Information processing system and information processing method
The system enhances data collection for fine-tuning generative AI models by using biometric data acquisition, addressing the inefficiency of manual dataset creation, thereby improving model responsiveness.
Patent Information
- Application Number
- JP2024084657
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-05-24
- Publication Date
- 2025-12-05
AI Technical Summary
The manual creation of datasets for fine-tuning generative AI models like GPT-4 and Llama2 is labor-intensive, necessitating improved data collection efficiency.
An information processing system and method that integrates a reception unit, output unit, biometric data acquisition, and processing units to collect and utilize user biometric data as training data for fine-tuning language models, including facial expressions, pulse rate, and gaze information, to enhance model responsiveness.
Efficiently collects training data for fine-tuning language models, improving their responsiveness to user intent.
Smart Images
Figure 2025177634000001_ABST
Abstract
Description
[Technical Field]
[0001] The present invention relates to information processing using a language model, particularly a generative AI (Artificial Intelligence) model, and a language model learning method.
[0002] Note that one embodiment of the present invention is not limited to the above technical field. Examples of the technical field of one embodiment of the present invention disclosed in this specification and the like include semiconductor devices, display devices, light-emitting devices, power storage devices, memory devices, electronic devices, lighting devices, input devices, input / output devices, driving methods thereof, and manufacturing methods thereof. A semiconductor device refers to any device that can function by utilizing semiconductor characteristics. [Background technology]
[0003] In recent years, there has been active development of language models using neural networks, with large-scale language models (LLMs) attracting particular attention. Large-scale language models are natural language processing models trained using large amounts of data. Large-scale language models can be used to realize, for example, dialogue models that respond to user instructions. Non-Patent Document 1 discloses GPT-4 (Generative Pre-trained Transformer 4), Llama2, and Llama3 as large-scale language models, and ChatGPT as a dialogue model.
[0004] In order for dialogue models such as ChatGPT to respond in line with user intent, they undergo a learning process called Policy Optimization (PO), which fine-tunes the language model (adjusting the language model parameters through re-learning) to produce responses that are preferable to humans.
[0005] PO includes methods called Proximal Policy Optimization (PPO) (Non-Patent Document 2), which uses reinforcement learning using an evaluation model that reflects human values, and Direct Policy Optimization (DPO) (Non-Patent Document 3), which allows LLM to learn value standards directly from data. [Prior art documents] [Non-patent literature]
[0006] [Non-Patent Document 1] Summary of ChatGPT / GPT-4 Research and Perspective Towards the Future of Large Language Models, Yiheng Liu et al. (Submitted on 4 Apr 2023, [online], Internet<URL:https: / / arxiv.org / abs / 2304.01852> [Non-patent document 2] Proximal Policy Optimization Algorithms, John Schulman et al. (Submitted on 20 Jul 2017, [online], Internet<URL:https:arxiv.org / abs / 1707.06347> ) [Non-patent document 3] Direct Preference Optimization: Your Language Model is Secretly a Reward Model, Rafael Rafailov et al. (Submitted on 29 May 2023, [online], Internet<URL:https:arxiv.org / abs / 2305.18290> ) Summary of the Invention [Problem to be solved by the invention]
[0007] However, in order to implement PPO and DPO, it is necessary to manually create a dataset, which requires a great deal of effort.
[0008] In view of the above problems, one aspect of the present invention aims to improve the efficiency of data collection required for fine-tuning generative AI.
[0009] Note that the description of these problems does not preclude the existence of other problems. Note that one embodiment of the present invention does not necessarily solve all of these problems. Note that problems other than these can be extracted from the description of the specification, drawings, claims, etc. [Means for solving the problem]
[0010] In view of the above problems, one aspect of the present invention is an information processing system including a reception unit, an output unit, a first processing unit, a second processing unit, a third processing unit, a biometric data acquisition unit, and a storage unit. The reception unit has a function of receiving an instruction and transmitting the instruction to the first processing unit. The first processing unit has a function of inputting the instruction to a first language model and transmitting an answer generated by the first language model to the output unit. The output unit has a function of presenting the answer to a user. The biometric data acquisition unit has a function of acquiring first biometric data of a user while the user is confirming the answer and a function of transmitting the first biometric data to the second processing unit. The second processing unit has a function of converting the first biometric data into vector data and a function of transmitting the vector data to the storage unit. The third processing unit has a function of updating parameters of the first language model using the instruction, the answer, and the vector data as training data.
[0011] In the information processing system, it is preferable that the biometric data acquisition unit has an imaging unit, and the imaging unit has a function of photographing the user.
[0012] In the information processing system, the biometric data acquisition unit preferably has a function of measuring the pulse rate of the user.
[0013] In the above information processing system, it is preferable that the biometric data acquisition unit has a function of acquiring second biometric data of the user before the user confirms the answer, and a function of transmitting the second biometric data to the second processing unit.
[0014] In the above information processing system, it is preferable that the biometric data acquisition unit has a function of determining whether the user is confirming the answer.
[0015] Another aspect of the present invention is an information processing method having first to sixth steps, in which in the first step, an instruction is received; in the second step, an answer is generated using a first language model based on the instruction; in the third step, the answer is presented to a user; in the fourth step, first biometric data of the user is obtained while the user is confirming the answer; in the fifth step, the first biometric data is converted into vector data; and in the sixth step, the instruction, answer, and vector data are used as training data to update parameters of the first language model.
[0016] Preferably, the information processing method further comprises a seventh step of acquiring second biometric data before the user confirms the answer, and the seventh, third and fourth steps are performed in this order. [Effects of the Invention]
[0017] According to one aspect of the present invention, it is possible to efficiently collect data required for fine-tuning a language model.
[0018] Note that the description of these effects does not preclude the existence of other effects. Note that one embodiment of the present invention does not necessarily have all of these effects. Note that effects other than these can be extracted from the description in the specification, drawings, claims, etc. [Brief explanation of the drawings]
[0019] [Figure 1]FIG. 1 is a schematic diagram illustrating an example of a configuration of an information processing system according to an embodiment. [Figure 2] FIG. 2 is a block diagram illustrating an example of a configuration of an information processing device according to an embodiment. [Figure 3] FIG. 3 is a flowchart illustrating an example of an information processing method according to the embodiment. [Figure 4] FIG. 4 shows an example of receiving an instruction, outputting a response, and acquiring biometric data according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION
[0020] Hereinafter, embodiments will be described with reference to the drawings. However, it will be readily understood by those skilled in the art that the embodiments can be implemented in many different ways and that various changes in form and details can be made without departing from the spirit and scope of the present invention. Therefore, the present invention should not be interpreted as being limited to the following description of the embodiments.
[0021] In the configuration of the invention described below, the same parts or parts having similar functions are denoted by the same reference numerals in different drawings, and repeated explanations thereof will be omitted. In addition, when referring to similar functions, no particular reference numerals may be used.
[0022] In the drawings described in this specification, the size of each component, the thickness of a layer, or an area may be exaggerated for clarity, and therefore, the drawings are not necessarily limited to the scale.
[0023] Note that ordinal numbers such as "first" and "second" used in this specification are used to avoid confusion between components and do not limit the number. They do not indicate any order or ranking, such as the order of processes or stacking. Even if a term does not have an ordinal number in this specification, an ordinal number may be added in the claims to avoid confusion between components. Even if a term has an ordinal number in this specification, a different ordinal number may be added in the claims. Even if a term has an ordinal number in this specification, the ordinal number may be omitted in the claims.
[0024] In this specification, a language model is based on the Transformer architecture and has undergone additional learning to become an interactive (also called conversational) model. A typical language model is a large-scale language model (LLM). An LLM processes data based on given text and is specialized for text generation functions, while a generative AI not only generates text but also has image generation functions that process data based on image data. In other words, a large-scale language model is a type of generative AI.
[0025] (Embodiment 1) In this embodiment, an information processing system and an information processing method according to one embodiment of the present invention will be described.
[0026] An information processing device and an information processing system according to one aspect of the present invention utilize a language model to output a response to an instruction input by a user. Furthermore, biometric data is acquired while the user is confirming the response. Furthermore, by storing the biometric data as vector data, it is possible to improve the efficiency of collecting training data required for fine-tuning the language model.
[0027] A more specific example will be described below with reference to the drawings.
[0028] <Configuration example 1 of information processing system> The information processing system of this embodiment preferably includes an information processing device 10, an information terminal 20, and a biometric data acquisition unit 30, as shown in Fig. 1. As shown in Fig. 1, the information terminal 20 is connected to the information processing device 10 via a network 40. The information terminal 20 is also connected to the biometric data acquisition unit 30 via a network 50.
[0029] <<Example of information terminal configuration>> In the exemplary configuration of the information processing system, the information terminal 20 is operated by a user and can also be called a client computer. While FIG. 1 shows a desktop computer, a smartphone, etc. as examples, a notebook computer, a tablet computer, etc. can also be used as the information terminal 20. A tablet computer can fold back a housing having a reception unit (typically a keyboard) and place it on top of the main body. In some tablet computers, the housing can be separated from the main body.
[0030] <<Configuration example of biometric data acquisition unit>> In the exemplary configuration of the information processing system, the biometric data acquisition unit 30 has the function of acquiring biometric data when the user is confirming their answers. While FIG. 1 shows a webcam and a smartwatch (registered trademark) as examples, it is preferable that these are capable of communicating with the information processing device 10 or the information terminal 20 and capable of measuring the user's biometric data. The biometric data may be data that changes depending on the user's emotions. Examples of such data include facial expressions, complexion, pupil size, and gaze information obtained from images (including videos) of the user's face, as well as information such as the user's words and tone of voice obtained by recording audio. Other examples include the user's vital signs, such as pulse wave, heart rate, pulse rate, blood pressure, respiratory frequency, respiratory depth, and body temperature.
[0031] It is preferable that the biometric data acquisition unit 30 be able to determine whether the user is checking the answer. For example, it is possible to determine whether the user is checking the answer by using gaze detection or the like from an image (including video) of the user's face, etc. Alternatively, it is preferable that the information processing device 10 extracts biometric data when the user is checking the answer and biometric data before the user checks the answer using image data and biometric data. In this case, it is preferable that the biometric data acquisition unit 30 or the information terminal 20 acquires image (including video) data of the user's face, etc., and transmits the image data to the information processing device 10. Furthermore, it is preferable that the biometric data acquisition unit 30 or the information terminal 20 transmits time data to the information processing device 10. This enables the information processing device 10 to temporally synchronize the image data and the biometric data.
[0032] <<Configuration Example of Information Processing Device>> Next, an example of the configuration of the information processing device will be described with reference to FIG.
[0033] As shown in Fig. 2, the information processing device 10 has a reception unit 110, an output unit 120, a first processing unit 130, a second processing unit 140, a third processing unit 150, a storage unit 160, and a transmission path 170. In addition to the information processing device 10, Fig. 2 also shows an information terminal 20 and a biometric data acquisition unit 30, with arrows indicating transmission and reception of data. The reception unit and output unit may be collectively referred to as a communication unit. The communication unit enables the information processing device 10 to transmit and receive data to and from the outside.
[0034] [Reception Section 110] The information processing device 10 can have a function of receiving data from the outside by the receiving unit 110. For example, the receiving unit 110 can receive data from the information terminal 20.
[0035] Examples of data that the reception unit 110 receives from the information terminal 20 include instructions input by the user, biometric data of the user, and the like.
[0036] The receiving unit 110 can supply the received data via a transmission path 170 to one or more selected from the memory unit 160, the first processing unit 130, the second processing unit 140, and the third processing unit 150.
[0037] [Output section 120] The information processing device 10 has a function of outputting calculation results and the like to the outside by the output unit 120. For example, the output unit 120 can transmit data to the information terminal 20.
[0038] [First Processing Unit 130, Second Processing Unit 140, and Third Processing Unit 150] The first processing unit 130, the second processing unit 140, and the third processing unit 150 have the function of performing processing such as calculation, analysis, and inference using data supplied from one or both of the receiving unit 110 and the storage unit 160. The first processing unit 130, the second processing unit 140, and the third processing unit 150 can supply the generated data (for example, calculation results, analysis results, inference results) to one or both of the storage unit 160 and the output unit 120.
[0039] The first processing unit 130, the second processing unit 140, and the third processing unit 150 may each include, for example, an arithmetic circuit. The first processing unit 130 may each include, for example, a central processing unit (CPU). Furthermore, the first processing unit 130, the second processing unit 140, and the third processing unit 150 may each include, for example, a graphics processing unit (GPU).
[0040] The first processing unit 130, the second processing unit 140, and the third processing unit 150 may each have a register and a main memory in addition to a CPU. The register and main memory may also be said to be owned by the CPU. The main memory is capable of sending and receiving data to and from a secondary cache or the like. The main memory has at least one of a volatile memory such as RAM (Random Access Memory) and a non-volatile memory such as ROM (Read Only Memory). The main memory may also have at least one of NOSRAM (registered trademark) and DOSRAM (registered trademark). The main memory can have one or both of OS transistors and Si transistors. The configuration of the register and main memory can be understood by replacing CPU in this paragraph with GPU.
[0041] NOSRAM is an abbreviation for "Nonvolatile Oxide Semiconductor Random Access Memory (RAM)." NOSRAM is a type of memory in which memory cells are two-transistor (2T) or three-transistor (3T) gain cells, and the transistors are transistors (also called OS transistors) that use metal oxides in the channel formation region. OS transistors have extremely low leakage current, i.e., the current that flows between the source and drain in the off state. NOSRAM can be used as a nonvolatile memory by retaining a charge corresponding to data in the memory cell using its extremely low leakage current. NOSRAM is particularly suitable for arithmetic processing that requires repeated large amounts of data read operations because it can read stored data without destroying it (nondestructive read). NOSRAM can increase its data capacity by stacking layers, so it can be used as a large-scale cache memory, main memory, or storage memory, thereby improving the performance of semiconductor devices.
[0042] DOSRAM is an abbreviation for "Dynamic Oxide Semiconductor RAM" and refers to RAM with 1T (transistor) 1C (capacitance) type memory cells. DOSRAM is a DRAM formed using OS transistors, and is a memory that temporarily stores information sent from outside. DOSRAM is a memory that takes advantage of the small off-current of OS transistors.
[0043] Examples of RAM include DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). The DRAM or SRAM can be used as a working space for the first processing unit 130, the second processing unit 140, and the third processing unit 150 by virtually allocating memory space thereto. The operating system, application programs, program modules, program data, lookup tables, and the like stored in the storage unit 160 are loaded into the RAM immediately before execution. The operating system, application programs, program modules, program data, and lookup tables loaded into the RAM can be accessed by the first processing unit 130, the second processing unit 140, and the third processing unit 150, respectively.
[0044] ROM can store systems that do not require rewriting. Examples of systems that do not require rewriting include BIOS (Basic Input / Output System) and firmware. Examples of ROM include mask ROM, OTPROM (One-Time Programmable Read Only Memory), and EPROM (Erasable Programmable Read Only Memory). Examples of EPROM include UV-EPROM (Ultra-Violet Erasable Programmable Read Only Memory), which allows stored data to be erased by exposure to ultraviolet light, EEPROM (Electrically Erasable Programmable Read Only Memory), and flash memory.
[0045] The first processing unit 130, the second processing unit 140, and the third processing unit 150 may each have a microprocessor such as a DSP (Digital Signal Processor) in addition to a CPU or a GPU. Since a DSP is specialized for digital signal processing, it is preferable to install a DSP to control peripheral circuits of the CPU or GPU. The microprocessor may be realized by a PLD (Programmable Logic Device) that operates on hardware such as an FPGA (Field Programmable Gate Array) or an FPAA (Field Programmable Analog Array).
[0046] [First Processing Unit 130] The first processing unit 130 can perform processing using a model that utilizes a large-scale language model (such as a document generation model or a dialogue model). For example, the first processing unit 130 can perform processing using a large-scale language model such as GPT-4, Llama2, or Llama3. Note that in this specification and the like, the term "language model" includes large-scale language models.
[0047] The first processing unit 130 has a function of inputting an instruction input by a user via the information terminal 20 into a language model via the reception unit 110 and generating a response to the instruction. Furthermore, the first processing unit 130 has a function of sending the response generated by the language model to the storage unit 160, a function of presenting the response to the user via the output unit 120, and the like.
[0048] [Second Processing Unit 140] The second processing unit 140 performs processing using neural networks, machine learning models, and information processing, converts biometric data into vector data, and can perform emotion estimation. For example, it can perform emotion analysis from facial expression images using a Python library such as FER (Face Emotion Recognizer) and output the result as vector data. It can also perform emotion estimation from the low-frequency and high-frequency components of the pulse rate and output the result as vector data.
[0049] The second processing unit 140 has a function of processing the biometric data acquired via the receiving unit 110 using a neural network, a machine learning model, and program processing, and outputting vector data. Furthermore, the second processing unit 140 has a function of sending the output vector data to the storage unit 160.
[0050] The second processing unit 140 preferably has a function of processing the biometric data before and while the user is checking the answer using a neural network, a machine learning model, and program processing, and outputting vector data. By using the biometric data before and after the user checks the answer, accurate emotion estimation can be performed.
[0051] [Third Processing Unit 150] The third processing unit 150 can fine-tune the language model and update it. As training data, a set of instructions input by a user, responses to the instructions, and vectorized biometric data output by the second processing unit 140 is used. By using the vectorized biometric data as training data, the language model is trained to be able to output responses that are meaningful to humans.
[0052] [Storage section 160] The storage unit 160 provides the information processing device 10 with a storage function. The storage unit 160 is a memory area and can store programs and / or data, etc. Representative programs include programs executed by the first processing unit 130, the second processing unit 140, and the third processing unit 150. The data includes data received by the reception unit 110 (e.g., instructions input by the user, biometric data acquired by the biometric data acquisition unit 30). The data also includes data generated by the first processing unit 130, the second processing unit 140, and the third processing unit 150 (e.g., answers, calculation results, analysis results, and inference results generated by the first processing unit 130).
[0053] The storage unit 160 may have a database. Furthermore, the information processing device 10 may have a database separate from the storage unit 160. The information processing device 10 may have a function to retrieve data from a database that exists outside the storage unit 160, outside the information processing device 10, or outside the information processing system. Furthermore, the information processing device 10 may have a function to retrieve data from both its own database and an external database.
[0054] Either or both of a storage and a file server can be used as the storage unit 160. Also, the storage unit 160 can be a database that records the paths of files stored in the file server.
[0055] The storage unit 160 has at least one of a volatile memory and a non-volatile memory. Examples of the volatile memory include DRAM and SRAM. Examples of the non-volatile memory include ReRAM (Resistive Random Access Memory, also called Resistive Memory), PRAM (Phase Change Random Access Memory), FeRAM (Ferroelectric Random Access Memory), MRAM (Magnetoresistive Random Access Memory, also called Magnetoresistive Memory), and flash memory. The storage unit 160 may also have at least one of NOSRAM and DOSRAM. The storage unit 160 may also have a recording media drive. Examples of the recording media drive include a hard disk drive (HDD) and a solid state drive (SSD).
[0056] In this specification and the like, the term "metal oxide" refers to an oxide of a metal in a broad sense. Metal oxides are classified into oxide insulators, oxide conductors (including transparent oxide conductors), oxide semiconductors (also referred to as oxide semiconductors or simply as OSs), and the like. For example, when a metal oxide is used in a semiconductor layer of a transistor, the metal oxide may be referred to as an oxide semiconductor.
[0057] The metal oxide contained in the channel formation region preferably contains indium (In). When the metal oxide contained in the channel formation region contains indium, the carrier mobility (electron mobility) of the OS transistor is increased. Furthermore, the metal oxide contained in the channel formation region is preferably an oxide semiconductor containing element M. The element M is preferably at least one of aluminum (Al), gallium (Ga), and tin (Sn). Other elements applicable to element M include boron (B), silicon (Si), titanium (Ti), iron (Fe), nickel (Ni), germanium (Ge), yttrium (Y), zirconium (Zr), molybdenum (Mo), lanthanum (La), cerium (Ce), neodymium (Nd), hafnium (Hf), tantalum (Ta), and tungsten (W). However, a combination of two or more of the above elements may be used as element M. The element M is, for example, an element having a high binding energy with oxygen. For example, it is an element whose bond energy with oxygen is higher than that of indium. Furthermore, the metal oxide contained in the channel formation region is preferably a metal oxide containing zinc (Zn). Metal oxides containing zinc may be more likely to crystallize.
[0058] The metal oxide contained in the channel formation region is not limited to a metal oxide containing indium, but may be, for example, a metal oxide containing zinc but not indium, such as zinc tin oxide or gallium tin oxide, a metal oxide containing gallium, or a metal oxide containing tin.
[0059] [Transmission Line 170] The transmission path 170 has a function of transmitting data. Data can be transmitted and received between the reception unit 110, the output unit 120, the first processing unit 130, the second processing unit 140, the third processing unit 150, and the storage unit 160 via the transmission path 170.
[0060] <<Examples of information processing methods>> FIG. 3 is an example of a flowchart showing steps relating to an information processing method according to one aspect of the present invention.
[0061] <Step S101> In step S101, the information processing device 10 receives an instruction from the user.
[0062] The user's input operation corresponding to step S101 is performed at the information terminal 20. The instruction is transmitted to the information processing device 10 via the reception unit 110 and input to the language model by the first processing unit 130. The input instruction is stored in the storage unit 160. The instruction is input as text data, voice data, or the like.
[0063] <Step S102> In step S102, the first processing unit 130 of the information processing device 10 transmits the answer output by the language model to the information terminal 20 via the output unit 120, and the information terminal 20 displays the answer to the user. The answer output by the language model is saved in the storage unit 160. The answer is output as text data, audio data, or the like.
[0064] In step S102, it is preferable that the biometric data acquisition unit 30 acquires the biometric data of the user before the information terminal 20 displays the answer to the user. Specifically, it is preferable that the acquisition unit 30 acquires the biometric data of the user before step S101, simultaneously with step S101, or between steps S101 and S102. The biometric data may be transmitted to the reception unit 110 of the information processing device 10. At this time, the biometric data may be sent to the information processing device 10 via the information terminal 20.
[0065] <Step S103> In step S103, while the user is checking the answer, the biometric data acquisition unit 30 acquires the biometric data of the user. Then, the biometric data acquisition unit 30 transmits the biometric data to the reception unit 110 of the information processing device 10. At this time, the biometric data may be sent to the information processing device 10 via the information terminal 20.
[0066] <Step S104> In step S104, the second processing unit 140 of the information processing device 10 vectorizes the biometric data using a neural network, a machine learning model, program processing, etc. Furthermore, the vectorized biometric data is stored in the storage unit 160 of the information processing device 10.
[0067] In step S104, the neural network or machine learning model is preferably a model that can determine whether an emotion is positive or negative in biometric data.
[0068] In step S104, the second processing unit 140 may use biometric data obtained before and while the user is checking the answer to determine whether the emotion is positive or negative.
[0069] <Step S105> In step S105, the third processing unit 150 of the information processing device 10 stores the vectorized biometric data in the storage unit 160.
[0070] <Step S106> In step S106, it is determined whether the number of vectorized biometric data stored in the storage unit 160 of the information processing device 10 exceeds a certain number.
[0071] If the determination in step S106 is "No", the process returns to step S101.
[0072] <Step S107> If the answer is "Yes" in step S106, the third processing unit 150 of the information processing device 10 performs fine tuning of the language model in step S107, and updates the language model.
[0073] In step S107, the language model is trained using the questions, answers, and vectorized biometric data stored in the storage unit 160 of the information processing device 10 as training data.
[0074] The information processing method according to one aspect of the present invention can end after step S107.
[0075] By using such an information processing device and information processing method, it is possible to efficiently collect training data required for fine-tuning a language model.
[0076] <<Example of acquiring biometric data from a user>> FIG. 4 shows a specific example of biometric data acquisition when a user uses an information processing device according to an embodiment of the present invention.
[0077] 4, when the user inputs an instruction to the information terminal 20, the answer output from the language model is displayed on the information terminal 20. While the user is checking the answer, the biometric data acquisition unit 30 acquires the user's biometric data. The data format of the question and answer may be text data, voice data, or the like.
[0078] The biometric data acquisition unit 30 is not limited to a camera, and may be a terminal having a sensor for measuring pulse, such as a smart watch (registered trademark).
[0079] According to the information processing method of one aspect of the present invention, it is possible to efficiently collect training data for fine-tuning a language model. [Explanation of symbols]
[0080] 10. Information processing equipment 20 Information terminal 30 Biometric data acquisition unit 40 Network 50 Network 110 Reception 120 Output section 130 First processing section 140 Second processing section 150 Third Processing Section 160 Storage section 170 Transmission Line
Claims
1. The device includes a receiving unit, an output unit, a first processing unit, a second processing unit, a third processing unit, a biometric data acquiring unit, and a storage unit, the reception unit has a function of receiving an instruction and transmitting the instruction to the first processing unit; the first processing unit has a function of inputting the instruction to a first language model and transmitting a response generated by the first language model to the output unit; the output unit has a function of presenting the answer to a user; the biometric data acquisition unit has a function of acquiring first biometric data of the user while the user is confirming the answer, and a function of transmitting the first biometric data to the second processing unit; the second processing unit has a function of converting the first biometric data into vector data and a function of transmitting the vector data to the storage unit; The third processing unit has a function of updating parameters of the first language model using the instruction, the response, and the vector data as training data.
2. 2. The method according to claim 1, wherein the biometric data acquisition unit includes an imaging unit, The imaging unit has a function of photographing the user.
3. 2. The information processing system according to claim 1, wherein the biometric data acquisition unit has a function of measuring the pulse rate of the user.
4. 2. An information processing system according to claim 1, wherein the biometric data acquisition unit has a function of acquiring second biometric data of the user before the user confirms the answer, and a function of transmitting the second biometric data to the second processing unit.
5. 2. The information processing system according to claim 1, wherein the biometric data acquisition unit has a function of determining whether the user is confirming the answer.
6. The method includes first to sixth steps, In the first step, an instruction is received, In the second step, a response is generated based on the instruction using a first language model; In the third step, the answer is presented to the user; In the fourth step, biometric data of the user is acquired while the user is confirming the answer; In the fifth step, the biometric data is converted into vector data; In the sixth step, the information processing method updates parameters of the first language model using the instruction, the response, and the vector data as training data.
7. 7. The method according to claim 6, further comprising a seventh step of acquiring biometric data before the user confirms the answer, An information processing method in which the seventh, third and fourth steps are performed in this order.