

[0592] Output: Answer data stored on the server (e.g., "AI will play an important role in many fields, such as medicine and education, in the future.")

[0593] Step 4: List the responses

[0594] The server generates a web page to display the answers of each generative AI model in a list format. The user views this page in a web browser on their device.

[0595] Input: Answer data saved on the server

[0596] Output: A web page displayed in list format (Example: Generative AI model 1: "AI will play an important role in many fields in the future, such as medicine and education.")

[0597] Step 5: Recognize differences in responses

[0598] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity.

[0599] Input: Answer data saved on the server

[0600] Output: A web page with visual indication of the differences (e.g. high similarity is shown in green, low similarity in red)

[0601] Step 6: Regenerate the hybrid answer

[0602] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[0603] Input: The percentage of each answer specified by the user (e.g. 70% of the answers from AI model 1, 20% of the answers from AI model 2, and 10% of the answers from AI model 3)

[0604] Output: A web page with a hybrid answer (e.g., "AI will play an important role in many fields, such as medicine and education, and will continue to evolve as a tool to make human life more convenient, contributing to the realization of autonomous driving and smart cities.")

[0605] (Application example 1)

[0606] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0607] In systems using conventional generative AI models, there is a lack of a means to compare answers from multiple generative AI models and obtain the optimal answer. In addition, it is difficult to visually recognize the differences between answers from each generative AI model, and there is insufficient support for users to select the optimal answer. Furthermore, when combining answers from multiple generative AI models to generate a hybrid answer, a means is required that can flexibly respond based on the ratio specified by the user.

[0608] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0609] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for recognizing the difference between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for calculating and visually displaying the similarity between each answer, and a means for the user to specify the ratio for generating the hybrid answer. This makes it possible to compare answers from a plurality of generative AI models and obtain an optimal answer. In addition, by visually recognizing the difference between each answer, support for the user to select the optimal answer is enhanced. Furthermore, it becomes possible to flexibly generate a hybrid answer based on the ratio specified by the user.

[0610] A "generative AI model" is a type of artificial intelligence that uses natural language processing to generate answers to user questions or prompts.

[0611] The "means of asking questions" is a function that allows users to send questions to multiple generative AI models.

[0612] "Means of displaying answers in a list" is a function for displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[0613] A "means of recognizing differences" is a function that can compare answers from multiple generative AI models and identify the differences between them.

[0614] A "means for regenerating hybrid answers" is a function for generating new answers by combining answers from multiple generative AI models in a ratio specified by the user.

[0615] "Means for calculating similarity" is a function for numerically evaluating the similarity between answers from multiple generative AI models.

[0616] "Means for visual display" refers to a function for displaying the calculated similarity or difference in answers in a visual format such as a graph or color coding.

[0617] "Means for specifying proportions" refers to a function that enables users to specify the proportions in which answers from each generative AI model should be combined.

[0618] The system for implementing the present invention operates in cooperation with a server, a terminal, and a user. A specific embodiment of the system will be described below.

[0619] System Configuration

[0620] server

[0621] The server includes means for asking the same question to multiple generative AI models, means for displaying a list of answers from each generative AI model, means for recognizing differences between each answer, means for combining each answer in a ratio specified by a user to regenerate a hybrid answer, means for calculating and visually displaying the similarity of each answer, and means for a user to specify the ratio for generating a hybrid answer.

[0622] The server uses the following software and hardware:

[0623] Generative AI Models: OpenAI API

[0624] Data processing: Python program

[0625] Visualization: Matplotlib, Scikit-learn

[0626] Terminal

[0627] The terminal provides an interface for the user to enter questions and view generated answers. The terminal may be a smartphone, smart glasses, a head-mounted display, or the like.

[0628] User

[0629] Users input questions through their devices, review answers from generative AI models, and specify the proportions for generating hybrid answers.

[0630] Processing flow

[0631] 1. Entering a question: The user enters a question through the terminal, for example, a prompt such as "Please tell me some ideas for a new blog post."

[0632] 2. Question submission: The server submits the input question to multiple generative AI models.

[0633] 3. Obtaining answers: The server receives answers from each generative AI model.

[0634] 4. Displaying a list of answers: The server displays the answers from each generative AI model in a list format on the terminal.

[0635] 5. Difference Recognition: The server calculates the similarity of each answer and displays it visually, for example using a heatmap to show the similarity between answers.

[0636] 6. Specify Proportion: The user specifies what proportion of each answer should be combined.

[0637] 7. Generate hybrid answer: The server generates a hybrid answer based on the specified percentages and displays it on the terminal.

[0638] Examples

[0639] When a user types a question like "What are some ideas for a new blog post?", the server sends the question to multiple generative AI models, which can produce answers like these:

[0640] Generative AI Model A: "Latest trends in AI technology"

[0641] Generative AI Model B: "Advantages and disadvantages of remote work"

[0642] Generative AI Model C: "Sustainable Energy Future"

[0643] The server lists these answers and visually displays the similarity between the answers in a heat map. For example, the user can specify that "Latest trends in AI technology" should be combined at 70%, and the other two at 20% and 10%, respectively. The server generates a hybrid answer based on this specification and provides it to the user.

[0644] In this way, users can compare answers from multiple generative AI models to obtain the best answer. Also, visual recognition of the differences between each answer provides enhanced support for selecting the more appropriate answer. Furthermore, it is possible to flexibly generate hybrid answers based on the ratio specified by the user.

[0645] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0646] Step 1:

[0647] The user inputs a question through a terminal.

[0648] Input: The prompt the user enters (e.g. "Can you give me some ideas for new blog posts?")

[0649] Output: The prompt text entered

[0650] Specific operations: The user inputs a question using the input interface of the terminal and presses the send button.

[0651] Step 2:

[0652] The server sends the input question to multiple generative AI models.

[0653] Input: The prompt text entered by the user.

[0654] Output: The prompt sent to each generative AI model

[0655] Specific operation: The server uses the API to send prompt sentences to each generative AI model.

[0656] Step 3:

[0657] The server receives the answers from each generative AI model.

[0658] Input: The prompt sent to each generative AI model

[0659] Output: The answer from each generative AI model

[0660] Specific operation: The server receives the API response from each generative AI model and obtains the answer.

[0661] Step 4:

[0662] The server displays the answers from each generative AI model in a list format on the terminal.

[0663] Input: The answer from each generative AI model

[0664] Output: A list of answers displayed on the terminal

[0665] Specific operation: The server compiles the answers from each generative AI model into a single list and sends it to the terminal's display interface.

[0666] Step 5:

[0667] The server calculates the similarity of each answer and displays it visually.

[0668] Input: The answer from each generative AI model

[0669] Output: A heatmap showing similarity

[0670] Specific operation: The server calculates the vector of each answer, calculates the similarity using cosine similarity, and generates a heat map of the result and displays it on the terminal.

[0671] Step 6:

[0672] The user specifies how much of each answer should be combined.

[0673] Input: The percentage of each answer that you specify (e.g. 70%, 20%, 10%)

[0674] Output: Specified percentage

[0675] Specific behavior: The user uses the device interface to specify the percentage of each answer using sliders and input fields.

[0676] Step 7:

[0677] The server generates a hybrid answer based on the specified percentage and displays it on the terminal.

[0678] Input: Answers from each generative AI model, in user-specified proportions

[0679] Output: Hybrid answer

[0680] Specific operation: The server combines each answer based on the specified ratio to generate a new hybrid answer, and sends the result to the device for display.

[0681] Example 2

[0682] Next, a description will be given of Example 2 of the embodiment 2. In the following description, the data processing device 12 is referred to as a "server" and the smart glasses 214 are referred to as a "terminal".

[0683] Conventional generative AI systems lacked the means to compare answers from multiple generative AIs and effectively integrate them. As a result, users had to check the answers from each generative AI individually and compare and integrate them manually, which was inefficient. In addition, there was no way to visually recognize the differences in the answers from each generative AI, making it difficult to intuitively understand the differences in the quality and content of the answers.

[0684] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0685] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually indicating the differences, and a means for combining the answers from each generation AI in a ratio designated by the user to regenerate a hybrid answer. This enables the user to efficiently compare and integrate answers from multiple generation AIs and obtain an optimal hybrid answer while visually recognizing the differences.

[0686] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[0687] "Generative AI" is an artificial intelligence model that generates natural language answers based on input prompts.

[0688] A "list format" is a format in which multiple items are displayed side by side on one screen or page.

[0689] "Visually showing" means displaying information using visual means such as color coding or graphs.

[0690] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[0691] The "proportion specified by the user" is a number that specifies the proportion in which the user will combine answers from each generation AI.

[0692] A "server" is a computer system that performs processes such as receiving questions, sending questions to the generation AI, receiving answers, displaying the answers, comparing answers, and generating hybrid answers.

[0693] This invention is a system that asks the same question to multiple generation AIs and compares and integrates their answers. A specific embodiment of this system will be described below.

[0694] Question input interface

[0695] A user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and includes a text box for the user to enter a question and a submit button. When the user enters a question and clicks the submit button, the question is sent to the server.

[0696] Submit a question

[0697] The server sends the questions received from the user to multiple generation AIs. The server uses a Python HTTP request library (e.g., requests) to send questions to the API endpoints of each generation AI. Specifically, it sends the following prompt text to the generation AI:

[0698] "What do you think the future of AI is?"

[0699] Receiving a response

[0700] The server receives answers from each AI generator. It waits for a response from the API of each AI generator and stores the received answers. For example, it receives an answer from AI generator A saying, "AI will continue to evolve in the future and support people's lives in various fields."

[0701] List of responses

[0702] The server displays the answers received from each AI generator in a list format. Using an HTML template engine (e.g. Jinja2), the name of each AI generator and its answer are displayed side by side on a web page. For example, it is displayed as follows:

[0703] Generative AI A: "AI will continue to evolve and support people's lives in a variety of areas."

[0704] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[0705] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[0706] Recognizing differences in answers

[0707] The server compares the answers from each generation AI and visually displays the differences. It uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using colors and graphs based on the similarity. For example, if the answers of generation AI A and generation AI B show a high similarity, they are displayed in the same color.

[0708] Regenerating a hybrid answer

[0709] The user specifies the percentage of each answer that should be used as input. The server combines the answers based on that percentage to generate a hybrid answer. For example, if the user specifies "70% of the answers should be from answer A, 20% from answer B, and 10% from answer C," the server will generate the following hybrid answer:

[0710] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[0711] In this way, users can efficiently compare and integrate answers from multiple generative AIs, visually recognizing the differences, and arrive at the optimal hybrid answer.

[0712] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0713] Step 1:

[0714] The user opens a web browser on the device and accesses the question input interface. The user enters a question in the text box and clicks the send button. The input is the question entered by the user, and the output is the question data sent to the server. In concrete terms, the user enters "Tell me about the future of AI" and clicks the send button.

[0715] Step 2:

[0716] The server sends questions received from the user to multiple generation AIs. The input is the question data received from the user, and the output is the prompt text sent to each generation AI. The server uses Python's HTTP request library (e.g., requests) to send questions to the API endpoint of each generation AI. Specifically, the server sends the question "Tell me about the future of AI" to generation AI A, generation AI B, and generation AI C.

[0717] Step 3:

[0718] The server receives the answers from each generation AI. The input is the API response from each generation AI, and the output is the received answer data. The server waits for a response from the API of each generation AI and stores the received answers. In concrete terms, the server receives the answer from generation AI A that "AI will continue to evolve in the future and support people's lives in various fields."

[0719] Step 4:

[0720] The server displays the answers received from each generation AI in a list format. The input is the received answer data, and the output is the answers in a list format displayed on a web page. The server uses an HTML template engine (e.g. Jinja2) to display the name of each generation AI and its answer side by side on the web page. Specifically, the server displays the following on the web page:

[0721] Generative AI A: "AI will continue to evolve and support people's lives in various fields."

[0722] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[0723] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[0724] Step 5:

[0725] The server compares the answers from each generation AI and visually displays the differences. The input is the received answer data, and the output is the visually displayed difference information. The server uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using color coding or graphs based on the similarity. Specifically, the server calculates the similarity between the answers of generation AI A and generation AI B, and displays them in the same color to indicate high similarity.

[0726] Step 6:

[0727] The user specifies the usage percentage of each answer as input. The input is the usage percentage specified by the user, and the output is the hybrid answer. The server combines each answer based on the percentage and generates the hybrid answer. In concrete terms, the user specifies "answer A will be used 70%, answer B will be used 20%, and answer C will be used 10%", and the server generates the following hybrid answer:

[0728] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[0729] (Application example 2)

[0730] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the smart glasses 214 will be referred to as a "terminal".

[0731] In conventional content generation systems using generative AI models, there was a lack of a way to effectively compare and integrate responses from multiple generative AIs, making it difficult for users to generate optimal content. In addition, there was also an issue of efficiency when applying the generated content to content distribution services.

[0732] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0733] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to each generation AI, a means for displaying answers from each generation AI in a list format, a means for visually indicating the difference between each answer, a means for generating a hybrid answer by combining each answer in a ratio specified by the user, and a means for applying the generated hybrid answer to a content distribution service. This enables a user to effectively compare and integrate answers from multiple generation AIs to generate optimal content. In addition, the generated content can be quickly and efficiently applied to a content distribution service.

[0734] The "question input interface" is an interface through which users can input questions to the generation AI.

[0735] "Generative AI" is a system that uses artificial intelligence technology to generate text and content.

[0736] An "answer" is the text or content that a generative AI generates in response to a question.

[0737] The "list format" is a format in which multiple answers are displayed side by side on one screen.

[0738] A "means of visually showing the differences" is a means of displaying the differences in the answers from each generation AI in a visual manner, such as color coding or graphs.

[0739] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[0740] A "content distribution service" is a service that provides generated content to users via the Internet.

[0741] A system for implementing this invention has the following configuration. The server has a means for receiving questions from users through a question input interface and sending the questions to each generation AI. Specifically, the user inputs a question using a smartphone or a head-mounted display. For example, the user inputs a prompt such as "Please tell me some ideas for a new blog post."

[0742] When the server receives a question, it sends the question to multiple generation AIs (e.g., GPT-4, BERT, T5, etc.). Each generation AI generates an answer to the question and sends it back to the server. The server has a means to display the answers from each generation AI in a list format. This allows the user to compare the answers from each generation AI at a glance.

[0743] Furthermore, the server has a means for visually indicating the differences between the answers. Specifically, the server calculates the similarity between the answers and displays the differences using colors and graphs. This allows the user to intuitively understand the differences between the answers.

[0744] The user can choose the answer they like from each answer and specify the percentage. For example, they can specify "70% of the answers from GPT-4, 20% of the answers from BERT, and 10% of the answers from T5". The server has a means to combine each answer based on these percentages and generate a hybrid answer. The generated hybrid answer is provided to the user.

[0745] Finally, the server has means for applying the generated hybrid answer to a content delivery service, thereby allowing users to quickly and efficiently deliver the generated content.

[0746] As a concrete example, when a user enters a question such as "Please tell me some ideas for new blog articles," the answers from each generative AI are displayed in a list, and the differences in the answers are visually shown. A hybrid answer is generated based on the ratio specified by the user, and the answer is applied to the content distribution service.

[0747] Example prompt:

[0748] "Can you give me some ideas for new blog posts?"

[0749] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0750] Step 1:

[0751] The user inputs a question using a terminal. Specifically, the user inputs a prompt sentence through a question input interface on a smartphone or a head-mounted display. The input question is then sent to the server.

[0752] Input: The question entered by the user (prompt)

[0753] Output: The question sent to the server.

[0754] Step 2:

[0755] The server sends the received questions to multiple generative AI models. Specifically, it sends questions to generative AI models such as GPT-4, BERT, and T5 through an API. Each generative AI model generates an answer to the question and sends it back to the server.

[0756] Input: The question sent to the server

[0757] Output: The answer from each generative AI model

[0758] Step 3:

[0759] The server displays the answers from each generative AI model in a list format. Specifically, the name of each generative AI model and its answer are displayed side by side on the screen. This allows the user to compare the answers from each generative AI model at a glance.

[0760] Input: The answer from each generative AI model

[0761] Output: Answers displayed in a list format

[0762] Step 4:

[0763] The server visually indicates the differences between answers. Specifically, it calculates the similarity between answers and displays the differences using colors and graphs. This allows users to intuitively understand the differences between answers.

[0764] Input: The answer from each generative AI model

[0765] Output: Visually indicated difference in answers

[0766] Step 5:

[0767] Users can choose their favorite answer from each answer and specify the percentage. Specifically, users can specify the percentages such as "70% of GPT-4's answers, 20% of BERT's answers, and 10% of T5's answers."

[0768] Input: Percentage of user-specified answers

[0769] Output: Specified percentage

[0770] Step 6:

[0771] The server combines the answers based on the specified percentages to generate a hybrid answer, that is, it combines the text of each answer in the specified percentages to generate a new answer.

[0772] Input: Answers from each generative AI model, in user-specified proportions

[0773] Output: Hybrid answer

[0774] Step 7:

[0775] The server applies the generated hybrid answer to a content delivery service, specifically, providing the generated content to users over the Internet.

[0776] Input: Hybrid answer

[0777] Output: Answers applied to content delivery services

[0778] In addition, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0779] "Example 1"

[0780] As one embodiment of the present invention, a system including a generation AI incorporating an emotion engine is provided. This system asks a question from a user to a plurality of generation AIs, and displays a list of answers from each generation AI. At least one of the generation AIs includes an emotion engine that recognizes the user's emotion. The emotion engine estimates the emotion from the user's tone of voice, facial expression, choice of words, etc. Then, the answer to the question is adjusted based on the estimation result. For example, if the user feels angry, the emotion engine conveys that information to the generation AI, and the generation AI adjusts to generate a more polite answer. Also, if the user feels happy, the generation AI adjusts to generate a more enjoyable answer. In this way, the emotion engine plays a role in optimizing the answer of the generation AI according to the user's emotion.

[0781] "Example 2"

[0782] Furthermore, as another embodiment of the present invention, a system is provided that generates hybrid answers by taking into account the output of an emotion engine. Specifically, when a user specifies the usage ratio of each generation AI's answer, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine conveys that information to the system, and the system generates hybrid answers that include more fun answers. In this way, by taking into account the output of the emotion engine, it is possible to generate hybrid answers that are more adapted to the user's emotions.

[0783] The process flow of each embodiment will be described below.

[0784] "Example 1"

[0785] Step 1: Receive a question from the user.

[0786] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[0787] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[0788] Step 4: Adjust the responses of each generation AI based on the estimation results. For example, if the user feels angry, the emotion engine will convey that information to the generation AI, and the generation AI will adjust to generate more polite responses.

[0789] Step 5: List the answers from each generation AI.

[0790] "Example 2"

[0791] Step 1: Receive a question from the user.

[0792] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[0793] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[0794] Step 4: When the user specifies the proportion of each AI response to be used, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine will convey that information to the system, and the system will generate hybrid responses that include more happy responses.

[0795] Step 5: Provide the hybrid answer to the user.

[0796] Example 1

[0797] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0798] In conventional generative AI systems, it is difficult to visually recognize the differences when comparing answers from multiple generative AIs. In addition, the system lacks the functionality to adjust the generative AI's answers according to the user's emotions, making it difficult to increase user satisfaction. Furthermore, there is no means to combine answers from multiple generative AIs to generate the optimal answer, making it difficult to provide users with the most optimal information.

[0799] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0800] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the differences, a means for combining each answer based on a ratio specified by the user to generate a hybrid answer, and a means for recognizing the user's emotions and adjusting the answer of the generation AI based on the emotions. This makes it possible to visually compare answers from multiple generation AIs and provide the optimal answer according to the user's emotions.

[0801] A "question input interface" is a user interface for users to input questions, and is implemented using a web browser or a dedicated application.

[0802] "Generative AI" is an artificial intelligence model that uses natural language processing technology to generate answers to input questions, and examples include GPT-4 and BERT.

[0803] The "list format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen, and the name of each generation AI and its answer are arranged so that they can be visually recognized.

[0804] "Visually showing the differences" means displaying the differences in answers from multiple generation AIs using visual means such as color coding or graphs, and calculating the similarity between answers and visually expressing the result.

[0805] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user, providing information optimized to the user's needs.

[0806] The "emotion engine" is an engine that has the ability to infer emotions from the user's tone of voice, facial expressions, choice of words, etc., and adjust the responses of the generative AI based on these inferred results.

[0807] "Recognizing the user's emotions" means using an emotion engine to estimate the user's emotional state and conveying that information to the generative AI.

[0808] This invention is a system that asks the same question to multiple AI generators, displays the answers, visually recognizes the differences, and generates hybrid answers in the ratio specified by the user. It also includes a function to recognize the user's emotions and adjust the AI ​​generator's answers based on those emotions.

[0809] System Configuration

[0810] Question input interface

[0811] A user inputs a question through a question input interface. This interface is implemented using a web browser or a dedicated application. For example, a user inputs "What's the weather like today?"

[0812] Submit a question

[0813] The server sends the question received from the user to multiple generative AIs. Specifically, the server formats the question as a prompt sentence and sends it to a generative AI model (e.g., GPT-4 or BERT). An example of a prompt sentence is "Please provide the weather forecast for today."

[0814] Receiving and listing responses

[0815] The server receives the answers from each AI generator and displays them in a list. The name of each AI generator and its answer are displayed side by side on the terminal screen. For example, it is displayed as follows:

[0816] Generative AI model 1: "The weather is sunny today."

[0817] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[0818] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[0819] Recognizing differences in answers

[0820] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[0821] Hybrid Answer Generation

[0822] The user specifies the usage ratio of each answer. The server combines each answer based on that ratio to generate a hybrid answer. For example, if the user specifies "50% of the answers from Generative AI Model 1, 30% of the answers from Generative AI Model 2, and 20% of the answers from Generative AI Model 3," the server combines these answers to generate a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[0823] Using the Emotion Engine

[0824] The server uses an emotion engine to recognize the user's emotions. The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc. Based on the inference results, the generation AI adjusts the answer. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with."

[0825] Examples

[0826] The user enters "What's the weather like today?" into the question input interface and clicks the send button. The server sends this question to the generation AI as a prompt sentence, "Please provide the weather forecast for today." The server receives the answer from the generation AI and displays it in a list format on the device screen. The server calculates the similarity of each answer and displays them in different colors. The user specifies the usage ratio of each answer, and the server generates a hybrid answer based on that ratio. In addition, an emotion engine is used to recognize the user's emotions and adjust the answer of the generation AI.

[0827] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0828] Step 1:

[0829] A user inputs a question through the question input interface. Specifically, the user inputs "What's the weather like today?" and clicks the send button. The input question is sent to the server. The input data is the question in text format, and the output is the question data sent to the server.

[0830] Step 2:

[0831] The server sends the questions received from the user to multiple generation AIs. Specifically, the server formats the questions as prompt sentences and sends them to a generation AI model (e.g., GPT-4 or BERT). The input data is the question text from the user, and the output is the prompt sentence sent to the generation AI. For example, the server generates a prompt sentence such as "Please provide the weather forecast for today."

[0832] Step 3:

[0833] The server receives the answers from each generation AI. The responses from the generation AI are returned to the server in a data format such as JSON. The input data is the answer data from the generation AI, and the output is the answer data received by the server. For example, GPT-4 returns the answer "Today's weather is sunny."

[0834] Step 4:

[0835] The server displays the received answers in a list format. The name of each AI generator and its answer are displayed side by side on the terminal screen. The input data are the answer data from each AI generator, and the output is the answer displayed on the terminal in a list format. For example, it is displayed as follows:

[0836] Generative AI model 1: "The weather is sunny today."

[0837] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[0838] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[0839] Step 5:

[0840] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. The input data is the answer data from each generation AI, and the output is the visually displayed difference information. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[0841] Step 6:

[0842] The user specifies the usage ratio of each answer. Specifically, the user specifies "50% of answers from AI model 1, 30% of answers from AI model 2, and 20% of answers from AI model 3." The input data is the usage ratio specified by the user, and the output is the ratio data sent to the server.

[0843] Step 7:

[0844] The server combines each answer based on the ratio and generates a hybrid answer. The input data is the usage ratio specified by the user and the answer data from each generation AI, and the output is the generated hybrid answer. For example, it generates a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[0845] Step 8:

[0846] The server recognizes the user's emotions using an emotion engine. The emotion engine estimates emotions from the user's tone of voice, facial expressions, choice of words, etc. The input data is the user's voice and video data, and the output is estimated emotion information.

[0847] Step 9:

[0848] The server adjusts the answer of the generation AI based on the estimation result. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with." The input data is the estimated emotion information and the answer data from the generation AI, and the output is the adjusted answer data.

[0849] (Application example 1)

[0850] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0851] In conventional generative AI systems, when comparing answers from multiple generative AIs, adjustments were not made to take the user's emotions into account, making it difficult to obtain the optimal answer according to the user's emotions. In addition, when combining answers from generative AIs, it was not possible to reflect the user's emotions, making it difficult to increase user satisfaction.

[0852] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0853] In this invention, the server includes a means for asking the same question to multiple generation AIs, a means for displaying a list of answers from each generation AI, a means for recognizing differences between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for recognizing the user's emotions, and a means for adjusting the answer of the generation AI based on the recognized emotions. This makes it possible to provide an optimal answer according to the user's emotions and increase user satisfaction.

[0854] "Generative AI" is a system that uses artificial intelligence technology to automatically generate answers to questions from users.

[0855] A "means for asking a question" is a device or program that has the function of sending questions from users to multiple generation AIs.

[0856] A "means for displaying a list of answers" is a device or program that has the function of displaying answers received from multiple generation AIs side-by-side on a single screen.

[0857] A "means for recognizing differences" is a device or program that has the function of comparing answers from multiple generation AIs and visually showing the differences.

[0858] A "means for regenerating a hybrid answer" is a device or program that has the function of combining answers from multiple generation AIs based on a ratio specified by the user to generate a new answer.

[0859] A "means for recognizing emotions" is a device or program that has the function of inferring emotions from a user's tone of voice, facial expressions, choice of words, etc.

[0860] A "means for adjusting an answer" is a device or program that has the function of optimizing the answer of the generating AI based on the recognized user's emotions.

[0861] The system for implementing this invention has the functions of asking the same question to multiple AI generators, displaying a list of answers from each AI generator, recognizing the differences between each answer, and regenerating a hybrid answer by combining each answer in a ratio specified by the user. It also includes a function of recognizing the user's emotions and adjusting the AI ​​generator's answer based on those emotions.

[0862] System Configuration

[0863] The system consists of the following main components:

[0864] 1. Question input interface: This is an interface for users to input questions. This is implemented on devices such as smartphones and computers.

[0865] 2. Generative AI module: A module that includes multiple generative AIs. Each generative AI automatically generates answers to questions posed by users.

[0866] 3. Answer display module: This module displays a list of answers from each generation AI. It displays the differences in answers in a format that allows visual recognition.

[0867] 4. Emotion recognition module: This module estimates emotions from the user’s tone of voice, facial expressions, choice of words, etc.

[0868] 5. Answer Adjustment Module: This module adjusts the generative AI’s answers based on the recognized emotions.

[0869] 6. Hybrid answer generation module: This module combines answers from each generation AI based on the ratio specified by the user to generate a hybrid answer.

[0870] Program processing explanation

[0871] The server receives questions from users through a question input interface. These questions are sent to the generation AI module, and multiple generation AIs generate their own answers. The generated answers are displayed in a list by the answer display module. At this time, the differences between each answer are displayed in a format that allows visual recognition.

[0872] Next, the emotion recognition module recognizes the user's emotions. For example, it estimates emotions from the user's tone of voice and facial expressions. This recognized emotion information is sent to the answer adjustment module and used to adjust the answer of the generation AI. The adjusted answer is then displayed again in the answer display module.

[0873] Finally, based on the ratio specified by the user, the hybrid answer generation module combines the answers from each generation AI to generate a hybrid answer, which is finally provided to the user.

[0874] Examples

[0875] For example, if a user asks, "What movie would you recommend to watch tonight?", each generative AI will suggest its own movie recommendation. If the emotion recognition module recognizes the user's emotions and determines that the user is tired, for example, the answer will be adjusted to suggest a relaxing movie. Finally, a hybrid recommendation list is generated that combines answers from multiple generative AIs based on the ratio specified by the user.

[0876] Example prompt:

[0877] What movie would you recommend to watch tonight?

[0878] In this way, it is possible to provide the user with an optimal answer according to his / her feelings, thereby increasing the user's satisfaction.

[0879] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0880] Step 1:

[0881] A user inputs a question into the question input interface using a terminal. The input question is transmitted to the server. The input data is a question about the content that the user wants to view.

[0882] Step 2:

[0883] The server sends the received questions to multiple generation AIs. Each generation AI automatically generates an answer to the question. The generated answer is sent back to the server. The input data is the user's question, and the output data is the answer from each generation AI.

[0884] Step 3:

[0885] The server sends the answers from each AI generator to the answer display module and displays them in a list. The answer display module displays the differences between the answers in a format that allows the differences to be visually recognized. The input data are the answers from each AI generator, and the output data is the visually displayed list of answers.

[0886] Step 4:

[0887] When a user inputs a question into the question input interface using a terminal, the emotion recognition module analyzes the user's tone of voice and facial expression to recognize the emotion. The input data is the user's tone of voice and facial expression, and the output data is the recognized emotion information.

[0888] Step 5:

[0889] The server sends the recognized emotion information to the answer adjustment module. The answer adjustment module adjusts the answer of each generation AI based on the emotion information. For example, if the user is recognized as tired, the answer is adjusted to suggest relaxing content. The input data are the recognized emotion information and the answer from each generation AI, and the output data are the adjusted answer.

[0890] Step 6:

[0891] The server again transmits the adjusted answers to the answer display module for displaying them in a list. The input data is the adjusted answers, and the output data is a visually displayed list of adjusted answers.

[0892] Step 7:

[0893] The user specifies the usage ratio of each answer using the terminal. For example, the user specifies "50% of the answers from AI1, 30% of the answers from AI2, and 20% of the answers from AI3." The input data is the usage ratio specified by the user.

[0894] Step 8:

[0895] The server generates a hybrid answer by combining the answers from each generation AI in the hybrid answer generation module based on the ratio specified by the user. The input data is the usage ratio specified by the user and the answers from each generation AI, and the output data is the hybrid answer.

[0896] Step 9:

[0897] The server provides the generated hybrid answer to the user. The user uses a terminal to confirm the hybrid answer. The input data is the hybrid answer, and the output data is the final answer provided to the user.

[0898] Example 2

[0899] Next, a description will be given of Example 2 of the embodiment 2. In the following description, the data processing device 12 is referred to as a "server" and the smart glasses 214 are referred to as a "terminal".

[0900] In conventional generative AI systems, it was difficult to compare answers from multiple generative AIs and combine them to generate the optimal answer. In addition, the system did not generate answers that took the user's feelings into account, so it was not possible to increase user satisfaction. This meant that users could not efficiently obtain the information they were looking for, and the system's value as a user decreased.

[0901] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0902] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the difference, a means for combining answers from each generation AI in a ratio specified by the user to generate a hybrid answer, and a means for adjusting the hybrid answer by taking into account the output of the emotion analysis engine. This allows the user to compare answers from multiple generation AIs and obtain the optimal answer. Furthermore, the generation of an answer that takes into account the user's emotions can increase the user's satisfaction.

[0903] A "question input interface" is a user interface that allows a user to input a question and send the question to the system.

[0904] "Generative AI" is an artificial intelligence model that generates answers in natural language in response to a given question.

[0905] The "list format" is a format in which a plurality of items are visually organized and displayed, making it easier for the user to compare the items.

[0906] "Visually showing" means displaying information using visual means such as color coding or graphs so that the user can understand it intuitively.

[0907] A "hybrid answer" is a new answer generated by combining answers from multiple generative AIs, incorporating the characteristics of each answer.

[0908] An "emotion analysis engine" is an engine that analyzes emotions from user input and actions and provides the results to the system.

[0909] A "weighted average" is a method of calculating an average by assigning different weights to each piece of data, and is used to emphasize the influence of certain data.

[0910] An "API Request" is a request made to another system or service through an Application Program Interface, which is a means of sending or receiving data.

[0911] The present invention is a system for generating optimal answers to questions from users using multiple generative AI models. A specific embodiment of this system will be described below.

[0912] The server receives questions from users through a question input interface. This interface is implemented using a web browser or a dedicated application. Specifically, a browser such as Google Chrome can be used. For example, the user inputs a question such as, "Please tell me about the future of AI technology."

[0913] The server sends the received questions to multiple generation AIs. OpenAI's GPT-3 and Google's BERT are used as generation AIs. Each generation AI generates an answer to the question and sends it back to the server. Specifically, the question is sent to each generation AI's endpoint using an HTTP POST request.

[0914] The server receives the answers from each AI and displays them in a list format. This display is implemented on a web page using HTML and CSS. For example, the name of each AI and its answer are displayed side by side on the screen in the form of "Answer from AI model A" and "Answer from AI model B."

[0915] The server compares the answers from each generation AI and visually displays the differences. Specifically, it uses the Python library scikit-learn to calculate the similarity between answers, and displays the differences using color coding and graphs based on the similarity. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red.

[0916] Users can specify the proportion of each answer to be used. The server receives input from the user and combines each answer based on that proportion to generate a hybrid answer. Specifically, it uses Python's NumPy library to weight and average each answer by the specified proportion to generate a new answer. For example, if a user specifies "70% of answers from Generative AI Model A and 30% from Generative AI Model B," each answer will be combined according to that proportion.

[0917] Furthermore, by using a sentiment analysis engine, it is possible to generate hybrid answers that adapt to the user's emotions. As a sentiment analysis engine, IBM's Watson Tone Analyzer is used. The sentiment analysis engine analyzes the user's input and actions to determine emotions and sends this information to the server. For example, if the sentiment analysis engine determines that the user is happy, the server adjusts the answer to include more happy content.

[0918] In this way, users can compare answers from multiple AI generators to obtain the best answer. In addition, by generating answers that take the user's emotions into account, user satisfaction can be increased.

[0919] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0920] Step 1:

[0921] The user enters a question.

[0922] A user accesses the question input interface through a web browser or a dedicated application and inputs a question. For example, a user accesses a web page using Google Chrome and inputs "Please tell me about the future of AI technology" in the text box. The input question is sent to the server.

[0923] Step 2:

[0924] The server receives the question and sends it to each generating AI.

[0925] The server sends the question received from the user to multiple generation AIs. Specifically, it uses an HTTP POST request to send the question to the endpoint of each generation AI. OpenAI's GPT-3 and Google's BERT are used as generation AIs. The input is the user's question, and the output is the question sent to each generation AI.

[0926] Step 3:

[0927] Each generation AI generates an answer to the question and sends it back to the server.

[0928] Each generation AI generates an answer to the question it receives. For example, GPT-3 generates the answer "AI technology will continue to evolve in the future and will be used in many fields such as medicine and education," and BERT generates the answer "AI technology will contribute to the realization of autonomous driving and smart cities." These answers are sent back to the server as API responses. The input is the question sent to each generation AI, and the output is the generated answer.

[0929] Step 4:

[0930] The server displays a list of answers from each generation AI.

[0931] The server uses HTML and CSS to display the answers received from each generation AI in a list format on a web page. For example, the name of each generation AI and its answer are displayed side by side on the screen in the form of "Answer from generation AI model A" and "Answer from generation AI model B." The input is the answer from each generation AI, and the output is the answer displayed in a list format.

[0932] Step 5:

[0933] The server visually indicates the difference between each answer.

[0934] The server uses Python's scikit-learn library to calculate the similarity between each answer. Based on the calculation results, the similarity of the answers is visually displayed using color coding or graphs. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red. The input is the answer from each generation AI, and the output is the difference between the visually displayed answers.

[0935] Step 6:

[0936] The user specifies the usage ratio of each answer.

[0937] The user accesses an interface to specify the usage ratio of each answer. For example, the user can use a slider or text box to input "70% of answers from AI model A and 30% of answers from AI model B." The input is the usage ratio specified by the user, and the output is the usage ratio sent to the server.

[0938] Step 7:

[0939] The server generates a hybrid answer based on the specified percentages.

[0940] The server combines each answer based on the ratio specified by the user. Specifically, it uses Python's NumPy library to weight and average each answer to generate a new hybrid answer. For example, it generates an answer such as "AI technology will continue to evolve in the future, contributing to the realization of healthcare, education, autonomous driving, and smart cities." The input is the usage ratio specified by the user and the answer from each generation AI, and the output is the generated hybrid answer.

[0941] Step 8:

[0942] The emotion analysis engine analyzes the user's emotions and transmits them to the server.

[0943] Sentiment analysis engines analyze user input and behavior to determine emotions. For example, IBM's Watson Tone Analyzer can be used to analyze emotions from a user's text input and determine that "the user is happy." This information is sent to a server. The input is the user's input and behavior, and the output is the analyzed emotional information.

[0944] Step 9:

[0945] The server adjusts the hybrid answer taking into account the output of the sentiment analysis engine.

[0946] The server adjusts the hybrid answer taking into account the output from the sentiment analysis engine. For example, if it is determined that the user is happy, the server adjusts the answer to include more fun content. Specifically, it generates an answer that emphasizes the fun element. The input is the emotion information from the sentiment analysis engine and the hybrid answer, and the output is the adjusted hybrid answer.

[0947] (Application example 2)

[0948] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the smart glasses 214 will be referred to as a "terminal".

[0949] In systems using conventional generative AI models, when comparing answers from multiple generative AI models, it was difficult to visually recognize the differences in answers, and it was not possible to generate hybrid answers that adapted to the user's emotions. As a result, it was necessary for users to make a great deal of effort to obtain the optimal answer, making it difficult to obtain information efficiently. Furthermore, since it was not possible to take the user's emotions into account when combining answers from generative AI models, it was not possible to increase user satisfaction.

[0950] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0951] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for regenerating a hybrid answer by combining each answer in a ratio specified by a user, a means for generating a hybrid answer taking into account the output of the emotion engine, and a means for calculating the similarity between answers and visually displaying the difference. This allows a user to efficiently compare answers from a plurality of generative AI models and visually recognize the difference. In addition, by generating a hybrid answer that takes into account the user's emotions, the user's satisfaction can be increased.

[0952] A "generative AI model" is an algorithm or system that uses artificial intelligence to generate content such as text and images.

[0953] A "means for asking questions" is an interface that has the ability to receive questions from users and send them to a generative AI model.

[0954] A "means for displaying answers in a list" is an interface with the function of displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[0955] A "means of recognizing differences" is a system that has the ability to compare answers from multiple generative AI models and visually show the differences.

[0956] A "means for regenerating hybrid answers" is a system that has the ability to generate new answers by combining answers from multiple generative AI models based on a ratio specified by the user.

[0957] An "emotion engine" is an algorithm or system that analyzes user emotions and adjusts the system's operation based on the results.

[0958] "Means for calculating similarity" is a system capable of numerically evaluating the similarity between answers from multiple generative AI models.

[0959] A "means for visually displaying differences" is a system that has the ability to visually display the differences between responses using color coding, graphs, or other means.

[0960] A system for implementing the present invention includes a server, a user terminal, and a generative AI model. The server has a function of receiving a question from a user and transmitting the question to a plurality of generative AI models. The user terminal provides a question input interface, allowing the user to input a question. The generative AI model generates an answer to the received question and transmits it back to the server.

[0961] The server receives answers from multiple generative AI models and displays them in a list. Specifically, the name of each generative AI model and its answer are displayed side by side. In addition, the server recognizes the differences between each answer and calculates the similarity between the answers. Based on the similarity, the differences between the answers are visually displayed using color coding and graphs.

[0962] The user can check the answers from each generative AI model and specify the proportion of each answer to be used. The server combines each answer based on the proportion specified by the user to generate a hybrid answer. At this time, the emotion engine analyzes the user's emotions and adjusts the hybrid answer based on the results. For example, if the emotion engine determines that the user is happy, it will adjust the hybrid answer to include more cheerful answers.

[0963] As a concrete example, if a user enters the question "What is the best travel destination for our next vacation?", the server sends this question to multiple generative AI models. The answers from each generative AI model are displayed in a list, and the differences between the answers are visually shown. The user specifies the proportion of each answer to be used, and a hybrid answer is generated based on the analysis results of the emotion engine.

[0964] The hardware used includes user devices (e.g., smartphones, smart glasses, head-mounted displays, etc.) and servers. The software used includes generative AI models (e.g., OpenAI APIs), emotion engines (e.g., TextBlob), and data visualization tools (e.g., Matplotlib).

[0965] An example of a prompt is as follows:

[0966] "What's your recommended destination for our next vacation?"

[0967] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0968] Step 1:

[0969] The user uses the terminal to input a question.

[0970] Input: The prompt text entered by the user (e.g. "What are some recommended travel destinations for our next vacation?")

[0971] Output: The entered question is sent to the server.

[0972] Specific operation: The user inputs a question using the question input interface of the terminal and presses the send button. The terminal sends the input question to the server.

[0973] Step 2:

[0974] The server receives the question and sends it to multiple generative AI models.

[0975] Input: The question submitted by the user

[0976] Output: The question sent to the generative AI model

[0977] How it works: The server sends the received question to multiple generative AI models, each of which receives the question and prepares to generate an answer.

[0978] Step 3:

[0979] The generative AI model generates answers to the questions and sends them back to the server.

[0980] Input: The question sent to the generative AI model

[0981] Output: The answer returned by the generative AI model.

[0982] Specific operation: Each generative AI model generates an answer based on the question and sends the answer back to the server. The server receives the answer from each generative AI model.

[0983] Step 4:

[0984] The server lists the answers from each generative AI model.

[0985] Input: The answer returned by the generative AI model

[0986] Output: List of answers

[0987] Specific operation: The server displays the answers from each generative AI model in a list format on the screen. The name of each generative AI model and its answer are displayed side by side.

[0988] Step 5:

[0989] The server recognizes the difference between each answer and displays it visually.

[0990] Input: Answer from a generative AI model

[0991] Output: Visually displayed difference in answers

[0992] Specific operation: The server calculates the similarity of each answer and visually displays the differences between the answers using color coding and graphs, allowing users to recognize the differences between each answer at a glance.

[0993] Step 6:

[0994] The user specifies the usage ratio of each answer.

[0995] Input: The percentage of each answer that the user specifies to be used (e.g. "70% of the time answer A, 20% of the time answer B, 10% of the time answer C")

[0996] Output: The specified usage percentage is sent to the server.

[0997] Specific operation: The user uses the terminal interface to specify the usage ratio of each answer and presses the send button. The terminal then sends the specified usage ratio to the server.

[0998] Step 7:

[0999] The server generates a hybrid answer based on the specified percentages.

[1000] Input: Usage rate of each answer specified by the user, answers from the generative AI model

[1001] Output: Hybrid answer

[1002] Specific behavior: The server combines each answer based on the user-specified percentage to generate a hybrid answer, which is then displayed to the user.

[1003] Step 8:

[1004] The server uses an emotion engine to analyze the user's emotions and tailor the hybrid response.

[1005] Input: Hybrid answers, user sentiment data

[1006] Output: A tailored hybrid answer

[1007] Specific operation: The server uses the emotion engine to analyze the user's emotions and adjusts the hybrid answers based on the results. For example, if the emotion engine determines that the user is happy, the answers will be adjusted to include more happy answers. The adjusted hybrid answers are finally displayed to the user.

[1008] 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 a voice indicating a user input for the result of the specific processing. The control unit 46A transmits the voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1009] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1010] Another example of generative AI is Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) are mentioned.

[1011] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[1012] [Third embodiment]

[1013] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1014] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1015] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[1016] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. 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. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1017] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[1018] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[1019] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[1020] Fig. 6 shows an example of main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1021] The specific processing program 56 is an example of a "program" according to the technology of the present 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 in accordance with the specific processing program 56 executed on the RAM 30.

[1022] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1023] In the headset type terminal 314, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1024] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1025] "Example 1"

[1026] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[1027] "Example 2"

[1028] In one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[1029] The process flow of each embodiment will be described below.

[1030] "Example 1"

[1031] Step 1: The user inputs a question through the question input interface. This question is then posed to multiple generation AIs.

[1032] Step 2: The system sends a question to each Generator AI, which answers the question using its own algorithm.

[1033] Step 3: The system receives answers from each AI generator. These answers are displayed in a list format. Specifically, the name of each AI generator and its answer are displayed side by side on the screen.

[1034] Step 4: The system compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using color coding and graphs based on the similarity.

[1035] Step 5: The user specifies the percentage of each answer to be used. For example, "Answer A will be 70%, Answer B will be 20%, and Answer C will be 10%."

[1036] Step 6: The system combines the answers based on the percentages specified by the user to generate a hybrid answer, which becomes the final output.

[1037] Example 1

[1038] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1039] Conventional generative AI systems rely on answers from a single generative AI, which limits the variety and accuracy of answers. In addition, there is no way to compare answers from multiple generative AIs, making it difficult for users to select the most appropriate answer. Furthermore, there is a lack of functionality to combine multiple answers to generate hybrid answers, which means that it is not possible to flexibly generate answers that meet the needs of users.

[1040] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1041] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually showing the differences, and a means for regenerating a hybrid answer by combining the answers from each generation AI based on a ratio specified by the user. This makes it possible to compare various answers from multiple generation AIs and select the optimal answer. It also makes it possible to flexibly generate hybrid answers according to the needs of the user.

[1042] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[1043] "Generative AI" is an artificial intelligence model that generates natural language responses based on a given prompt.

[1044] An "API endpoint" is an interface for sending questions to the generation AI and receiving answers.

[1045] "List format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen.

[1046] "Visually showing" means showing the differences in responses using visual means such as color coding or graphs.

[1047] "Similarity" is an index showing the degree of agreement between the contents of multiple answers.

[1048] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user.

[1049] The "proportion" is the ratio that indicates the extent to which the user will use the answers from each generation AI.

[1050] This invention is a system that asks the same question to multiple generative AI models and compares and integrates their answers. A specific embodiment of this system will be described below.

[1051] Question input interface

[1052] The user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and provides a form where the user can enter a question in text format. Once the user enters a question and clicks the "Submit" button, the question is sent to the server.

[1053] Submit a question

[1054] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt. The generative AI models used include, for example, natural language processing models and machine learning models.

[1055] Receiving and listing responses

[1056] The server receives the answers from each generative AI model and temporarily stores them. Next, the server generates a web page to display the received answers in a list format. This web page displays the name of each generative AI model and its answer side by side. The user can view this page in the web browser of the terminal.

[1057] Recognizing differences in answers

[1058] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity. This process is carried out using text analysis algorithms and data visualization tools.

[1059] Regenerating a hybrid answer

[1060] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[1061] Examples

[1062] Example of question input

[1063] The user enters the following question into the question entry interface:

[1064] "What do you think the future of AI is?"

[1065] Examples of prompt statements

[1066] Here is an example of a prompt to be input to the generative AI model:

[1067] "What do you think the future of AI is?"

[1068] Example of answer list display

[1069] Generative AI Model 1: “AI will play an important role in many fields in the future, including medicine and education.”

[1070] Generative AI Model 2: "AI will help make autonomous driving and smart cities a reality."

[1071] Generative AI Model 3: “AI will continue to evolve as a tool to make human life more convenient.”

[1072] Examples of visual display of response differences

[1073] The parts of the answers from Generative AI Model 1 and Generative AI Model 2 that are highly similar are displayed in green.

[1074] The areas where the answers from Generative AI Model 1 and Generative AI Model 3 have low similarity are displayed in red.

[1075] A concrete example of hybrid answer regeneration

[1076] The user specifies "70% of the answers from Generative AI Model 1, 20% from Generative AI Model 2, and 10% from Generative AI Model 3." The server combines each answer based on that ratio and generates a hybrid answer like this:

[1077] "Going forward, AI will play an important role in many fields, including healthcare and education, and will also contribute to the realization of autonomous driving and smart cities, continuing to evolve as a tool to make human life more convenient."

[1078] In this way, users can compare various answers from multiple generative AI models and select the optimal answer. It also enables flexible generation of hybrid answers according to the needs of users.

[1079] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1080] Step 1: Enter your question

[1081] The user opens a web browser on the terminal and accesses the question input interface. The user enters a question in the text box and clicks the "Send" button. This operation sends the question to the server.

[1082] Input: A question typed in by the user (e.g., "Tell me about the future of AI.")

[1083] Output: The question data sent to the server

[1084] Step 2: Submit your question

[1085] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt.

[1086] Input: The query data sent to the server

[1087] Output: A prompt sent to each generative AI model (e.g., "Tell me about the future of AI.")

[1088] Step 3: Receiving a response

[1089] The server receives answers from each generative AI model. Each generative AI model generates an answer based on the prompt sentence and returns it to the server. The server temporarily stores these answers.

[1090] Input: Answer data from each generative AI model

[1091] Output: Answer data stored on the server (e.g., "AI will play an important role in many fields, such as medicine and education, in the future.")

[1092] Step 4: List the responses

[1093] The server generates a web page to display the answers of each generative AI model in a list format. The user views this page in a web browser on their device.

[1094] Input: Answer data saved on the server

[1095] Output: A web page displayed in list format (Example: Generative AI model 1: "AI will play an important role in many fields in the future, such as medicine and education.")

[1096] Step 5: Recognize differences in responses

[1097] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity.

[1098] Input: Answer data saved on the server

[1099] Output: A web page with visual indication of the differences (e.g. high similarity is shown in green, low similarity in red)

[1100] Step 6: Regenerate the hybrid answer

[1101] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[1102] Input: The percentage of each answer specified by the user (e.g. 70% of the answers from AI model 1, 20% of the answers from AI model 2, and 10% of the answers from AI model 3)

[1103] Output: A web page with a hybrid answer (e.g., "AI will play an important role in many fields, such as medicine and education, and will continue to evolve as a tool to make human life more convenient, contributing to the realization of autonomous driving and smart cities.")

[1104] (Application example 1)

[1105] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1106] In systems using conventional generative AI models, there is a lack of a means to compare answers from multiple generative AI models and obtain the optimal answer. In addition, it is difficult to visually recognize the differences between answers from each generative AI model, and there is insufficient support for users to select the optimal answer. Furthermore, when combining answers from multiple generative AI models to generate a hybrid answer, a means is required that can flexibly respond based on the ratio specified by the user.

[1107] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1108] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for recognizing the difference between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for calculating and visually displaying the similarity between each answer, and a means for the user to specify the ratio for generating the hybrid answer. This makes it possible to compare answers from a plurality of generative AI models and obtain an optimal answer. In addition, by visually recognizing the difference between each answer, support for the user to select the optimal answer is enhanced. Furthermore, it becomes possible to flexibly generate a hybrid answer based on the ratio specified by the user.

[1109] A "generative AI model" is a type of artificial intelligence that uses natural language processing to generate answers to user questions or prompts.

[1110] The "means of asking questions" is a function that allows users to send questions to multiple generative AI models.

[1111] "Means of displaying answers in a list" is a function for displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[1112] A "means of recognizing differences" is a function that can compare answers from multiple generative AI models and identify the differences between them.

[1113] A "means for regenerating hybrid answers" is a function for generating new answers by combining answers from multiple generative AI models in a ratio specified by the user.

[1114] "Means for calculating similarity" is a function for numerically evaluating the similarity between answers from multiple generative AI models.

[1115] "Means for visual display" refers to a function for displaying the calculated similarity or difference in answers in a visual format such as a graph or color coding.

[1116] "Means for specifying proportions" refers to a function that enables users to specify the proportions in which answers from each generative AI model should be combined.

[1117] The system for implementing the present invention operates in cooperation with a server, a terminal, and a user. A specific embodiment of the system will be described below.

[1118] System Configuration

[1119] server

[1120] The server includes means for asking the same question to multiple generative AI models, means for displaying a list of answers from each generative AI model, means for recognizing differences between each answer, means for combining each answer in a ratio specified by a user to regenerate a hybrid answer, means for calculating and visually displaying the similarity of each answer, and means for a user to specify the ratio for generating a hybrid answer.

[1121] The server uses the following software and hardware:

[1122] Generative AI Models: OpenAI API

[1123] Data processing: Python program

[1124] Visualization: Matplotlib, Scikit-learn

[1125] Terminal

[1126] The terminal provides an interface for the user to enter questions and view generated answers. The terminal may be a smartphone, smart glasses, a head-mounted display, or the like.

[1127] User

[1128] Users input questions through their devices, review answers from generative AI models, and specify the proportions for generating hybrid answers.

[1129] Processing flow

[1130] 1. Entering a question: The user enters a question through the terminal, for example, a prompt such as "Please tell me some ideas for a new blog post."

[1131] 2. Question submission: The server submits the input question to multiple generative AI models.

[1132] 3. Obtaining answers: The server receives answers from each generative AI model.

[1133] 4. Displaying a list of answers: The server displays the answers from each generative AI model in a list format on the terminal.

[1134] 5. Difference Recognition: The server calculates the similarity of each answer and displays it visually, for example using a heatmap to show the similarity between answers.

[1135] 6. Specify Proportion: The user specifies what proportion of each answer should be combined.

[1136] 7. Generate hybrid answer: The server generates a hybrid answer based on the specified percentages and displays it on the terminal.

[1137] Examples

[1138] When a user types a question like "What are some ideas for a new blog post?", the server sends the question to multiple generative AI models, which can produce answers like these:

[1139] Generative AI Model A: "Latest trends in AI technology"

[1140] Generative AI Model B: "Advantages and disadvantages of remote work"

[1141] Generative AI Model C: "Sustainable Energy Future"

[1142] The server lists these answers and visually displays the similarity between the answers in a heat map. For example, the user can specify that "Latest trends in AI technology" should be combined at 70%, and the other two at 20% and 10%, respectively. The server generates a hybrid answer based on this specification and provides it to the user.

[1143] In this way, users can compare answers from multiple generative AI models to obtain the best answer. Also, visual recognition of the differences between each answer provides enhanced support for selecting the more appropriate answer. Furthermore, it is possible to flexibly generate hybrid answers based on the ratio specified by the user.

[1144] The flow of the specific process in the application example 1 will be described with reference to FIG.

[1145] Step 1:

[1146] The user inputs a question through a terminal.

[1147] Input: The prompt the user enters (e.g. "Can you give me some ideas for new blog posts?")

[1148] Output: The prompt text entered

[1149] Specific operations: The user inputs a question using the input interface of the terminal and presses the send button.

[1150] Step 2:

[1151] The server sends the input question to multiple generative AI models.

[1152] Input: The prompt text entered by the user.

[1153] Output: The prompt sent to each generative AI model

[1154] Specific operation: The server uses the API to send prompt sentences to each generative AI model.

[1155] Step 3:

[1156] The server receives the answers from each generative AI model.

[1157] Input: The prompt sent to each generative AI model

[1158] Output: The answer from each generative AI model

[1159] Specific operation: The server receives the API response from each generative AI model and obtains the answer.

[1160] Step 4:

[1161] The server displays the answers from each generative AI model in a list format on the terminal.

[1162] Input: The answer from each generative AI model

[1163] Output: A list of answers displayed on the terminal

[1164] Specific operation: The server compiles the answers from each generative AI model into a single list and sends it to the terminal's display interface.

[1165] Step 5:

[1166] The server calculates the similarity of each answer and displays it visually.

[1167] Input: The answer from each generative AI model

[1168] Output: A heatmap showing similarity

[1169] Specific operation: The server calculates the vector of each answer, calculates the similarity using cosine similarity, and generates a heat map of the result and displays it on the terminal.

[1170] Step 6:

[1171] The user specifies how much of each answer should be combined.

[1172] Input: The percentage of each answer that you specify (e.g. 70%, 20%, 10%)

[1173] Output: Specified percentage

[1174] Specific behavior: The user uses the device interface to specify the percentage of each answer using sliders and input fields.

[1175] Step 7:

[1176] The server generates a hybrid answer based on the specified percentage and displays it on the terminal.

[1177] Input: Answers from each generative AI model, in user-specified proportions

[1178] Output: Hybrid answer

[1179] Specific operation: The server combines each answer based on the specified ratio to generate a new hybrid answer, and sends the result to the device for display.

[1180] Example 2

[1181] Next, a second embodiment of the second form will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1182] Conventional generative AI systems lacked the means to compare answers from multiple generative AIs and effectively integrate them. As a result, users had to check the answers from each generative AI individually and compare and integrate them manually, which was inefficient. In addition, there was no way to visually recognize the differences in the answers from each generative AI, making it difficult to intuitively understand the differences in the quality and content of the answers.

[1183] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1184] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually indicating the differences, and a means for combining the answers from each generation AI in a ratio designated by the user to regenerate a hybrid answer. This enables the user to efficiently compare and integrate answers from multiple generation AIs and obtain an optimal hybrid answer while visually recognizing the differences.

[1185] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[1186] "Generative AI" is an artificial intelligence model that generates natural language answers based on input prompts.

[1187] A "list format" is a format in which multiple items are displayed side by side on one screen or page.

[1188] "Visually showing" means displaying information using visual means such as color coding or graphs.

[1189] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[1190] The "proportion specified by the user" is a number that specifies the proportion in which the user will combine answers from each generation AI.

[1191] A "server" is a computer system that performs processes such as receiving questions, sending questions to the generation AI, receiving answers, displaying the answers, comparing answers, and generating hybrid answers.

[1192] This invention is a system that asks the same question to multiple generation AIs and compares and integrates their answers. A specific embodiment of this system will be described below.

[1193] Question input interface

[1194] A user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and includes a text box for the user to enter a question and a submit button. When the user enters a question and clicks the submit button, the question is sent to the server.

[1195] Submit a question

[1196] The server sends the questions received from the user to multiple generation AIs. The server uses a Python HTTP request library (e.g., requests) to send questions to the API endpoints of each generation AI. Specifically, it sends the following prompt text to the generation AI:

[1197] "What do you think the future of AI is?"

[1198] Receiving a response

[1199] The server receives answers from each AI generator. It waits for a response from the API of each AI generator and stores the received answers. For example, it receives an answer from AI generator A saying, "AI will continue to evolve in the future and support people's lives in various fields."

[1200] List of responses

[1201] The server displays the answers received from each AI generator in a list format. Using an HTML template engine (e.g. Jinja2), the name of each AI generator and its answer are displayed side by side on a web page. For example, it is displayed as follows:

[1202] Generative AI A: "AI will continue to evolve and support people's lives in various fields."

[1203] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[1204] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[1205] Recognizing differences in answers

[1206] The server compares the answers from each generation AI and visually displays the differences. It uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using colors and graphs based on the similarity. For example, if the answers of generation AI A and generation AI B show a high similarity, they are displayed in the same color.

[1207] Regenerating a hybrid answer

[1208] The user specifies the percentage of each answer that should be used as input. The server combines the answers based on that percentage to generate a hybrid answer. For example, if the user specifies "70% of the answers should be from answer A, 20% from answer B, and 10% from answer C," the server will generate the following hybrid answer:

[1209] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[1210] In this way, users can efficiently compare and integrate answers from multiple generative AIs, visually recognizing the differences, and arrive at the optimal hybrid answer.

[1211] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1212] Step 1:

[1213] The user opens a web browser on the device and accesses the question input interface. The user enters a question in the text box and clicks the send button. The input is the question entered by the user, and the output is the question data sent to the server. In concrete terms, the user enters "Tell me about the future of AI" and clicks the send button.

[1214] Step 2:

[1215] The server sends questions received from the user to multiple generation AIs. The input is the question data received from the user, and the output is the prompt text sent to each generation AI. The server uses Python's HTTP request library (e.g., requests) to send questions to the API endpoint of each generation AI. Specifically, the server sends the question "Tell me about the future of AI" to generation AI A, generation AI B, and generation AI C.

[1216] Step 3:

[1217] The server receives the answers from each generation AI. The input is the API response from each generation AI, and the output is the received answer data. The server waits for a response from the API of each generation AI and stores the received answers. In concrete terms, the server receives the answer from generation AI A that "AI will continue to evolve in the future and support people's lives in various fields."

[1218] Step 4:

[1219] The server displays the answers received from each generation AI in a list format. The input is the received answer data, and the output is the answers in a list format displayed on a web page. The server uses an HTML template engine (e.g. Jinja2) to display the name of each generation AI and its answer side by side on the web page. Specifically, the server displays the following on the web page:

[1220] Generative AI A: "AI will continue to evolve and support people's lives in various fields."

[1221] Generative AI B: “The future of AI is very bright and will bring about innovations in many fields, including medicine, education, and entertainment.”

[1222] Generative AI C: "AI technology is developing rapidly and may surpass human intelligence in the future."

[1223] Step 5:

[1224] The server compares the answers from each generation AI and visually displays the differences. The input is the received answer data, and the output is the visually displayed difference information. The server uses a natural language processing library (e.g., NLTK or spaCy) to calculate the similarity between the answers, and displays the differences using color coding or graphs based on the similarity. Specifically, the server calculates the similarity between the answers of generation AI A and generation AI B, and displays them in the same color to indicate high similarity.

[1225] Step 6:

[1226] The user specifies the usage percentage of each answer as input. The input is the usage percentage specified by the user, and the output is the hybrid answer. The server combines each answer based on the percentage and generates the hybrid answer. In concrete terms, the user specifies "answer A will be used 70%, answer B will be used 20%, and answer C will be used 10%", and the server generates the following hybrid answer:

[1227] Hybrid answer: "AI will continue to evolve and will support people's lives in many areas, including medicine, education, and entertainment."

[1228] (Application example 2)

[1229] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1230] In conventional content generation systems using generative AI models, there was a lack of a way to effectively compare and integrate responses from multiple generative AIs, making it difficult for users to generate optimal content. In addition, there was also an issue of efficiency when applying the generated content to content distribution services.

[1231] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1232] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to each generation AI, a means for displaying answers from each generation AI in a list format, a means for visually indicating the difference between each answer, a means for generating a hybrid answer by combining each answer in a ratio specified by the user, and a means for applying the generated hybrid answer to a content distribution service. This enables a user to effectively compare and integrate answers from multiple generation AIs to generate optimal content. In addition, the generated content can be quickly and efficiently applied to a content distribution service.

[1233] The "question input interface" is an interface through which users can input questions to the generation AI.

[1234] "Generative AI" is a system that uses artificial intelligence technology to generate text and content.

[1235] An "answer" is the text or content that a generative AI generates in response to a question.

[1236] The "list format" is a format in which multiple answers are displayed side by side on one screen.

[1237] A "means of visually showing the differences" is a means of displaying the differences in the answers from each generation AI in a visual manner, such as color coding or graphs.

[1238] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs in proportions specified by the user.

[1239] A "content distribution service" is a service that provides generated content to users via the Internet.

[1240] A system for implementing this invention has the following configuration. The server has a means for receiving questions from users through a question input interface and sending the questions to each generation AI. Specifically, the user inputs a question using a smartphone or a head-mounted display. For example, the user inputs a prompt such as "Please tell me some ideas for a new blog post."

[1241] When the server receives a question, it sends the question to multiple generation AIs (e.g., GPT-4, BERT, T5, etc.). Each generation AI generates an answer to the question and sends it back to the server. The server has a means to display the answers from each generation AI in a list format. This allows the user to compare the answers from each generation AI at a glance.

[1242] Furthermore, the server has a means for visually indicating the differences between the answers. Specifically, the server calculates the similarity between the answers and displays the differences using colors and graphs. This allows the user to intuitively understand the differences between the answers.

[1243] The user can choose the answer they like from each answer and specify the percentage. For example, they can specify "70% of the answers from GPT-4, 20% of the answers from BERT, and 10% of the answers from T5". The server has a means to combine each answer based on these percentages and generate a hybrid answer. The generated hybrid answer is provided to the user.

[1244] Finally, the server has means for applying the generated hybrid answer to a content delivery service, thereby allowing users to quickly and efficiently deliver the generated content.

[1245] As a concrete example, when a user enters a question such as "Please tell me some ideas for new blog articles," the answers from each generative AI are displayed in a list, and the differences in the answers are visually shown. A hybrid answer is generated based on the ratio specified by the user, and the answer is applied to the content distribution service.

[1246] Example prompt:

[1247] "Can you give me some ideas for new blog posts?"

[1248] The flow of the specific process in the application example 2 will be described with reference to FIG.

[1249] Step 1:

[1250] The user inputs a question using a terminal. Specifically, the user inputs a prompt sentence through a question input interface on a smartphone or a head-mounted display. The input question is then sent to the server.

[1251] Input: The question entered by the user (prompt)

[1252] Output: The question sent to the server.

[1253] Step 2:

[1254] The server sends the received questions to multiple generative AI models. Specifically, it sends questions to generative AI models such as GPT-4, BERT, and T5 through an API. Each generative AI model generates an answer to the question and sends it back to the server.

[1255] Input: The question sent to the server

[1256] Output: The answer from each generative AI model

[1257] Step 3:

[1258] The server displays the answers from each generative AI model in a list format. Specifically, the name of each generative AI model and its answer are displayed side by side on the screen. This allows the user to compare the answers from each generative AI model at a glance.

[1259] Input: The answer from each generative AI model

[1260] Output: Answers displayed in a list format

[1261] Step 4:

[1262] The server visually indicates the differences between answers. Specifically, it calculates the similarity between answers and displays the differences using colors and graphs. This allows users to intuitively understand the differences between answers.

[1263] Input: The answer from each generative AI model

[1264] Output: Visually indicated difference in answers

[1265] Step 5:

[1266] Users can choose their favorite answer from each answer and specify the percentage. Specifically, users can specify the percentages such as "70% of GPT-4's answers, 20% of BERT's answers, and 10% of T5's answers."

[1267] Input: Percentage of user-specified answers

[1268] Output: Specified percentage

[1269] Step 6:

[1270] The server combines the answers based on the specified percentages to generate a hybrid answer, that is, it combines the text of each answer in the specified percentages to generate a new answer.

[1271] Input: Answers from each generative AI model, in user-specified proportions

[1272] Output: Hybrid answer

[1273] Step 7:

[1274] The server applies the generated hybrid answer to a content delivery service, specifically, providing the generated content to users over the Internet.

[1275] Input: Hybrid answer

[1276] Output: Answers applied to content delivery services

[1277] In addition, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1278] "Example 1"

[1279] As one embodiment of the present invention, a system including a generation AI incorporating an emotion engine is provided. This system asks a question from a user to a plurality of generation AIs, and displays a list of answers from each generation AI. At least one of the generation AIs includes an emotion engine that recognizes the user's emotion. The emotion engine estimates the emotion from the user's tone of voice, facial expression, choice of words, etc. Then, the answer to the question is adjusted based on the estimation result. For example, if the user feels angry, the emotion engine conveys that information to the generation AI, and the generation AI adjusts to generate a more polite answer. Also, if the user feels happy, the generation AI adjusts to generate a more enjoyable answer. In this way, the emotion engine plays a role in optimizing the answer of the generation AI according to the user's emotion.

[1280] "Example 2"

[1281] Furthermore, as another embodiment of the present invention, a system is provided that generates hybrid answers by taking into account the output of an emotion engine. Specifically, when a user specifies the usage ratio of each generation AI's answer, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine conveys that information to the system, and the system generates hybrid answers that include more fun answers. In this way, by taking into account the output of the emotion engine, it is possible to generate hybrid answers that are more adapted to the user's emotions.

[1282] The process flow of each embodiment will be described below.

[1283] "Example 1"

[1284] Step 1: Receive a question from the user.

[1285] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[1286] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[1287] Step 4: Adjust the responses of each generation AI based on the estimation results. For example, if the user feels angry, the emotion engine will convey that information to the generation AI, and the generation AI will adjust to generate more polite responses.

[1288] Step 5: List the answers from each generation AI.

[1289] "Example 2"

[1290] Step 1: Receive a question from the user.

[1291] Step 2: The received question is posed to multiple generative AIs, at least one of which includes an emotion engine that recognizes the user's emotions.

[1292] Step 3: The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc.

[1293] Step 4: When the user specifies the proportion of each AI response to be used, the output from the emotion engine is also taken into account. For example, if the user feels happy, the emotion engine will convey that information to the system, and the system will generate hybrid responses that include more happy responses.

[1294] Step 5: Provide the hybrid answer to the user.

[1295] Example 1

[1296] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1297] In conventional generative AI systems, it is difficult to visually recognize the differences when comparing answers from multiple generative AIs. In addition, the system lacks the functionality to adjust the generative AI's answers according to the user's emotions, making it difficult to increase user satisfaction. Furthermore, there is no means to combine answers from multiple generative AIs to generate the optimal answer, making it difficult to provide users with the most optimal information.

[1298] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1299] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the differences, a means for combining each answer based on a ratio specified by the user to generate a hybrid answer, and a means for recognizing the user's emotions and adjusting the answer of the generation AI based on the emotions. This makes it possible to visually compare answers from multiple generation AIs and provide the optimal answer according to the user's emotions.

[1300] A "question input interface" is a user interface for users to input questions, and is implemented using a web browser or a dedicated application.

[1301] "Generative AI" is an artificial intelligence model that uses natural language processing technology to generate answers to input questions, and examples include GPT-4 and BERT.

[1302] The "list format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen, and the name of each generation AI and its answer are arranged so that they can be visually recognized.

[1303] "Visually showing the differences" means displaying the differences in answers from multiple generation AIs using visual means such as color coding or graphs, and calculating the similarity between answers and visually expressing the result.

[1304] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user, providing information optimized to the user's needs.

[1305] The "emotion engine" is an engine that has the ability to infer emotions from the user's tone of voice, facial expressions, choice of words, etc., and adjust the responses of the generative AI based on these inferred results.

[1306] "Recognizing the user's emotions" means using an emotion engine to estimate the user's emotional state and conveying that information to the generative AI.

[1307] This invention is a system that asks the same question to multiple AI generators, displays the answers, visually recognizes the differences, and generates hybrid answers in the ratio specified by the user. It also includes a function to recognize the user's emotions and adjust the AI ​​generator's answers based on those emotions.

[1308] System Configuration

[1309] Question input interface

[1310] A user inputs a question through a question input interface. This interface is implemented using a web browser or a dedicated application. For example, a user inputs "What's the weather like today?"

[1311] Submit a question

[1312] The server sends the question received from the user to multiple generative AIs. Specifically, the server formats the question as a prompt sentence and sends it to a generative AI model (e.g., GPT-4 or BERT). An example of a prompt sentence is "Please provide the weather forecast for today."

[1313] Receiving and listing responses

[1314] The server receives the answers from each AI generator and displays them in a list. The name of each AI generator and its answer are displayed side by side on the terminal screen. For example, it is displayed as follows:

[1315] Generative AI model 1: "The weather is sunny today."

[1316] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[1317] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[1318] Recognizing differences in answers

[1319] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[1320] Hybrid Answer Generation

[1321] The user specifies the usage ratio of each answer. The server combines each answer based on that ratio to generate a hybrid answer. For example, if the user specifies "50% of the answers from Generative AI Model 1, 30% of the answers from Generative AI Model 2, and 20% of the answers from Generative AI Model 3," the server combines these answers to generate a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[1322] Using the Emotion Engine

[1323] The server uses an emotion engine to recognize the user's emotions. The emotion engine infers emotions from the user's tone of voice, facial expressions, choice of words, etc. Based on the inference results, the generation AI adjusts the answer. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with."

[1324] Examples

[1325] The user enters "What's the weather like today?" into the question input interface and clicks the send button. The server sends this question to the generation AI as a prompt sentence, "Please provide the weather forecast for today." The server receives the answer from the generation AI and displays it in a list format on the device screen. The server calculates the similarity of each answer and displays them in different colors. The user specifies the usage ratio of each answer, and the server generates a hybrid answer based on that ratio. In addition, an emotion engine is used to recognize the user's emotions and adjust the answer of the generation AI.

[1326] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1327] Step 1:

[1328] A user inputs a question through the question input interface. Specifically, the user inputs "What's the weather like today?" and clicks the send button. The input question is sent to the server. The input data is the question in text format, and the output is the question data sent to the server.

[1329] Step 2:

[1330] The server sends the questions received from the user to multiple generation AIs. Specifically, the server formats the questions as prompt sentences and sends them to a generation AI model (e.g., GPT-4 or BERT). The input data is the question text from the user, and the output is the prompt sentence sent to the generation AI. For example, the server generates a prompt sentence such as "Please provide the weather forecast for today."

[1331] Step 3:

[1332] The server receives the answers from each generation AI. The responses from the generation AI are returned to the server in a data format such as JSON. The input data is the answer data from the generation AI, and the output is the answer data received by the server. For example, GPT-4 returns the answer "Today's weather is sunny."

[1333] Step 4:

[1334] The server displays the received answers in a list format. The name of each AI generator and its answer are displayed side by side on the terminal screen. The input data are the answer data from each AI generator, and the output is the answer displayed on the terminal in a list format. For example, it is displayed as follows:

[1335] Generative AI model 1: "The weather is sunny today."

[1336] Generative AI model 2: "It will be cloudy with occasional sunny spells today."

[1337] Generative AI model 3: "It's sunny today, but it might get a little cloudy in the afternoon."

[1338] Step 5:

[1339] The server compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using colors and graphs based on the similarity. The input data is the answer data from each generation AI, and the output is the visually displayed difference information. For example, the answers from generation AI model 1 and generation AI model 2 are highly similar, so they are displayed in the same color.

[1340] Step 6:

[1341] The user specifies the usage ratio of each answer. Specifically, the user specifies "50% of answers from AI model 1, 30% of answers from AI model 2, and 20% of answers from AI model 3." The input data is the usage ratio specified by the user, and the output is the ratio data sent to the server.

[1342] Step 7:

[1343] The server combines each answer based on the ratio and generates a hybrid answer. The input data is the usage ratio specified by the user and the answer data from each generation AI, and the output is the generated hybrid answer. For example, it generates a hybrid answer such as "It's sunny today, but it may be a little cloudy in the afternoon."

[1344] Step 8:

[1345] The server recognizes the user's emotions using an emotion engine. The emotion engine estimates emotions from the user's tone of voice, facial expressions, choice of words, etc. The input data is the user's voice and video data, and the output is estimated emotion information.

[1346] Step 9:

[1347] The server adjusts the answer of the generation AI based on the estimation result. For example, if the emotion engine recognizes that the user is angry, the generation AI generates a more polite answer such as "It's sunny today. Please let me know if there is anything I can help you with." The input data is the estimated emotion information and the answer data from the generation AI, and the output is the adjusted answer data.

[1348] (Application example 1)

[1349] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1350] In conventional generative AI systems, when comparing answers from multiple generative AIs, adjustments were not made to take the user's emotions into account, making it difficult to obtain the optimal answer according to the user's emotions. In addition, when combining answers from generative AIs, it was not possible to reflect the user's emotions, making it difficult to increase user satisfaction.

[1351] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1352] In this invention, the server includes a means for asking the same question to multiple generation AIs, a means for displaying a list of answers from each generation AI, a means for recognizing differences between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for recognizing the user's emotions, and a means for adjusting the answer of the generation AI based on the recognized emotions. This makes it possible to provide an optimal answer according to the user's emotions and increase user satisfaction.

[1353] "Generative AI" is a system that uses artificial intelligence technology to automatically generate answers to questions from users.

[1354] A "means for asking a question" is a device or program that has the function of sending questions from users to multiple generation AIs.

[1355] A "means for displaying a list of answers" is a device or program that has the function of displaying answers received from multiple generation AIs side-by-side on a single screen.

[1356] A "means for recognizing differences" is a device or program that has the function of comparing answers from multiple generation AIs and visually showing the differences.

[1357] A "means for regenerating a hybrid answer" is a device or program that has the function of combining answers from multiple generation AIs based on a ratio specified by the user to generate a new answer.

[1358] A "means for recognizing emotions" is a device or program that has the function of inferring emotions from a user's tone of voice, facial expressions, choice of words, etc.

[1359] A "means for adjusting an answer" is a device or program that has the function of optimizing the answer of the generating AI based on the recognized user's emotions.

[1360] The system for implementing this invention has the functions of asking the same question to multiple AI generators, displaying a list of answers from each AI generator, recognizing the differences between each answer, and regenerating a hybrid answer by combining each answer in a ratio specified by the user. It also includes a function of recognizing the user's emotions and adjusting the AI ​​generator's answer based on those emotions.

[1361] System Configuration

[1362] The system consists of the following main components:

[1363] 1. Question input interface: This is an interface for users to input questions. This is implemented on devices such as smartphones and computers.

[1364] 2. Generative AI module: A module that includes multiple generative AIs. Each generative AI automatically generates answers to questions posed by users.

[1365] 3. Answer display module: This module displays a list of answers from each generation AI. It displays the differences in answers in a format that allows visual recognition.

[1366] 4. Emotion recognition module: This module estimates emotions from the user’s tone of voice, facial expressions, choice of words, etc.

[1367] 5. Answer Adjustment Module: This module adjusts the generative AI’s answers based on the recognized emotions.

[1368] 6. Hybrid answer generation module: This module combines answers from each generation AI based on the ratio specified by the user to generate a hybrid answer.

[1369] Program processing explanation

[1370] The server receives questions from users through a question input interface. These questions are sent to the generation AI module, and multiple generation AIs generate their own answers. The generated answers are displayed in a list by the answer display module. At this time, the differences between each answer are displayed in a format that allows visual recognition.

[1371] Next, the emotion recognition module recognizes the user's emotions. For example, it estimates emotions from the user's tone of voice and facial expressions. This recognized emotion information is sent to the answer adjustment module and used to adjust the answer of the generation AI. The adjusted answer is then displayed again in the answer display module.

[1372] Finally, based on the ratio specified by the user, the hybrid answer generation module combines the answers from each generation AI to generate a hybrid answer, which is finally provided to the user.

[1373] Examples

[1374] For example, if a user asks, "What movie would you recommend to watch tonight?", each generative AI will suggest its own movie recommendation. If the emotion recognition module recognizes the user's emotions and determines that the user is tired, for example, the answer will be adjusted to suggest a relaxing movie. Finally, a hybrid recommendation list is generated that combines answers from multiple generative AIs based on the ratio specified by the user.

[1375] Example prompt:

[1376] What movie would you recommend to watch tonight?

[1377] In this way, it is possible to provide the user with an optimal answer according to his / her feelings, thereby increasing the user's satisfaction.

[1378] The flow of the specific process in the application example 1 will be described with reference to FIG.

[1379] Step 1:

[1380] A user inputs a question into the question input interface using a terminal. The input question is transmitted to the server. The input data is a question about the content that the user wants to view.

[1381] Step 2:

[1382] The server sends the received questions to multiple generation AIs. Each generation AI automatically generates an answer to the question. The generated answer is sent back to the server. The input data is the user's question, and the output data is the answer from each generation AI.

[1383] Step 3:

[1384] The server sends the answers from each AI generator to the answer display module and displays them in a list. The answer display module displays the differences between the answers in a format that allows the differences to be visually recognized. The input data are the answers from each AI generator, and the output data is the visually displayed list of answers.

[1385] Step 4:

[1386] When a user inputs a question into the question input interface using a terminal, the emotion recognition module analyzes the user's tone of voice and facial expression to recognize the emotion. The input data is the user's tone of voice and facial expression, and the output data is the recognized emotion information.

[1387] Step 5:

[1388] The server sends the recognized emotion information to the answer adjustment module. The answer adjustment module adjusts the answer of each generation AI based on the emotion information. For example, if the user is recognized as tired, the answer is adjusted to suggest relaxing content. The input data are the recognized emotion information and the answer from each generation AI, and the output data are the adjusted answer.

[1389] Step 6:

[1390] The server again transmits the adjusted answers to the answer display module for displaying them in a list. The input data is the adjusted answers, and the output data is a visually displayed list of adjusted answers.

[1391] Step 7:

[1392] The user specifies the usage ratio of each answer using the terminal. For example, the user specifies "50% of the answers from AI1, 30% of the answers from AI2, and 20% of the answers from AI3." The input data is the usage ratio specified by the user.

[1393] Step 8:

[1394] The server generates a hybrid answer by combining the answers from each generation AI in the hybrid answer generation module based on the ratio specified by the user. The input data is the usage ratio specified by the user and the answers from each generation AI, and the output data is the hybrid answer.

[1395] Step 9:

[1396] The server provides the generated hybrid answer to the user. The user uses a terminal to confirm the hybrid answer. The input data is the hybrid answer, and the output data is the final answer provided to the user.

[1397] Example 2

[1398] Next, a second embodiment of the second form will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1399] In conventional generative AI systems, it was difficult to compare answers from multiple generative AIs and combine them to generate the optimal answer. In addition, the system did not generate answers that took the user's feelings into account, so it was not possible to increase user satisfaction. This meant that users could not efficiently obtain the information they were looking for, and the system's value as a user decreased.

[1400] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1401] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing answers from each generation AI and visually showing the difference, a means for combining answers from each generation AI in a ratio specified by the user to generate a hybrid answer, and a means for adjusting the hybrid answer by taking into account the output of the emotion analysis engine. This allows the user to compare answers from multiple generation AIs and obtain the optimal answer. Furthermore, the generation of an answer that takes into account the user's emotions can increase the user's satisfaction.

[1402] A "question input interface" is a user interface that allows a user to input a question and send the question to the system.

[1403] "Generative AI" is an artificial intelligence model that generates answers in natural language in response to a given question.

[1404] The "list format" is a format in which a plurality of items are visually organized and displayed, making it easier for the user to compare the items.

[1405] "Visually showing" means displaying information using visual means such as color coding or graphs so that the user can understand it intuitively.

[1406] A "hybrid answer" is a new answer generated by combining answers from multiple generative AIs, incorporating the characteristics of each answer.

[1407] An "emotion analysis engine" is an engine that analyzes emotions from user input and actions and provides the results to the system.

[1408] A "weighted average" is a method of calculating an average by assigning different weights to each piece of data, and is used to emphasize the influence of certain data.

[1409] An "API Request" is a request made to another system or service through an Application Program Interface, which is a means of sending or receiving data.

[1410] The present invention is a system for generating optimal answers to questions from users using multiple generative AI models. A specific embodiment of this system will be described below.

[1411] The server receives questions from users through a question input interface. This interface is implemented using a web browser or a dedicated application. Specifically, a browser such as Google Chrome can be used. For example, the user inputs a question such as, "Please tell me about the future of AI technology."

[1412] The server sends the received questions to multiple generation AIs. OpenAI's GPT-3 and Google's BERT are used as generation AIs. Each generation AI generates an answer to the question and sends it back to the server. Specifically, the question is sent to each generation AI's endpoint using an HTTP POST request.

[1413] The server receives the answers from each AI and displays them in a list format. This display is implemented on a web page using HTML and CSS. For example, the name of each AI and its answer are displayed side by side on the screen in the form of "Answer from AI model A" and "Answer from AI model B."

[1414] The server compares the answers from each generation AI and visually displays the differences. Specifically, it uses the Python library scikit-learn to calculate the similarity between answers, and displays the differences using color coding and graphs based on the similarity. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red.

[1415] Users can specify the proportion of each answer to be used. The server receives input from the user and combines each answer based on that proportion to generate a hybrid answer. Specifically, it uses Python's NumPy library to weight and average each answer by the specified proportion to generate a new answer. For example, if a user specifies "70% of answers from Generative AI Model A and 30% from Generative AI Model B," each answer will be combined according to that proportion.

[1416] Furthermore, by using a sentiment analysis engine, it is possible to generate hybrid answers that adapt to the user's emotions. As a sentiment analysis engine, IBM's Watson Tone Analyzer is used. The sentiment analysis engine analyzes the user's input and actions to determine emotions and sends this information to the server. For example, if the sentiment analysis engine determines that the user is happy, the server adjusts the answer to include more happy content.

[1417] In this way, users can compare answers from multiple AI generators to obtain the best answer. In addition, by generating answers that take the user's emotions into account, user satisfaction can be increased.

[1418] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1419] Step 1:

[1420] The user enters a question.

[1421] A user accesses the question input interface through a web browser or a dedicated application and inputs a question. For example, a user accesses a web page using Google Chrome and inputs "Please tell me about the future of AI technology" in the text box. The input question is sent to the server.

[1422] Step 2:

[1423] The server receives the question and sends it to each generating AI.

[1424] The server sends the question received from the user to multiple generation AIs. Specifically, it uses an HTTP POST request to send the question to the endpoint of each generation AI. OpenAI's GPT-3 and Google's BERT are used as generation AIs. The input is the user's question, and the output is the question sent to each generation AI.

[1425] Step 3:

[1426] Each generation AI generates an answer to the question and sends it back to the server.

[1427] Each generation AI generates an answer to the question it receives. For example, GPT-3 generates the answer "AI technology will continue to evolve in the future and will be used in many fields such as medicine and education," and BERT generates the answer "AI technology will contribute to the realization of autonomous driving and smart cities." These answers are sent back to the server as API responses. The input is the question sent to each generation AI, and the output is the generated answer.

[1428] Step 4:

[1429] The server displays a list of answers from each generation AI.

[1430] The server uses HTML and CSS to display the answers received from each generation AI in a list format on a web page. For example, the name of each generation AI and its answer are displayed side by side on the screen in the form of "Answer from generation AI model A" and "Answer from generation AI model B." The input is the answer from each generation AI, and the output is the answer displayed in a list format.

[1431] Step 5:

[1432] The server visually indicates the difference between each answer.

[1433] The server uses Python's scikit-learn library to calculate the similarity between each answer. Based on the calculation results, the similarity of the answers is visually displayed using color coding or graphs. For example, answers with high similarity are displayed in green, and answers with low similarity are displayed in red. The input is the answer from each generation AI, and the output is the difference between the visually displayed answers.

[1434] Step 6:

[1435] The user specifies the usage ratio of each answer.

[1436] The user accesses an interface to specify the usage ratio of each answer. For example, the user can use a slider or text box to input "70% of answers from AI model A and 30% of answers from AI model B." The input is the usage ratio specified by the user, and the output is the usage ratio sent to the server.

[1437] Step 7:

[1438] The server generates a hybrid answer based on the specified percentages.

[1439] The server combines each answer based on the ratio specified by the user. Specifically, it uses Python's NumPy library to weight and average each answer to generate a new hybrid answer. For example, it generates an answer such as "AI technology will continue to evolve in the future, contributing to the realization of healthcare, education, autonomous driving, and smart cities." The input is the usage ratio specified by the user and the answer from each generation AI, and the output is the generated hybrid answer.

[1440] Step 8:

[1441] The emotion analysis engine analyzes the user's emotions and transmits them to the server.

[1442] Sentiment analysis engines analyze user input and behavior to determine emotions. For example, IBM's Watson Tone Analyzer can be used to analyze emotions from a user's text input and determine that "the user is happy." This information is sent to a server. The input is the user's input and behavior, and the output is the analyzed emotional information.

[1443] Step 9:

[1444] The server adjusts the hybrid answer taking into account the output of the sentiment analysis engine.

[1445] The server adjusts the hybrid answer taking into account the output from the sentiment analysis engine. For example, if it is determined that the user is happy, the server adjusts the answer to include more fun content. Specifically, it generates an answer that emphasizes the fun element. The input is the emotion information from the sentiment analysis engine and the hybrid answer, and the output is the adjusted hybrid answer.

[1446] (Application example 2)

[1447] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1448] In systems using conventional generative AI models, when comparing answers from multiple generative AI models, it was difficult to visually recognize the differences in answers, and it was not possible to generate hybrid answers that adapted to the user's emotions. As a result, it was necessary for users to make a great deal of effort to obtain the optimal answer, making it difficult to obtain information efficiently. Furthermore, since it was not possible to take the user's emotions into account when combining answers from generative AI models, it was not possible to increase user satisfaction.

[1449] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1450] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for regenerating a hybrid answer by combining each answer in a ratio specified by a user, a means for generating a hybrid answer taking into account the output of the emotion engine, and a means for calculating the similarity between answers and visually displaying the difference. This allows a user to efficiently compare answers from a plurality of generative AI models and visually recognize the difference. In addition, by generating a hybrid answer that takes into account the user's emotions, the user's satisfaction can be increased.

[1451] A "generative AI model" is an algorithm or system that uses artificial intelligence to generate content such as text and images.

[1452] A "means for asking questions" is an interface that has the ability to receive questions from users and send them to a generative AI model.

[1453] A "means for displaying answers in a list" is an interface with the function of displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[1454] A "means of recognizing differences" is a system that has the ability to compare answers from multiple generative AI models and visually show the differences.

[1455] A "means for regenerating hybrid answers" is a system that has the ability to generate new answers by combining answers from multiple generative AI models based on a ratio specified by the user.

[1456] An "emotion engine" is an algorithm or system that analyzes user emotions and adjusts the system's operation based on the results.

[1457] "Means for calculating similarity" is a system capable of numerically evaluating the similarity between answers from multiple generative AI models.

[1458] A "means for visually displaying differences" is a system that has the ability to visually display the differences between responses using color coding, graphs, or other means.

[1459] A system for implementing the present invention includes a server, a user terminal, and a generative AI model. The server has a function of receiving a question from a user and transmitting the question to a plurality of generative AI models. The user terminal provides a question input interface, allowing the user to input a question. The generative AI model generates an answer to the received question and transmits it back to the server.

[1460] The server receives answers from multiple generative AI models and displays them in a list. Specifically, the name of each generative AI model and its answer are displayed side by side. In addition, the server recognizes the differences between each answer and calculates the similarity between the answers. Based on the similarity, the differences between the answers are visually displayed using color coding and graphs.

[1461] The user can check the answers from each generative AI model and specify the proportion of each answer to be used. The server combines each answer based on the proportion specified by the user to generate a hybrid answer. At this time, the emotion engine analyzes the user's emotions and adjusts the hybrid answer based on the results. For example, if the emotion engine determines that the user is happy, it will adjust the hybrid answer to include more cheerful answers.

[1462] As a concrete example, if a user enters the question "What is the best travel destination for our next vacation?", the server sends this question to multiple generative AI models. The answers from each generative AI model are displayed in a list, and the differences between the answers are visually shown. The user specifies the proportion of each answer to be used, and a hybrid answer is generated based on the analysis results of the emotion engine.

[1463] The hardware used includes user devices (e.g., smartphones, smart glasses, head-mounted displays, etc.) and servers. The software used includes generative AI models (e.g., OpenAI APIs), emotion engines (e.g., TextBlob), and data visualization tools (e.g., Matplotlib).

[1464] The following are examples of prompt sentences:

[1465] "What's your recommended destination for our next vacation?"

[1466] The flow of the specific process in the application example 2 will be described with reference to FIG.

[1467] Step 1:

[1468] The user uses the terminal to input a question.

[1469] Input: The prompt text entered by the user (e.g. "What are some recommended travel destinations for our next vacation?")

[1470] Output: The entered question is sent to the server.

[1471] Specific operation: The user inputs a question using the question input interface of the terminal and presses the send button. The terminal sends the input question to the server.

[1472] Step 2:

[1473] The server receives the question and sends it to multiple generative AI models.

[1474] Input: The question submitted by the user

[1475] Output: The question sent to the generative AI model

[1476] How it works: The server sends the received question to multiple generative AI models, each of which receives the question and prepares to generate an answer.

[1477] Step 3:

[1478] The generative AI model generates answers to the questions and sends them back to the server.

[1479] Input: The question sent to the generative AI model

[1480] Output: The answer returned by the generative AI model.

[1481] Specific operation: Each generative AI model generates an answer based on the question and sends the answer back to the server. The server receives the answer from each generative AI model.

[1482] Step 4:

[1483] The server lists the answers from each generative AI model.

[1484] Input: The answer returned by the generative AI model

[1485] Output: List of answers

[1486] Specific operation: The server displays the answers from each generative AI model in a list format on the screen. The name of each generative AI model and its answer are displayed side by side.

[1487] Step 5:

[1488] The server recognizes the difference between each answer and displays it visually.

[1489] Input: Answer from a generative AI model

[1490] Output: Visually displayed difference in answers

[1491] Specific operation: The server calculates the similarity of each answer and visually displays the differences between the answers using color coding and graphs, allowing users to recognize the differences between each answer at a glance.

[1492] Step 6:

[1493] The user specifies the usage ratio of each answer.

[1494] Input: The percentage of each answer that the user specifies to be used (e.g. "70% of the time answer A, 20% of the time answer B, 10% of the time answer C")

[1495] Output: The specified usage percentage is sent to the server.

[1496] Specific operation: The user uses the terminal interface to specify the usage ratio of each answer and presses the send button. The terminal then sends the specified usage ratio to the server.

[1497] Step 7:

[1498] The server generates a hybrid answer based on the specified percentages.

[1499] Input: Usage rate of each answer specified by the user, answers from the generative AI model

[1500] Output: Hybrid answer

[1501] Specific behavior: The server combines each answer based on the user-specified percentage to generate a hybrid answer, which is then displayed to the user.

[1502] Step 8:

[1503] The server uses an emotion engine to analyze the user's emotions and tailor the hybrid response.

[1504] Input: Hybrid answers, user sentiment data

[1505] Output: A tailored hybrid answer

[1506] Specific operation: The server uses the emotion engine to analyze the user's emotions and adjusts the hybrid answers based on the results. For example, if the emotion engine determines that the user is happy, the answers will be adjusted to include more happy answers. The adjusted hybrid answers are finally displayed to the user.

[1507] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input for the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1508] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1509] Another example of generative AI is Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) are mentioned.

[1510] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1511] [Fourth embodiment]

[1512] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1513] 7, a 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.

[1514] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[1515] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. 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. In addition, the microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1516] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[1517] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[1518] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[1519] The control target 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, legs, etc. The posture and behavior of the robot 414 are controlled by controlling the motors of the arms, hands, legs, etc. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1520] Fig. 8 shows an example of main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1521] The specific processing program 56 is an example of a "program" according to the technology of the present 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 in accordance with the specific processing program 56 executed on the RAM 30.

[1522] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1523] In the robot 414, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1524] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1525] "Example 1"

[1526] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[1527] "Example 2"

[1528] As one embodiment of the present invention, the means for asking the same question to a plurality of generation AIs has a function of receiving a question from a user through a question input interface and sending the question to each generation AI. The means for displaying a list of answers from each generation AI has a function of receiving answers from each generation AI and displaying them in a list format. Specifically, the names of each generation AI and their answers are displayed side by side on the screen. The means for recognizing the differences between each answer has a function of comparing answers from each generation AI and visually indicating the differences. Specifically, the similarity between the answers is calculated, and the differences are displayed using color coding or graphs based on the similarity. The means for regenerating a hybrid answer by combining each answer in a ratio specified by the user has a function of receiving the usage ratio of each answer from the user as an input, combining each answer based on that ratio, and generating a hybrid answer. Specifically, if the user specifies "70% of answer A, 20% of answer B, and 10% of answer C," the answers are combined according to that ratio.

[1529] The process flow of each embodiment will be described below.

[1530] "Example 1"

[1531] Step 1: The user inputs a question through the question input interface. This question is then posed to multiple generation AIs.

[1532] Step 2: The system sends a question to each Generator AI, which answers the question using its own algorithm.

[1533] Step 3: The system receives answers from each AI generator. These answers are displayed in a list format. Specifically, the name of each AI generator and its answer are displayed side by side on the screen.

[1534] Step 4: The system compares the answers from each generation AI and visually indicates the differences. Specifically, it calculates the similarity between the answers and displays the differences using color coding and graphs based on the similarity.

[1535] Step 5: The user specifies the percentage of each answer to be used. For example, "Answer A will be 70%, Answer B will be 20%, and Answer C will be 10%."

[1536] Step 6: The system combines the answers based on the percentages specified by the user to generate a hybrid answer, which becomes the final output.

[1537] Example 1

[1538] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1539] Conventional generative AI systems rely on answers from a single generative AI, which limits the variety and accuracy of answers. In addition, there is no way to compare answers from multiple generative AIs, making it difficult for users to select the most appropriate answer. Furthermore, there is a lack of functionality to combine multiple answers to generate hybrid answers, which means that it is not possible to flexibly generate answers that meet the needs of users.

[1540] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1541] In this invention, the server includes a means for receiving a question from a user through a question input interface and sending the question to multiple generation AIs, a means for receiving answers from each generation AI and displaying them in a list format, a means for comparing the answers from each generation AI and visually showing the differences, and a means for regenerating a hybrid answer by combining the answers from each generation AI based on a ratio specified by the user. This makes it possible to compare various answers from multiple generation AIs and select the optimal answer. It also makes it possible to flexibly generate hybrid answers according to the needs of the user.

[1542] The "question input interface" is a user interface that allows a user to input a question and transmit the question to the server.

[1543] "Generative AI" is an artificial intelligence model that generates natural language responses based on a given prompt.

[1544] An "API endpoint" is an interface for sending questions to the generation AI and receiving answers.

[1545] "List format" is a format in which answers from multiple generation AIs are displayed side by side on a single screen.

[1546] "Visually showing" means showing the differences in responses using visual means such as color coding or graphs.

[1547] "Similarity" is an index showing the degree of agreement between the contents of multiple answers.

[1548] A "hybrid answer" is a new answer generated by combining answers from multiple generation AIs based on the ratio specified by the user.

[1549] The "proportion" is the ratio that indicates the extent to which the user will use the answers from each generation AI.

[1550] This invention is a system that asks the same question to multiple generative AI models and compares and integrates their answers. A specific embodiment of this system will be described below.

[1551] Question input interface

[1552] The user accesses the question input interface using a web browser on the terminal. This interface is built using HTML and JavaScript, and provides a form where the user can enter a question in text format. Once the user enters a question and clicks the "Submit" button, the question is sent to the server.

[1553] Submit a question

[1554] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt. The generative AI models used include, for example, natural language processing models and machine learning models.

[1555] Receiving and listing responses

[1556] The server receives the answers from each generative AI model and temporarily stores them. Next, the server generates a web page to display the received answers in a list format. This web page displays the name of each generative AI model and its answer side by side. The user can view this page in the web browser of the terminal.

[1557] Recognizing differences in answers

[1558] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity. This process is carried out using text analysis algorithms and data visualization tools.

[1559] Regenerating a hybrid answer

[1560] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[1561] Examples

[1562] Example of question input

[1563] The user enters the following question into the question entry interface:

[1564] "What do you think the future of AI is?"

[1565] Examples of prompt statements

[1566] Here is an example of a prompt to be input to the generative AI model:

[1567] "What do you think the future of AI is?"

[1568] Example of answer list display

[1569] Generative AI Model 1: “AI will play an important role in many fields in the future, including medicine and education.”

[1570] Generative AI Model 2: "AI will help make autonomous driving and smart cities a reality."

[1571] Generative AI Model 3: “AI will continue to evolve as a tool to make human life more convenient.”

[1572] Examples of visual display of response differences

[1573] The parts of the answers from Generative AI Model 1 and Generative AI Model 2 that are highly similar are displayed in green.

[1574] The areas where the answers from Generative AI Model 1 and Generative AI Model 3 have low similarity are displayed in red.

[1575] A concrete example of hybrid answer regeneration

[1576] The user specifies "70% of the answers from Generative AI Model 1, 20% from Generative AI Model 2, and 10% from Generative AI Model 3." The server combines each answer based on that ratio and generates a hybrid answer like this:

[1577] "Going forward, AI will play an important role in many fields, including healthcare and education, and will also contribute to the realization of autonomous driving and smart cities, continuing to evolve as a tool to make human life more convenient."

[1578] In this way, users can compare various answers from multiple generative AI models and select the optimal answer. It also enables flexible generation of hybrid answers according to the needs of users.

[1579] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1580] Step 1: Enter your question

[1581] The user opens a web browser on the terminal and accesses the question input interface. The user enters a question in the text box and clicks the "Send" button. This operation sends the question to the server.

[1582] Input: A question typed in by the user (e.g., "Tell me about the future of AI.")

[1583] Output: The question data sent to the server

[1584] Step 2: Submit your question

[1585] The server sends the question received from the user to multiple generative AI models. The server sends an HTTP request to the API endpoint of each generative AI model, passing the question as a prompt.

[1586] Input: The query data sent to the server

[1587] Output: A prompt sent to each generative AI model (e.g., "Tell me about the future of AI.")

[1588] Step 3: Receiving a response

[1589] The server receives answers from each generative AI model. Each generative AI model generates an answer based on the prompt sentence and returns it to the server. The server temporarily stores these answers.

[1590] Input: Answer data from each generative AI model

[1591] Output: Answer data stored on the server (e.g., "AI will play an important role in many fields, such as medicine and education, in the future.")

[1592] Step 4: List the responses

[1593] The server generates a web page to display the answers of each generative AI model in a list format. The user views this page in a web browser on their device.

[1594] Input: Answer data saved on the server

[1595] Output: A web page displayed in list format (Example: Generative AI model 1: "AI will play an important role in many fields in the future, such as medicine and education.")

[1596] Step 5: Recognize differences in responses

[1597] The server compares the answers from each generative AI model and visually indicates the differences by calculating the similarity between the answers and displaying the differences using color coding and graphs based on the similarity.

[1598] Input: Answer data saved on the server

[1599] Output: A web page with visual indication of the differences (e.g. high similarity is shown in green, low similarity in red)

[1600] Step 6: Regenerate the hybrid answer

[1601] The user uses an interface to specify the proportion of each answer to be used. Based on the proportions specified by the user, the server combines the answers from each generative AI model to generate a hybrid answer, which is then displayed on the web page.

[1602] Input: The percentage of each answer specified by the user (e.g. 70% of the answers from AI model 1, 20% of the answers from AI model 2, and 10% of the answers from AI model 3)

[1603] Output: A web page with a hybrid answer (e.g., "AI will play an important role in many fields, such as medicine and education, and will continue to evolve as a tool to make human life more convenient, contributing to the realization of autonomous driving and smart cities.")

[1604] (Application example 1)

[1605] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1606] In systems using conventional generative AI models, there is a lack of a means to compare answers from multiple generative AI models and obtain the optimal answer. In addition, it is difficult to visually recognize the differences between answers from each generative AI model, and there is insufficient support for users to select the optimal answer. Furthermore, when combining answers from multiple generative AI models to generate a hybrid answer, a means is required that can flexibly respond based on the ratio specified by the user.

[1607] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1608] In this invention, the server includes a means for asking the same question to a plurality of generative AI models, a means for displaying a list of answers from each generative AI model, a means for recognizing the difference between each answer, a means for regenerating a hybrid answer by combining each answer in a ratio specified by the user, a means for calculating and visually displaying the similarity between each answer, and a means for the user to specify the ratio for generating the hybrid answer. This makes it possible to compare answers from a plurality of generative AI models and obtain an optimal answer. In addition, by visually recognizing the difference between each answer, support for the user to select the optimal answer is enhanced. Furthermore, it becomes possible to flexibly generate a hybrid answer based on the ratio specified by the user.

[1609] A "generative AI model" is a type of artificial intelligence that uses natural language processing to generate answers to user questions or prompts.

[1610] The "means of asking questions" is a function that allows users to send questions to multiple generative AI models.

[1611] "Means of displaying answers in a list" is a function for displaying answers obtained from multiple generative AI models side-by-side on a single screen.

[1612] A "means of recognizing differences" is a function that can compare answers from multiple generative AI models and identify the differences between them.

[1613] A "means for regenerating hybrid answers" is a function for generating new answers by combining answers from multiple generative AI models in a ratio specified by the user.

[1614] "Means for calculating similarity" is a function for numerically evaluating the similarity between answers from multiple generative AI models.

[1615] "Means for visual display" refers to a function for displaying the calculated similarity or difference in answers in a visual format such as a graph or color coding.

[1616] "Means for specifying proportions" refers to a function that enables users to specify the proportions in which answers from each generative AI model should be combined.

[1617] The system for implementing the present invention operates in cooperation with a server, a terminal, and a user. A specific embodiment of the system will be described below.

[1618] System Configuration

[1619] server

[1620] The server includes means for asking the same question to multiple generative AI models, means for displaying a list of answers from each generative AI model, means for recognizing differences between each answer, means for combining each answer in a ratio specified by a user to regenerate a hybrid answer, means for calculating and visually displaying the similarity of each answer, and means for a user to specify the ratio for generating a hybrid answer.

[1621] The server uses the following software and hardware:

[1622] Generative AI Models: OpenAI API

[1623] Data processing: Python program

[1624] Visualization: Matplotlib, Scikit-learn

[1625] Terminal

[1626] The terminal provides an interface for the user to enter questions and view generated answers. The terminal may be a smartphone, smart glasses, a head-mounted display, or the like.

[1627] User

[1628] Users input questions through their devices, review answers from generative AI models, and specify the proportions for generating hybrid answers.

[1629] Processing flow

[1630] 1. Entering a question: The user enters a question through the terminal, for example, a prompt such as "Please give me some ideas for a new blog post."

[1631] 2. Question submission: The server submits the input question to multiple generative AI models.

[1632] 3. Obtaining answers: The server receives answers from each generative AI model.

[1633] 4. Displaying a list of answers: The server displays the answers from each generative AI model in a list format on the terminal.

[1634] 5. Difference Recognition: The server calculates the similarity of each answer and displays it visually, for example using a heatmap to show the similarity between answers.

[1635] 6. Specify Proportion: The user specifies what proportion of each answer should be combined.

[1636] 7. Generate hybrid answer: The server generates a hybrid answer based on the specified percentages and displays it on the terminal.

[1637] Examples

[1638] When a user types a question like "What are some ideas for a new blog post?", the server sends the question to multiple generative AI models, which can produce answers like these:

[1639] Generative AI Model A: "Latest trends in AI technology"

[1640] Generative AI Model B: "Advantages and disadvantages of remote work"

[1641] Generative AI Model C: "Sustainable Energy Future"

[1642] The server lists these answers and visually displays the similarity between the answers in a heat map. For example, the user can specify that "Latest trends in AI technology" should be combined at 70%, and the other two at 20% and 10%, respectively. The server generates a hybrid answer based on this specification and provides it to the user.

[1643] In this way, users can compare answers from multiple generative AI models to obtain the best answer. Also, visual recognition of the differences between each answer provides enhanced support for selecting the more appropriate answer. Furthermore, it is possible to flexibly generate hybrid answers based on the ratio specified by the user.

[1644] The flow of the specific process in the application example 1 will be described with reference to FIG.

[1645] Step 1:

[1646] The user inputs a question through a terminal.

[1647] Input: The prompt the user enters (e.g. "Can you give me some ideas for new blog posts?")

[1648] Output: The prompt text entered

[1649] Specific operations: The user inputs a question using the input interface of the terminal and presses the send button.

[1650] Step 2:

[1651] The server sends the input question to multiple generative AI models.

[1652] Input: The prompt text entered by the user.

[1653] Output: The prompt sent to each generative AI model

[1654] Specific operation: The server uses the API to send prompt sentences to e...

Claims

1. A means for inputting the same question to a plurality of generative AI models; A means for displaying a list of each of the answers to the question generated by the plurality of generative AI models; A means for regenerating a new answer by combining each of the answers in a specified ratio; means for outputting the regenerated new answer; A system including:

2. means for recognizing a user's emotion and adjusting the new answer based on emotion information indicative of the recognized emotion of the user; The system of claim 1 .

3. The means for displaying the list displays the differences between the answers in a visually recognizable format. The system of claim 1 .

Images