system

Abstract

Provide a system. [Solution] A means for collecting data used by the generating AI for learning, a means for analyzing the collected data and verifying misinformation and bias, a means for conducting real-time security diagnosis while the generating AI is operating, a means for detecting misinformation and security risks and feeding that information back to the generating AI, and a means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyber attacks. A system including:

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 an instruction sentence related to a description of 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] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] In recent years, advances in generative AI technology have led to the provision of services utilizing generative AI in many fields. However, this has also created several challenges. First, there is a risk that generative AI will generate incorrect answers by learning false information. Second, there is a risk that the output data of generative AI may be tampered with by cyberattacks targeting generative AI. These issues can reduce the reliability and security of generative AI, hindering the widespread adoption of generative AI services. Therefore, a means to effectively resolve these challenges is desired. [Means for solving the problem]

[0005] In order to solve the above problems, the present invention provides the following means. First, a means for collecting data used by the generative AI for learning is provided. This means collects data from a wide range of data sources and uses it as learning material for the generative AI. Next, a means for analyzing the collected data and verifying misinformation and bias is provided. This means prevents the generative AI from learning misinformation. Furthermore, a means for performing security diagnosis in real time while the generative AI is operating is provided. This means can immediately detect when the generative AI is subjected to a cyber-attack. By providing a means for detecting misinformation and security risks and feeding that information back to the generative AI, it is possible to keep the generative AI's learning model accurate and secure. Furthermore, a means for retraining the generative AI's learning model and eliminating the effects of inaccurate data and cyber-attacks is provided. This makes it possible to improve the reliability and security of the generative AI.

[0006] "Generative AI" is a type of artificial intelligence that automatically generates data such as text, images, and audio.

[0007] "Data collection means" refers to the process or method for collecting data from various data sources that the generative AI uses for training.

[0008] "Data analysis methods" are algorithms and processes used to analyze collected data and verify whether it contains misinformation or bias.

[0009] "Security diagnostic measures" are technologies and procedures for monitoring the security status of the system in real time while the generating AI is operating, and for detecting abnormalities or signs of cyber attacks.

[0010] "Feedback measures" are processes or methods for feeding back detected misinformation or security risks to the generative AI.

[0011] "Learning model retraining measures" are techniques and procedures for retraining the learning model of generative AI and eliminating the effects of inaccurate data or cyber attacks. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 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. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 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. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 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. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0013] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0014] First, the terms used in the following description will be explained.

[0015] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single 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), and an APU (Accelerated Processing Unit).

[0016] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0017] 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.

[0018] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0019] 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. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0020] [First embodiment]

[0021] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0022] 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.

[0023] 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 WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0024] 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.

[0025] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. 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.

[0026] 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 (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0027] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0028] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0029] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" 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 process 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.

[0030] 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.

[0031] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. 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.

[0032] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. 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."

[0033] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0034] System Overview

[0035] 1. Data Collection and Analysis

[0036] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0037] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to verify whether it contains misinformation or bias. Unreliable data is filtered out, and reliable data is used as training material for the generative AI.

[0038] 2. Security Monitoring and Diagnostics

[0039] The device constantly monitors network traffic and system logs while the generated AI is running.

[0040] If the device detects any abnormal patterns or signs of a cyber attack, it will send that information to a server.

[0041] The server diagnoses the problem based on the information received from the device and notifies the user of the results. For example, if a high frequency of access from a specific IP address is detected, it will diagnose that there is a high possibility of a DDoS attack.

[0042] 3. Feedback and Corrections

[0043] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects based on this feedback, for example by preventing unreliable data from being reused.

[0044] Users receive notifications from the server, review reported issues, and can manually correct the generated AI's operations and settings if necessary.

[0045] 4. Automatic recovery

[0046] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data and cyber attacks.

[0047] The device will continue to monitor the generation AI during the recovery process to prevent any further abnormalities.

[0048] Specific examples

[0049] 1. Specific examples of data collection and analysis

[0050] The server crawls news sites to collect the latest medical information, adds metadata such as the acquisition date and time, data source, and data category to the collected data, and stores it in a data store.

[0051] The server performs NLP processing on the collected data to evaluate the reliability of the medical information. For example, if the same information is collected from multiple reliable sources, the data is assigned a high reliability score.

[0052] 2. Specific examples of security monitoring

[0053] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack.

[0054] The device sends this information to a server, which identifies any abnormal activity and diagnoses whether it is a DDoS attack.

[0055] 3. Examples of feedback and correction

[0056] The server detects misinformation generated by the AI ​​and feeds that information back to the AI. For example, if incorrect medical information is output, the server instructs the AI ​​to correct it based on a reliable data source.

[0057] Users receive feedback and change how the generating AI operates to prevent further misinformation from being generated.

[0058] 4. Specific examples of automatic recovery

[0059] The server retrains the generation AI's learning model and filters out data affected by the DDoS attack.

[0060] The device continuously monitors the operation of the generation AI during the recovery process to ensure that recurrence is prevented.

[0061] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[0062] The processing flow will be explained below.

[0063] Step 1:

[0064] The server launches a web crawler to collect data from specified data sources (news sites, academic paper databases, etc.). Specifically, the crawler accesses each data source, retrieves data via HTML or API, and stores it in a database.

[0065] Step 2:

[0066] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. For example, it tokenizes the acquired text data and applies a reliability evaluation algorithm to detect misinformation and biased data. It calculates a reliability score and adds low-quality data to a warning list.

[0067] Step 3:

[0068] The device monitors the system environment in real time while the generative AI is running, monitoring network traffic and analyzing system logs, and applying anomaly detection algorithms to detect abnormal activity and patterns.

[0069] Step 4:

[0070] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, the device reports a detailed log and the details of the detection to the server.

[0071] Step 5:

[0072] The server analyzes the abnormal data received from the terminal and determines the type of problem and the scope of its impact. For example, it analyzes abnormal access patterns, evaluates whether this is a DDoS attack, and identifies the affected services and data range.

[0073] Step 6:

[0074] The server generates a report based on the diagnostic results and notifies the user of that information. For example, if a DDoS attack is occurring, a report will be sent to the user containing details of the attack, the extent of its impact, and recommended countermeasures.

[0075] Step 7:

[0076] The user checks the notification from the server and manually adjusts the operation and settings of the generated AI as necessary, for example by blocking the IP address from which the unauthorized access originated.

[0077] Step 8:

[0078] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, specifically by filtering out unreliable data and updating the learning model with new, reliable data.

[0079] Step 9:

[0080] The device monitors the operation of the generation AI during the recovery process to ensure that abnormalities do not occur again. For example, it monitors data access and analysis processes during retraining in real time, and immediately reports any abnormalities to the server.

[0081] This series of processes reduces the risk of misinformation and cyberattacks from the generative AI, improving reliability and security.

[0082] Example 1

[0083] Next, a description will be given of 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."

[0084] If the data used by generative AI for training contains false information, there is a risk that false information will be generated. Furthermore, if generative AI becomes the target of a cyberattack, there is a high possibility that the data and the system will be adversely affected. It is necessary to prevent such problems and ensure the reliability and security of generative AI.

[0085] 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.

[0086] In this invention, the server includes means for collecting data used by the generative AI for learning, means for analyzing the reliability of the collected data using natural language processing and machine learning algorithms and verifying misinformation and bias, means for monitoring network traffic and system logs in real time while the generative AI is operating and detecting abnormal patterns and signs of cyberattacks, means for feeding back detected misinformation and security risks to the generative AI and adjusting or retraining the model, and means for retraining the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. This reduces the risk of the generative AI learning misinformation and makes it possible to protect against cyberattacks.

[0087] "Generative AI" is a system that uses artificial intelligence (AI) technology to generate new content and information from data.

[0088] "Data collection methods" are the processes and tools used to obtain the required information from designated data sources.

[0089] "Natural language processing (NLP)" is a technology that enables computers to understand, interpret, and generate natural language.

[0090] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications.

[0091] A "reliability score" is an index used to evaluate the accuracy and reliability of collected data.

[0092] "Network traffic" is the flow of data sent and received over a network.

[0093] A "system log" is a file or database that records system activity and events.

[0094] An "anomalous pattern" refers to a behavior or data trend that deviates from normal behavior.

[0095] A "cyber attack" is a malicious attack on a computer system or network.

[0096] "Feedback means" is a process that returns information to a system or model based on detected information, allowing for improvements or corrections.

[0097] "Retraining" is the process of retraining an existing learning model using new data to improve its performance.

[0098] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0099] Data collection and analysis

[0100] The server uses a web crawler to collect data from specified data sources (e.g., news sites or academic paper databases). The collected data is then accompanied by metadata such as the acquisition date and time, data source, and author. The server then evaluates the reliability of the collected data using natural language processing (NLP) tools (e.g., SpaCy or NLTK) or machine learning algorithms (e.g., Random Forest or SVM). Data with a low reliability score is filtered and not used as training material for the generative AI. For example, if the same information is obtained from multiple reliable sources, that information is assigned a high reliability score.

[0101] Security Monitoring and Diagnostics

[0102] While the generative AI is running, the device monitors network traffic and system logs in real time. It uses an IDS (intrusion detection system), such as Snort or Suricata. If the device detects an abnormal pattern or signs of a cyber attack, it sends that information to a server. The server diagnoses the problem based on the anomaly detection information received from the device. For example, frequent access from a specific IP address may be diagnosed as a DDoS attack. The diagnosis results are then notified to the user.

[0103] Feedback and Corrections

[0104] The server feeds back the results of the analysis and security diagnosis to the generative AI. For example, if the generative AI learns false information, it will provide feedback so that it can make corrections based on a reliable data source. The user receives notifications from the server and checks the reported issues. If necessary, it is also possible to manually correct the settings and operation of the generative AI model.

[0105] Automatic recovery

[0106] The server retrains the affected Generative AI's learning model to remove the effects of misinformation and cyberattacks. The retraining process uses a new data set that removes the effects of inaccurate data and attacks. The device monitors the Generative AI's behavior during this retraining process to ensure that recurrence is prevented.

[0107] Specific examples

[0108] 1. Specific examples of data collection

[0109] The server uses a web crawler to collect the latest news about the new coronavirus from news sites, and adds metadata to the collected data, such as the article title, text, publication date, and author.

[0110] 2. Specific examples of data analysis

[0111] The server analyzes the collected medical information using NLP tools. As part of natural language processing, key topics and sentiment are extracted. Machine learning models are also used to score reliability, with matching information from multiple reliable sources being given a high reliability score. Information with low reliability is filtered out.

[0112] 3. Specific examples of security monitoring

[0113] The device monitors the network traffic of the generated AI, and if more than 100 requests per second come from a specific IP address, it sends this as an anomaly to the server.

[0114] 4. Examples of Problem Diagnosis and Notification

[0115] The server diagnoses the anomaly, determines that "high frequency access from a specific IP address is likely a DDoS attack," and notifies the user. The user receives this information and takes appropriate measures.

[0116] 5. Examples of feedback and correction

[0117] The server provides feedback to the generative AI model that it contains misinformation and provides reliable alternative information. For example, if the model contains incorrect medical information, the server instructs the generative AI to refer to accurate information from an official medical institution.

[0118] 6. Specific examples of automatic recovery

[0119] The server uses the retraining dataset to retrain the generative AI model, removing data affected by attacks or misinformation, and the device monitors this retraining process to ensure no new anomalies arise.

[0120] Prompt Sentence Examples

[0121] Here are some examples of prompts for generative AI models:

[0122] "Collect medical information from news site A and news site B and rate the reliability of each."

[0123] "Monitor high-frequency access from specific IP addresses and report to the server if any signs of a DDoS attack are detected."

[0124] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[0125] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0126] Step 1: Data collection

[0127] The server uses a web crawler to collect data from news sites and academic paper databases. The input is the URL of the specified data source and a search query. The server accesses these URLs and retrieves the HTML content. It then uses an HTML parser to extract the necessary information (article title, body text, publication date, author, etc.). The extracted data is stored in a database along with metadata. The output is the collected dataset.

[0128] Step 2: Data analysis

[0129] The server analyzes the collected data by applying natural language processing (NLP) and machine learning algorithms. The input is the collected dataset. The server uses NLP tools (e.g., SpaCy, NLTK) to tokenize the text and apply sentiment analysis and topic models. The server scores the reliability of the data using machine learning algorithms (e.g., Random Forest, SVM). Data with low reliability is filtered. The output is the analyzed dataset with reliability ratings.

[0130] Step 3: Security monitoring

[0131] The terminal monitors network traffic and system logs in real time while the generative AI is running. The input is the generative AI's network traffic and system logs. The terminal uses an intrusion detection system (e.g., Snort, Suricata) to detect abnormal patterns and signs of cyber attacks. If an anomaly is detected, the information is sent to the server. The output is information about the detected anomaly.

[0132] Step 4: Problem diagnosis and notification

[0133] The server diagnoses the problem based on the anomaly information received from the terminal. The input is the anomaly information sent from the terminal. The server uses an anomaly detection algorithm (e.g., k-means clustering, isolated forest) to identify the type of anomaly and its scope of impact. For example, it determines whether high frequency access is a DDoS attack. The diagnosis result is notified to the user. The output is the diagnosis result indicating the type of problem and the scope of impact, along with a notification message.

[0134] Step 5: Feedback and revisions

[0135] The server feeds detected misinformation and security risks back to the generative AI. The inputs are the results from problem diagnosis and additional accurate data. The generative AI self-corrects based on the feedback, for example, by instructing it to correct misinformation based on reliable data sources. The user receives notifications from the server and manually corrects the settings and operation of the generative AI model as needed. The output is a corrected generative AI model.

[0136] Step 6: Automatic Recovery

[0137] The server retrains the generative AI's learning model to remove the effects of misinformation and cyberattacks. The input is the dataset used for retraining. The server uses a machine learning algorithm to train new data. During this process, it filters out inaccurate data or data affected by attacks. The device monitors the generative AI's behavior during the retraining process to prevent recurrence. The output is a retrained generative AI model.

[0138] (Application example 1)

[0139] Next, a description will be given of Application 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."

[0140] In content distribution services using generative AI, the training data may contain misinformation or misleading biased information. Furthermore, generative AI systems are vulnerable to cyberattacks, increasing the risk of security vulnerabilities. Ensuring that users receive reliable information and can use services safely is a key challenge.

[0141] 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.

[0142] In this invention, the server includes means for collecting data used by the generation AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing security diagnosis in real time while the generation AI is operating, means for detecting misinformation and security risks and feeding that information back to the generation AI, means for retraining the generation AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for evaluating reliability scores on user terminals and scoring collected content, means for performing natural language processing on the collected data and extracting features, and means for monitoring network traffic on user terminals and detecting abnormal behavior. This allows users to obtain reliable information and ensures the safety of the generation AI system.

[0143] "Generative AI" is artificial intelligence that learns from data and automatically generates content such as text and images.

[0144] A "data collection tool" is a method or device for collecting data from a specified source.

[0145] "Data analysis means" refers to methods or devices for analyzing collected data and detecting misinformation or bias.

[0146] "Security diagnostic means" refers to a method or device for evaluating and diagnosing the security status of a system in real time while the generating AI is operating.

[0147] "Feedback means" refers to a method or device for detecting misinformation or security risks and sending that information back to the generating AI to instruct it to make corrections.

[0148] "Retraining means" refers to methods or devices for updating the learning model of generative AI and eliminating the effects of inaccurate data or cyber attacks.

[0149] A "trust score evaluation means" is a method or device for evaluating collected content and quantifying its trust.

[0150] "Natural language processing" is a technology that allows computers to understand, analyze, and process human language.

[0151] A "feature extraction means" is a method or device for extracting useful information or patterns from data.

[0152] A "network traffic monitor" is a method or device for monitoring communication flows and detecting anomalous patterns or unauthorized behavior.

[0153] This invention is a system that provides safe and reliable information in a content distribution service using generative AI. This system operates in cooperation with three parties: a server, a terminal, and a user.

[0154] 1. Data Collection and Analysis

[0155] The server collects data from specified sources (such as news sites or academic paper databases) using tools such as web crawlers and analyzes the collected data using natural language processing (NLP) techniques. It calculates a reliability score and filters out unreliable data. Specifically, it accesses websites using the requests module and extracts text using BeautifulSoup.

[0156] 2. Security Monitoring and Diagnostics

[0157] The user device monitors network traffic and system logs in real time while the AI ​​generator is running. If any abnormal behavior or patterns are detected, the information is sent to the server, which then uses the information to diagnose the problem. The logging library is used to record network traffic.

[0158] 3. Feedback and Corrections

[0159] The server detects misinformation and security risks and provides feedback to the generation AI. For example, if a user reports a specific article as misinformation, the generation AI's learning data is corrected based on that information. The user receives a notification from the server and can manually correct the generation AI's settings as needed.

[0160] 4. Automatic recovery

[0161] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data or cyberattacks. The retrained model replaces the old model and begins working normally again.

[0162] Specific examples

[0163] For example, to ensure the reliability of medical information, the server collects medical news articles and rates their reliability. The same information collected from multiple reliable sources is given a high reliability score, and less reliable information is filtered out.

[0164] Prompt Sentence Examples

[0165] "Collect news articles from the URL 'https: / / example.com / news1' and calculate their confidence scores. Adjust the generative AI based on user feedback and perform automatic retraining."

[0166] This system allows users to obtain highly reliable information and use generative AI systems in a secure environment.

[0167] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0168] Step 1:

[0169] The server collects data from specified sources by accessing URLs using the requests module and extracting text from HTML using BeautifulSoup, taking a list of URLs to collect as input and producing raw text data as output.

[0170] Step 2:

[0171] The server analyzes the collected text data. It tokenizes the text data using natural language processing (NLP) techniques and extracts features using TfidfVectorizer. The input is the output data from step 1, and the output is a feature vector. It then calculates a reliability score and excludes data that is deemed unreliable.

[0172] Step 3:

[0173] The terminal monitors network traffic and system logs while the generation AI is running. It detects abnormal patterns and frequent access and sends this information to the server. The input is real-time network traffic data, and the output is log information when an abnormality is detected. Specifically, the logging library is used to record logs.

[0174] Step 4:

[0175] The server analyzes the received anomaly log and determines the extent of the impact of the security risk that occurred. It then feeds the results back to the user and the generating AI, correcting the generating AI model as needed. The input is the anomaly log information from Step 3, and the output is the feedback information.

[0176] Step 5:

[0177] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. Specifically, it retrains the model using a new dataset that removes misinformation and security risks. The input is the corrected dataset, and the output is an updated generative AI model.

[0178] Step 6:

[0179] The user checks the feedback information provided by the server and modifies the operation and settings of the generative AI system. At this time, the input is the feedback information from the server, and the output is the modified settings of the generative AI system.

[0180] Through these steps, content distribution services using generative AI are operated with high reliability and security.

[0181] Furthermore, an emotion engine that estimates the user's emotion may be 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.

[0182] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[0183] System Overview

[0184] 1. Data Collection and Analysis

[0185] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0186] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to check for misinformation and bias, analyze trends, and eliminate unreliable data.

[0187] 2. Emotion engine integration

[0188] The device collects emotional data through interactions with the user, extracted from voice tone, facial expressions, text input, and more.

[0189] The emotion engine analyzes the collected emotion data and provides the results as feedback, adjusting the generative AI's responses and output based on the user's emotions.

[0190] 3. Security Monitoring and Diagnostics

[0191] The device monitors the system environment in real time while the generative AI is running, analyzing network traffic and system logs and reporting any abnormalities detected to the server.

[0192] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers countermeasures.

[0193] 4. Feedback and Corrections

[0194] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects and generates highly reliable responses.

[0195] The emotion engine feeds user emotional data back to the generative AI, which then adapts to provide responses based on the user's emotions.

[0196] 5. Automatic Recovery

[0197] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, more reliable data for the retraining process.

[0198] The device will continue to monitor the generating AI during the recovery process and strengthen measures to prevent recurrence.

[0199] Specific examples

[0200] 1. Specific examples of data collection and analysis

[0201] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[0202] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[0203] 2. Specific examples of emotion engines

[0204] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[0205] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[0206] 3. Specific examples of security monitoring

[0207] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[0208] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[0209] 4. Examples of feedback and correction

[0210] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[0211] The user checks the notification from the server and manually corrects the operation of the generating AI to prevent further misinformation from being generated.

[0212] 5. Specific examples of automatic recovery

[0213] The server retrains the learning model of the generated AI to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[0214] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[0215] This system not only reduces the risk of misinformation and cyberattacks from generative AI, but also enables adaptive responses based on user emotions, improving reliability and user experience.

[0216] The processing flow will be explained below.

[0217] Step 1:

[0218] The server collects data from specified data sources (news sites, academic paper databases, etc.) by launching a web crawler, accessing each data source, acquiring data, and storing it in a database with metadata (acquisition date and time, data source, data category).

[0219] Step 2:

[0220] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. This involves tokenizing and contextual analysis to check for misinformation and bias. A reliability score is calculated, and any low scores are added to a warning list.

[0221] Step 3:

[0222] The device monitors the system environment in real time while the generative AI is running, specifically monitoring network traffic and system logs to detect abnormal patterns and signs of cyberattacks using an anomaly detection algorithm.

[0223] Step 4:

[0224] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, a warning will be reported to the server along with a detailed log of the situation.

[0225] Step 5:

[0226] The server analyzes the abnormal data received from the device to determine the type of problem and the extent of its impact. If a DDoS attack is suspected, it identifies the extent of the impact and takes necessary measures.

[0227] Step 6:

[0228] The server generates a report based on the diagnosis results and notifies the user of that information. For example, it generates a report that a DDoS attack has occurred, including details of the attack, the extent of its impact, and recommended countermeasures, and sends it to the user.

[0229] Step 7:

[0230] The user receives notifications from the server and manually adjusts the AI's operations and settings as needed, for example by blocking the IP address of the attacking source.

[0231] Step 8:

[0232] The server retrains the generative AI's learning model, which includes a retraining process to remove the effects of inaccurate data or cyberattacks, and updates the old model with new, more reliable data.

[0233] Step 9:

[0234] The terminal continues to monitor the generation AI in real time during the recovery process to ensure that the abnormality does not recur. Active monitoring is applied as a preventative measure, and if an abnormality is detected again, it is immediately reported to the server.

[0235] Step 10:

[0236] The device collects emotional data through interactions with the user, specifically analyzing the tone of voice during conversations, facial expressions, and text input to quantify the user's emotional state.

[0237] Step 11:

[0238] The emotion engine analyzes the collected emotional data and feeds the analysis results back to the generative AI. If the user is feeling stressed, they provide that information to the generative AI, which then adapts its response.

[0239] Step 12:

[0240] The emotion engine adjusts the generative AI's responses based on the user's emotions: for example, if the user is determined to be relaxed, the generative AI will use a friendly, calming tone.

[0241] This series of processes enables generative AI to reduce misinformation, protect against cyberattacks, and respond in a way that adapts to the user's emotions.

[0242] Example 2

[0243] Next, a description will be given of 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."

[0244] Generative AI relies on its performance to generate information from a wide range of data, but if that data contains misinformation or bias, the quality of the generated information will be reduced. Generative AI systems are also at risk of cyberattacks, making real-time security monitoring essential. Furthermore, there is a demand for improved user experience, making it necessary to respond based on user emotions.

[0245] 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.

[0246] In this invention, the server includes means for collecting data, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is running, means for collecting emotional data using interactions with users, means for analyzing the emotional data and adjusting the generating AI's response, means for detecting misinformation and security risks and feeding that information back to the generating AI, and means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyberattacks. This provides a system that reduces the risk of generating misinformation and cyberattacks by the generating AI and enables responses based on the user's emotions.

[0247] "Data" refers to the information and materials that generative AI uses for learning, and is collected from news sites, academic paper databases, etc.

[0248] "Collection means" refers to the technology or tools used to automatically obtain the required data from the designated data sources.

[0249] "Analytical means" refers to techniques or tools used to analyze collected data and verify whether it contains misinformation or bias.

[0250] "Generative AI" refers to artificial intelligence systems that generate new information or responses based on input data.

[0251] "Means for performing security diagnosis" refers to technologies or tools that monitor the environment in which the generative AI is operating in real time and detect cyber attacks and system anomalies.

[0252] A "user" is someone who interacts with the system and uses its services.

[0253] "Interaction" refers to the exchange of information and operations between a user and a system.

[0254] "Emotional data" is data that indicates the user's emotional state and is extracted from voice tone, facial expressions, text input, and the like.

[0255] "Means for collecting emotional data" refers to techniques or tools for capturing a user's emotional state from voice tone, facial expressions, text input, etc.

[0256] "Means for analyzing emotional data" refers to technology or tools that analyze the user's emotional state based on collected emotional data and provide feedback on the results to the generating AI.

[0257] "Feedback means" refers to technologies or tools that provide the generative AI with information such as detected misinformation, security risks, and user emotional data, allowing it to adjust or modify its response.

[0258] "Means for retraining learning models" refers to techniques or tools for rebuilding the learning models of generative AI using new, reliable data to remove the effects of inaccurate data or cyberattacks.

[0259] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[0260] System Overview

[0261] Data collection and analysis

[0262] The server uses a web crawler (e.g., Apache (registered trademark) Nutch) to collect data from news sites and academic paper databases. The collected data is assigned metadata such as date, time, source, and category. The server then analyzes the collected data using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TENSORFLOW (registered trademark)) to verify whether it contains misinformation or bias. Unreliable data is then removed.

[0263] Specific working example:

[0264] The server crawls the URL "https: / / example-academic-database.com" and downloads the text data and metadata.

[0265] The server uses SpaCy to perform grammatical analysis on the collected data.

[0266] A TensorFlow model is used to calculate a reliability score for the data, and data below a threshold is removed.

[0267] Emotion engine integration

[0268] The device collects emotion data through user interaction using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or the Google® Cloud Speech-to-Text API). The collected emotion data is sent to the emotion engine for analysis.

[0269] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[0270] Specific working example:

[0271] Facial expressions are captured using a laptop camera and analyzed using the OpenCV library.

[0272] Convert audio recorded by the microphone into text using the Google Cloud Speech-to-Text API.

[0273] The emotion engine identifies the user's stress level from the tone and rate of the recorded voice and sends that information to the generative AI.

[0274] Security Monitoring and Diagnostics

[0275] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[0276] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers the necessary countermeasures.

[0277] Specific working example:

[0278] The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[0279] The server analyzes the logs for abnormal traffic and finds that frequent requests are coming from a specific IP address, which is then added to a blacklist.

[0280] Feedback and Corrections

[0281] The server feeds detected misinformation and security risks back to the AI ​​generator, which then self-corrects based on this feedback to generate more reliable responses.

[0282] The emotion engine feeds the user's emotional data back to the generative AI, adapting it to provide responses that correspond to the user's emotions.

[0283] Specific working example:

[0284] The server analyzes the output text of the generated AI, detects that it contains incorrect information, and feeds back the error log to the generated AI.

[0285] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, which then adjusts the response it provides to be more relaxing.

[0286] Automatic recovery

[0287] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[0288] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[0289] Specific working example:

[0290] The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the generative AI's learning model.

[0291] The device uses Splunk to monitor the retraining logs in real time to check for any unusual activity.

[0292] Specific examples

[0293] 1. Specific examples of data collection and analysis

[0294] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[0295] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[0296] 2. Specific examples of emotion engines

[0297] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[0298] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[0299] 3. Specific examples of security monitoring

[0300] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[0301] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[0302] 4. Examples of feedback and correction

[0303] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[0304] If the emotion engine detects that the user is feeling stressed, it feeds that data back to the generative AI, which then adapts its responses to help the user relax.

[0305] 5. Specific examples of automatic recovery

[0306] The server retrains the generated AI's learning model to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[0307] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[0308] Prompt Sentence Examples

[0309] "Analyze the latest medical research data using NLP processing and evaluate its reliability."

[0310] "Determine whether the user is stressed and tailor your response accordingly."

[0311] "Detect abnormal network traffic and diagnose DDoS attacks."

[0312] "Detect misinformation and correct it based on reliable data."

[0313] "Retrain your generative AI learning models and restore your systems based on reliable data."

[0314] This system will reduce the risk of misinformation and cyberattacks from generative AI, and will also enable responses based on user emotions, improving reliability and user experience.

[0315] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0316] Step 1: Data collection

[0317] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases, and the collected data is annotated with metadata such as date, origin, and category.

[0318] Input: A list of URLs for the specified data sources

[0319] Specific example of operation: The server crawls the URL "https: / / example-academic-database.com" and downloads text data and metadata.

[0320] Output: Collected dataset (text data and metadata)

[0321] Step 2: Data analysis

[0322] The data collected by the server is analyzed using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow). The data is verified for misinformation and bias, and unreliable data is eliminated.

[0323] Input: Collected dataset

[0324] Example of how it works: The server uses SpaCy to perform grammatical analysis on the collected data, calculates a reliability score for the data using a TensorFlow model, and deletes data that falls below a threshold.

[0325] Output: A reliable dataset

[0326] Step 3: Collecting emotion data

[0327] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or the Google Cloud Speech-to-Text API).

[0328] Input: User interaction

[0329] Specific example: Facial expressions are captured using a laptop camera and analyzed using the OpenCV library. Audio recorded by a microphone is converted to text using the Google Cloud Speech-to-Text API.

[0330] Output: Collected emotional data (voice tone, facial expression analysis data, text)

[0331] Step 4: Analyze the sentiment data

[0332] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[0333] Input: Collected emotion data

[0334] Example of how it works: The emotion engine identifies the user's stress level from the tone and speed of the recorded voice and sends that information to the generative AI.

[0335] Output: Feedback data to the generation AI (user emotion analysis results)

[0336] Step 5: Security Monitoring

[0337] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[0338] Input: Real-time data from the generative AI system (network traffic, system logs)

[0339] Specific example of operation: The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[0340] Output: Security alerts and anomaly data

[0341] Step 6: Analysis of abnormal data and feedback

[0342] The server analyzes the abnormal data received from the device, diagnoses the type of problem and the extent of its impact, and, if necessary, provides feedback to the generating AI to correct the problem.

[0343] Input: Security alerts and anomaly data

[0344] Specific operation example: The server analyzes the log of abnormal traffic and finds that frequent requests are coming from a specific IP address. The server then adds that IP address to the blacklist.

[0345] Output: Feedback data and system correction instructions for the generative AI

[0346] Step 7: Retrain the learning model

[0347] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[0348] Input: New high-confidence dataset, inaccurate data detection results

[0349] Specific example of operation: The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the learning model of the generated AI.

[0350] Output: Retrained learning model

[0351] Step 8: Monitoring During Retraining

[0352] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[0353] Input: System state data during retraining

[0354] Example of how it works: The device uses Splunk to monitor the retraining logs in real time and check for any abnormal activity.

[0355] Output: Security status and monitoring reports

[0356] The above are the specific processing steps of the program for this system. By explaining in detail the specific operations performed at each step and their inputs and outputs, the specific method of implementing the invention will become clearer.

[0357] (Application example 2)

[0358] Next, a description will be given of Application 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."

[0359] Conventional generative AI systems have issues with insufficient resistance to learning misinformation and cyberattacks, and lack of appropriate countermeasures. Furthermore, they do not take user emotions into consideration and are unable to respond adaptively, which has the problem of not improving the user experience. The present invention aims to solve these issues and provide a system that achieves the reliability and safety of generative AI and an improved user experience.

[0360] 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.

[0361] In this invention, the server includes means for collecting data used by the generating AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is in operation, means for detecting misinformation and security risks and feeding that information back to the generating AI, means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for analyzing and recognizing passenger emotions and adjusting the generating AI's operation based on that information, and means for monitoring the network traffic of the autonomous vehicle, detecting abnormal activity, and taking immediate measures. This improves the reliability and safety of the generating AI and enables adaptive responses based on user emotions.

[0362] "Generative AI" is an artificial intelligence system that automatically generates information based on data and interacts with users.

[0363] A "data collection means" is a machine or software device that collects data for use in generative AI training.

[0364] "Data analysis tools" are algorithms and software tools used to analyze collected data and detect misinformation and bias.

[0365] "Security diagnostic means" is a function for checking the security of the system in real time while the generating AI is operating.

[0366] "Feedback measures" are mechanisms for detecting misinformation and security risks and returning that information to the generating AI.

[0367] "Retraining measures" are methods for retraining the generative AI's learning model to remove the effects of inaccurate data or cyber attacks.

[0368] "Emotion analysis means" refers to equipment or software that analyzes passenger emotions and can identify emotions based on voice, facial expressions, etc.

[0369] The "behavior adjustment means" is a function for adjusting the behavior of the generating AI based on the analyzed emotional data.

[0370] "Network traffic monitoring means" refers to devices or software that monitor communication data from autonomous vehicles, detect abnormal activity, and take immediate action.

[0371] MODE FOR CARRYING OUT THE INVENTION

[0372] Overall system configuration

[0373] This invention builds a system to realize the reliability and safety of self-driving vehicles, as well as to improve the user experience. The system is mainly composed of a server, a terminal, a generative AI, and an emotion engine working together. The roles and specific operations of each are explained below.

[0374] Hardware and software used

[0375] Hardware:

[0376] On-board computer (high performance computing unit)

[0377] Smartphone (ANDROID (registered trademark) or iOS)

[0378] In-vehicle camera (high resolution camera)

[0379] software:

[0380] Emotion recognition engine (commonly used API examples: Microsoft® Azure® Emotion API, Amazon Rekognition)

[0381] Machine learning platform (commonly used framework examples: TensorFlow, PyTorch)

[0382] Natural language processing (general software examples: SpaCy, NLTK)

[0383] Network monitoring tools (common tools examples: Wireshark, Snort)

[0384] System operation explanation

[0385] Data collection and analysis

[0386] The server collects data for the generative AI to use in its training. It uses a web crawler to gather the latest map and traffic information. The collected data is then analyzed through natural language processing (NLP) algorithms to verify its reliability. The server compares the data with reliable data sources such as Google Maps and OpenStreetMap to eliminate misinformation.

[0387] Emotion Recognition and Behavior Regulation

[0388] The devices (in this case, the onboard camera and smartphone) collect passenger facial expressions and vocal tones in real time. Based on this, the emotion recognition engine analyzes the passenger's emotions. The results of this analysis are fed back to the generative AI and used as a basis for adjusting the behavior of the autonomous vehicle.

[0389] For example, if the emotion recognition engine determines that a passenger is feeling stressed, the generative AI will adjust the vehicle's behavior, such as slowing down the vehicle's speed or driving more smoothly.

[0390] Security Monitoring

[0391] The terminal also monitors the autonomous vehicle's network traffic in real time. For example, if an abnormally high number of data requests is coming from a specific IP address, it could be detected as a sign of a cyber attack. This information is then sent to a server, which can then take appropriate measures.

[0392] Feedback and Corrections

[0393] The server feeds misinformation and detected security risks back to the Generator AI, which then self-corrects based on this feedback to generate more reliable responses. The retraining process uses newly collected, reliable data to keep the Generator AI's model up to date.

[0394] Examples and prompts

[0395] Specific examples

[0396] For example, if an onboard camera analyzes a passenger's facial expression and determines that the passenger is anxious, the autonomous vehicle will slow down. Also, if the latest map data contains incorrect information, that information will be immediately fed back to the generating AI, and the route of the autonomous vehicle will be corrected.

[0397] Prompt Sentence Examples

[0398] Prompt: Generate reliable routes based on up-to-date map and traffic data. Adjust responses based on real-time feedback, including sentiment data.

[0399] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0400] Step 1:

[0401] The server uses data collection methods to collect data for the generative AI to use for learning. It uses a web crawler to obtain the latest map data and traffic information from multiple reliable data sources (e.g., Google Maps, OpenStreetMap). The input is the URL of the data source and the data category to be collected, and the output is the collected raw data. The data is accompanied by metadata such as the acquisition date and time, data source, and data category.

[0402] Step 2:

[0403] The server processes the collected data using analytical tools to verify misinformation and bias. It uses natural language processing (NLP) algorithms to perform text analysis of the data and detect inaccurate or biased information. The input is the collected raw data, and the output is verified, reliable data. Specific operations include keyword extraction, theme classification, and reliability assessment.

[0404] Step 3:

[0405] The device uses emotion recognition means to collect passengers' facial expressions and vocal tones in real time. Input data is acquired using the onboard camera and microphone, and analyzed by an emotion recognition engine (API example: Microsoft Azure Emotion API). The input is facial expression data and voice data, and the output is analyzed emotion data. Specific operations include facial recognition, tone analysis, and emotion classification.

[0406] Step 4:

[0407] The terminal uses an action adjustment means based on the analyzed emotional data to adjust the behavior of the generating AI. If it determines that the passenger is feeling stressed, it feeds that information back to the generating AI, which then adjusts the vehicle's speed and driving pattern. The input is emotional data, and the output is adjusted driving instructions. Specifically, it generates and sends speed adjustment commands.

[0408] Step 5:

[0409] The terminal uses security monitoring means to monitor the autonomous vehicle's network traffic in real time. If it detects an abnormal traffic pattern (e.g., a high frequency of data requests), it immediately reports it to the server. The input is network traffic data, and the output is an anomaly detection report. Specifically, it monitors and analyzes data packets and stores abnormal traffic logs.

[0410] Step 6:

[0411] The server uses a feedback mechanism to feed back misinformation and detected security risks to the generative AI. It then reanalyzes the data using an NLP algorithm or machine learning model to update the generative AI's learning model. The input is the feedback information and new data, and the output is an updated generative AI model. Specific operations include retraining the model and evaluating its accuracy.

[0412] Step 7:

[0413] The server uses a retraining method to retrain the generative AI's learning model. It uses new data that has been cleaned of inaccurate data and the effects of cyberattacks to retrain the learning model so that it can generate reliable responses. The input is new, reliable data, and the output is a retrained generative AI model. Specific operations include preparing and cleaning the data and training the model.

[0414] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio 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 audio data.

[0415] 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 engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with 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.

[0416] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0417] [Second embodiment]

[0418] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0419] 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.

[0420] 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 WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0421] 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.

[0422] 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 audio in accordance with instructions from the processor 46.

[0423] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and 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 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0424] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0425] Fig. 4 shows an example of the 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.

[0426] 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.

[0427] 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.

[0428] 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.

[0429] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0430] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0431] System Overview

[0432] 1. Data Collection and Analysis

[0433] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0434] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to verify whether it contains misinformation or bias. Unreliable data is filtered out, and reliable data is used as training material for the generative AI.

[0435] 2. Security Monitoring and Diagnostics

[0436] The device constantly monitors network traffic and system logs while the generated AI is running.

[0437] If the device detects any abnormal patterns or signs of a cyber attack, it will send that information to a server.

[0438] The server diagnoses the problem based on the information received from the device and notifies the user of the results. For example, if a high frequency of access from a specific IP address is detected, it will diagnose that there is a high possibility of a DDoS attack.

[0439] 3. Feedback and Corrections

[0440] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects based on this feedback, for example by preventing unreliable data from being reused.

[0441] Users receive notifications from the server, review reported issues, and can manually correct the generated AI's operations and settings if necessary.

[0442] 4. Automatic recovery

[0443] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data and cyber attacks.

[0444] The device will continue to monitor the generation AI during the recovery process to prevent any further abnormalities.

[0445] Specific examples

[0446] 1. Specific examples of data collection and analysis

[0447] The server crawls news sites to collect the latest medical information, adds metadata such as the acquisition date and time, data source, and data category to the collected data, and stores it in a data store.

[0448] The server performs NLP processing on the collected data to evaluate the reliability of the medical information. For example, if the same information is collected from multiple reliable sources, the data is assigned a high reliability score.

[0449] 2. Specific examples of security monitoring

[0450] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack.

[0451] The device sends this information to a server, which identifies any abnormal activity and diagnoses whether it is a DDoS attack.

[0452] 3. Examples of feedback and correction

[0453] The server detects misinformation generated by the AI ​​and feeds that information back to the AI. For example, if incorrect medical information is output, the server instructs the AI ​​to correct it based on a reliable data source.

[0454] Users receive feedback and change how the generating AI operates to prevent further misinformation from being generated.

[0455] 4. Specific examples of automatic recovery

[0456] The server retrains the generation AI's learning model and filters out data affected by the DDoS attack.

[0457] The device continuously monitors the operation of the generation AI during the recovery process to ensure that recurrence is prevented.

[0458] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[0459] The processing flow will be explained below.

[0460] Step 1:

[0461] The server launches a web crawler to collect data from specified data sources (news sites, academic paper databases, etc.). Specifically, the crawler accesses each data source, retrieves data via HTML or API, and stores it in a database.

[0462] Step 2:

[0463] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. For example, it tokenizes the acquired text data and applies a reliability evaluation algorithm to detect misinformation and biased data. It calculates a reliability score and adds low-quality data to a warning list.

[0464] Step 3:

[0465] The device monitors the system environment in real time while the generative AI is running, monitoring network traffic and analyzing system logs, and applying anomaly detection algorithms to detect abnormal activity and patterns.

[0466] Step 4:

[0467] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, the device reports a detailed log and the details of the detection to the server.

[0468] Step 5:

[0469] The server analyzes the abnormal data received from the terminal and determines the type of problem and the scope of its impact. For example, it analyzes abnormal access patterns, evaluates whether this is a DDoS attack, and identifies the affected services and data range.

[0470] Step 6:

[0471] The server generates a report based on the diagnostic results and notifies the user of that information. For example, if a DDoS attack is occurring, a report will be sent to the user containing details of the attack, the extent of its impact, and recommended countermeasures.

[0472] Step 7:

[0473] The user checks the notification from the server and manually adjusts the operation and settings of the generated AI as necessary, for example by blocking the IP address from which the unauthorized access originated.

[0474] Step 8:

[0475] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, specifically by filtering out unreliable data and updating the learning model with new, reliable data.

[0476] Step 9:

[0477] The device monitors the operation of the generation AI during the recovery process to ensure that abnormalities do not occur again. For example, it monitors data access and analysis processes during retraining in real time, and immediately reports any abnormalities to the server.

[0478] This series of processes reduces the risk of misinformation and cyberattacks from the generative AI, improving reliability and security.

[0479] Example 1

[0480] Next, a description will be given of 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."

[0481] If the data used by generative AI for training contains false information, there is a risk that false information will be generated. Furthermore, if generative AI becomes the target of a cyberattack, there is a high possibility that the data and the system will be adversely affected. It is necessary to prevent such problems and ensure the reliability and security of generative AI.

[0482] 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.

[0483] In this invention, the server includes means for collecting data used by the generative AI for learning, means for analyzing the reliability of the collected data using natural language processing and machine learning algorithms and verifying misinformation and bias, means for monitoring network traffic and system logs in real time while the generative AI is operating and detecting abnormal patterns and signs of cyberattacks, means for feeding back detected misinformation and security risks to the generative AI and adjusting or retraining the model, and means for retraining the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. This reduces the risk of the generative AI learning misinformation and makes it possible to protect against cyberattacks.

[0484] "Generative AI" is a system that uses artificial intelligence (AI) technology to generate new content and information from data.

[0485] "Data collection methods" are the processes and tools used to obtain the required information from designated data sources.

[0486] "Natural language processing (NLP)" is a technology that enables computers to understand, interpret, and generate natural language.

[0487] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications.

[0488] A "reliability score" is an index used to evaluate the accuracy and reliability of collected data.

[0489] "Network traffic" is the flow of data sent and received over a network.

[0490] A "system log" is a file or database that records system activity and events.

[0491] An "anomalous pattern" refers to a behavior or data trend that deviates from normal behavior.

[0492] A "cyber attack" is a malicious attack on a computer system or network.

[0493] "Feedback means" is a process that returns information to a system or model based on detected information, allowing for improvements or corrections.

[0494] "Retraining" is the process of retraining an existing learning model using new data to improve its performance.

[0495] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0496] Data collection and analysis

[0497] The server uses a web crawler to collect data from specified data sources (e.g., news sites or academic paper databases). The collected data is then accompanied by metadata such as the acquisition date and time, data source, and author. The server then evaluates the reliability of the collected data using natural language processing (NLP) tools (e.g., SpaCy or NLTK) or machine learning algorithms (e.g., Random Forest or SVM). Data with a low reliability score is filtered and not used as training material for the generative AI. For example, if the same information is obtained from multiple reliable sources, that information is assigned a high reliability score.

[0498] Security Monitoring and Diagnostics

[0499] While the generative AI is running, the device monitors network traffic and system logs in real time. It uses an IDS (intrusion detection system), such as Snort or Suricata. If the device detects an abnormal pattern or signs of a cyber attack, it sends that information to a server. The server diagnoses the problem based on the anomaly detection information received from the device. For example, frequent access from a specific IP address may be diagnosed as a DDoS attack. The diagnosis results are then notified to the user.

[0500] Feedback and Corrections

[0501] The server feeds back the results of the analysis and security diagnosis to the generative AI. For example, if the generative AI learns false information, it will provide feedback so that it can make corrections based on a reliable data source. The user receives notifications from the server and checks the reported issues. If necessary, it is also possible to manually correct the settings and operation of the generative AI model.

[0502] Automatic recovery

[0503] The server retrains the affected Generative AI's learning model to remove the effects of misinformation and cyberattacks. The retraining process uses a new data set that removes the effects of inaccurate data and attacks. The device monitors the Generative AI's behavior during this retraining process to ensure that recurrence is prevented.

[0504] Specific examples

[0505] 1. Specific examples of data collection

[0506] The server uses a web crawler to collect the latest news about the new coronavirus from news sites, and adds metadata to the collected data, such as the article title, text, publication date, and author.

[0507] 2. Specific examples of data analysis

[0508] The server analyzes the collected medical information using NLP tools. As part of natural language processing, key topics and sentiment are extracted. Machine learning models are also used to score reliability, with matching information from multiple reliable sources being given a high reliability score. Information with low reliability is filtered out.

[0509] 3. Specific examples of security monitoring

[0510] The device monitors the network traffic of the generated AI, and if more than 100 requests per second come from a specific IP address, it sends this as an anomaly to the server.

[0511] 4. Examples of Problem Diagnosis and Notification

[0512] The server diagnoses the anomaly, determines that "high frequency access from a specific IP address is likely a DDoS attack," and notifies the user. The user receives this information and takes appropriate measures.

[0513] 5. Examples of feedback and correction

[0514] The server provides feedback to the generative AI model that it contains misinformation and provides reliable alternative information. For example, if the model contains incorrect medical information, the server instructs the generative AI to refer to accurate information from an official medical institution.

[0515] 6. Specific examples of automatic recovery

[0516] The server uses the retraining dataset to retrain the generative AI model, removing data affected by attacks or misinformation, and the device monitors this retraining process to ensure no new anomalies arise.

[0517] Prompt Sentence Examples

[0518] Here are some examples of prompts for generative AI models:

[0519] "Collect medical information from news site A and news site B and rate the reliability of each."

[0520] "Monitor high-frequency access from specific IP addresses and report to the server if any signs of a DDoS attack are detected."

[0521] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[0522] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0523] Step 1: Data collection

[0524] The server uses a web crawler to collect data from news sites and academic paper databases. The input is the URL of the specified data source and a search query. The server accesses these URLs and retrieves the HTML content. It then uses an HTML parser to extract the necessary information (article title, body text, publication date, author, etc.). The extracted data is stored in a database along with metadata. The output is the collected dataset.

[0525] Step 2: Data analysis

[0526] The server analyzes the collected data by applying natural language processing (NLP) and machine learning algorithms. The input is the collected dataset. The server uses NLP tools (e.g., SpaCy, NLTK) to tokenize the text and apply sentiment analysis and topic models. The server scores the reliability of the data using machine learning algorithms (e.g., Random Forest, SVM). Data with low reliability is filtered. The output is the analyzed dataset with reliability ratings.

[0527] Step 3: Security monitoring

[0528] The terminal monitors network traffic and system logs in real time while the generative AI is running. The input is the generative AI's network traffic and system logs. The terminal uses an intrusion detection system (e.g., Snort, Suricata) to detect abnormal patterns and signs of cyber attacks. If an anomaly is detected, the information is sent to the server. The output is information about the detected anomaly.

[0529] Step 4: Problem diagnosis and notification

[0530] The server diagnoses the problem based on the anomaly information received from the terminal. The input is the anomaly information sent from the terminal. The server uses an anomaly detection algorithm (e.g., k-means clustering, isolated forest) to identify the type of anomaly and its scope of impact. For example, it determines whether high frequency access is a DDoS attack. The diagnosis result is notified to the user. The output is the diagnosis result indicating the type of problem and the scope of impact, along with a notification message.

[0531] Step 5: Feedback and revisions

[0532] The server feeds detected misinformation and security risks back to the generative AI. The inputs are the results from problem diagnosis and additional accurate data. The generative AI self-corrects based on the feedback, for example, by instructing it to correct misinformation based on reliable data sources. The user receives notifications from the server and manually corrects the settings and operation of the generative AI model as needed. The output is a corrected generative AI model.

[0533] Step 6: Automatic Recovery

[0534] The server retrains the generative AI's learning model to remove the effects of misinformation and cyberattacks. The input is the dataset used for retraining. The server uses a machine learning algorithm to train new data. During this process, it filters out inaccurate data or data affected by attacks. The device monitors the generative AI's behavior during the retraining process to prevent recurrence. The output is a retrained generative AI model.

[0535] (Application example 1)

[0536] Next, a description will be given of Application 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."

[0537] In content distribution services using generative AI, the training data may contain misinformation or misleading biased information. Furthermore, generative AI systems are vulnerable to cyberattacks, increasing the risk of security vulnerabilities. Ensuring that users receive reliable information and can use services safely is a key challenge.

[0538] 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.

[0539] In this invention, the server includes means for collecting data used by the generation AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing security diagnosis in real time while the generation AI is operating, means for detecting misinformation and security risks and feeding that information back to the generation AI, means for retraining the generation AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for evaluating reliability scores on user terminals and scoring collected content, means for performing natural language processing on the collected data and extracting features, and means for monitoring network traffic on user terminals and detecting abnormal behavior. This allows users to obtain reliable information and ensures the safety of the generation AI system.

[0540] "Generative AI" is artificial intelligence that learns from data and automatically generates content such as text and images.

[0541] A "data collection tool" is a method or device for collecting data from a specified source.

[0542] "Data analysis means" refers to methods or devices for analyzing collected data and detecting misinformation or bias.

[0543] "Security diagnostic means" refers to a method or device for evaluating and diagnosing the security status of a system in real time while the generating AI is operating.

[0544] "Feedback means" refers to a method or device for detecting misinformation or security risks and sending that information back to the generating AI to instruct it to make corrections.

[0545] "Retraining means" refers to methods or devices for updating the learning model of generative AI and eliminating the effects of inaccurate data or cyber attacks.

[0546] A "trust score evaluation means" is a method or device for evaluating collected content and quantifying its trust.

[0547] "Natural language processing" is a technology that allows computers to understand, analyze, and process human language.

[0548] A "feature extraction means" is a method or device for extracting useful information or patterns from data.

[0549] A "network traffic monitor" is a method or device for monitoring communication flows and detecting anomalous patterns or unauthorized behavior.

[0550] This invention is a system that provides safe and reliable information in a content distribution service using generative AI. This system operates in cooperation with three parties: a server, a terminal, and a user.

[0551] 1. Data Collection and Analysis

[0552] The server collects data from specified sources (such as news sites or academic paper databases) using tools such as web crawlers and analyzes the collected data using natural language processing (NLP) techniques. It calculates a reliability score and filters out unreliable data. Specifically, it accesses websites using the requests module and extracts text using BeautifulSoup.

[0553] 2. Security Monitoring and Diagnostics

[0554] The user device monitors network traffic and system logs in real time while the AI ​​generator is running. If any abnormal behavior or patterns are detected, the information is sent to the server, which then uses the information to diagnose the problem. The logging library is used to record network traffic.

[0555] 3. Feedback and Corrections

[0556] The server detects misinformation and security risks and provides feedback to the generation AI. For example, if a user reports a specific article as misinformation, the generation AI's learning data is corrected based on that information. The user receives a notification from the server and can manually correct the generation AI's settings as needed.

[0557] 4. Automatic recovery

[0558] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data or cyberattacks. The retrained model replaces the old model and begins working normally again.

[0559] Specific examples

[0560] For example, to ensure the reliability of medical information, the server collects medical news articles and rates their reliability. The same information collected from multiple reliable sources is given a high reliability score, and less reliable information is filtered out.

[0561] Prompt Sentence Examples

[0562] "Collect news articles from the URL 'https: / / example.com / news1' and calculate their confidence scores. Adjust the generative AI based on user feedback and perform automatic retraining."

[0563] This system allows users to obtain highly reliable information and use generative AI systems in a secure environment.

[0564] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0565] Step 1:

[0566] The server collects data from specified sources by accessing URLs using the requests module and extracting text from HTML using BeautifulSoup, taking a list of URLs to collect as input and producing raw text data as output.

[0567] Step 2:

[0568] The server analyzes the collected text data. It tokenizes the text data using natural language processing (NLP) techniques and extracts features using TfidfVectorizer. The input is the output data from step 1, and the output is a feature vector. It then calculates a reliability score and excludes data that is deemed unreliable.

[0569] Step 3:

[0570] The terminal monitors network traffic and system logs while the generation AI is running. It detects abnormal patterns and frequent access and sends this information to the server. The input is real-time network traffic data, and the output is log information when an abnormality is detected. Specifically, the logging library is used to record logs.

[0571] Step 4:

[0572] The server analyzes the received anomaly log and determines the extent of the impact of the security risk that occurred. It then feeds the results back to the user and the generating AI, correcting the generating AI model as needed. The input is the anomaly log information from Step 3, and the output is the feedback information.

[0573] Step 5:

[0574] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. Specifically, it retrains the model using a new dataset that removes misinformation and security risks. The input is the corrected dataset, and the output is an updated generative AI model.

[0575] Step 6:

[0576] The user checks the feedback information provided by the server and modifies the operation and settings of the generative AI system. At this time, the input is the feedback information from the server, and the output is the modified settings of the generative AI system.

[0577] Through these steps, content distribution services using generative AI are operated with high reliability and security.

[0578] Furthermore, 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.

[0579] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[0580] System Overview

[0581] 1. Data Collection and Analysis

[0582] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0583] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to check for misinformation and bias, analyze trends, and eliminate unreliable data.

[0584] 2. Emotion engine integration

[0585] The device collects emotional data through interactions with the user, extracted from voice tone, facial expressions, text input, and more.

[0586] The emotion engine analyzes the collected emotion data and provides the results as feedback, adjusting the generative AI's responses and output based on the user's emotions.

[0587] 3. Security Monitoring and Diagnostics

[0588] The device monitors the system environment in real time while the generative AI is running, analyzing network traffic and system logs and reporting any abnormalities detected to the server.

[0589] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers countermeasures.

[0590] 4. Feedback and Corrections

[0591] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects and generates highly reliable responses.

[0592] The emotion engine feeds user emotional data back to the generative AI, which then adapts to provide responses based on the user's emotions.

[0593] 5. Automatic Recovery

[0594] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, more reliable data for the retraining process.

[0595] The device will continue to monitor the generating AI during the recovery process and strengthen measures to prevent recurrence.

[0596] Specific examples

[0597] 1. Specific examples of data collection and analysis

[0598] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[0599] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[0600] 2. Specific examples of emotion engines

[0601] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[0602] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[0603] 3. Specific examples of security monitoring

[0604] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[0605] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[0606] 4. Examples of feedback and correction

[0607] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[0608] The user checks the notification from the server and manually corrects the operation of the generating AI to prevent further misinformation from being generated.

[0609] 5. Specific examples of automatic recovery

[0610] The server retrains the learning model of the generated AI to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[0611] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[0612] This system not only reduces the risk of misinformation and cyberattacks from generative AI, but also enables adaptive responses based on user emotions, improving reliability and user experience.

[0613] The processing flow will be explained below.

[0614] Step 1:

[0615] The server collects data from specified data sources (news sites, academic paper databases, etc.) by launching a web crawler, accessing each data source, acquiring data, and storing it in a database with metadata (acquisition date and time, data source, data category).

[0616] Step 2:

[0617] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. This involves tokenizing and contextual analysis to check for misinformation and bias. A reliability score is calculated, and any low scores are added to a warning list.

[0618] Step 3:

[0619] The device monitors the system environment in real time while the generative AI is running, specifically monitoring network traffic and system logs to detect abnormal patterns and signs of cyberattacks using an anomaly detection algorithm.

[0620] Step 4:

[0621] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, a warning will be reported to the server along with a detailed log of the situation.

[0622] Step 5:

[0623] The server analyzes the abnormal data received from the device to determine the type of problem and the extent of its impact. If a DDoS attack is suspected, it identifies the extent of the impact and takes necessary measures.

[0624] Step 6:

[0625] The server generates a report based on the diagnosis results and notifies the user of that information. For example, it generates a report that a DDoS attack has occurred, including details of the attack, the extent of its impact, and recommended countermeasures, and sends it to the user.

[0626] Step 7:

[0627] The user receives notifications from the server and manually adjusts the AI's operations and settings as needed, for example by blocking the IP address of the attacking source.

[0628] Step 8:

[0629] The server retrains the generative AI's learning model, which includes a retraining process to remove the effects of inaccurate data or cyberattacks, and updates the old model with new, more reliable data.

[0630] Step 9:

[0631] The terminal continues to monitor the generation AI in real time during the recovery process to ensure that the abnormality does not recur. Active monitoring is applied as a preventative measure, and if an abnormality is detected again, it is immediately reported to the server.

[0632] Step 10:

[0633] The device collects emotional data through interactions with the user, specifically analyzing the tone of voice during conversations, facial expressions, and text input to quantify the user's emotional state.

[0634] Step 11:

[0635] The emotion engine analyzes the collected emotional data and feeds the analysis results back to the generative AI. If the user is feeling stressed, they provide that information to the generative AI, which then adapts its response.

[0636] Step 12:

[0637] The emotion engine adjusts the generative AI's responses based on the user's emotions: for example, if the user is determined to be relaxed, the generative AI will use a friendly, calming tone.

[0638] This series of processes enables generative AI to reduce misinformation, protect against cyberattacks, and respond in a way that adapts to the user's emotions.

[0639] Example 2

[0640] Next, a description will be given of 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."

[0641] Generative AI relies on its performance to generate information from a wide range of data, but if that data contains misinformation or bias, the quality of the generated information will be reduced. Generative AI systems are also at risk of cyberattacks, making real-time security monitoring essential. Furthermore, there is a demand for improved user experience, making it necessary to respond based on user emotions.

[0642] 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.

[0643] In this invention, the server includes means for collecting data, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is running, means for collecting emotional data using interactions with users, means for analyzing the emotional data and adjusting the generating AI's response, means for detecting misinformation and security risks and feeding that information back to the generating AI, and means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyberattacks. This provides a system that reduces the risk of generating misinformation and cyberattacks by the generating AI and enables responses based on the user's emotions.

[0644] "Data" refers to the information and materials that generative AI uses for learning, and is collected from news sites, academic paper databases, etc.

[0645] "Collection means" refers to the technology or tools used to automatically obtain the required data from the designated data sources.

[0646] "Analytical means" refers to techniques or tools used to analyze collected data and verify whether it contains misinformation or bias.

[0647] "Generative AI" refers to artificial intelligence systems that generate new information or responses based on input data.

[0648] "Means for performing security diagnosis" refers to technologies or tools that monitor the environment in which the generative AI is operating in real time and detect cyber attacks and system anomalies.

[0649] A "user" is someone who interacts with the system and uses its services.

[0650] "Interaction" refers to the exchange of information and operations between a user and a system.

[0651] "Emotional data" is data that indicates the user's emotional state and is extracted from voice tone, facial expressions, text input, and the like.

[0652] "Means for collecting emotional data" refers to techniques or tools for capturing a user's emotional state from voice tone, facial expressions, text input, etc.

[0653] "Means for analyzing emotional data" refers to technology or tools that analyze the user's emotional state based on collected emotional data and provide feedback on the results to the generating AI.

[0654] "Feedback means" refers to technologies or tools that provide the generative AI with information such as detected misinformation, security risks, and user emotional data, allowing it to adjust or modify its response.

[0655] "Means for retraining learning models" refers to techniques or tools for rebuilding the learning models of generative AI using new, reliable data to remove the effects of inaccurate data or cyberattacks.

[0656] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[0657] System Overview

[0658] Data collection and analysis

[0659] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases. The collected data is accompanied by metadata such as date, time, source, and category. The server then analyzes the collected data using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow) to verify whether it contains misinformation or bias. Unreliable data is then removed.

[0660] Specific working example:

[0661] The server crawls the URL "https: / / example-academic-database.com" and downloads the text data and metadata.

[0662] The server uses SpaCy to perform grammatical analysis on the collected data.

[0663] A TensorFlow model is used to calculate a reliability score for the data, and data below a threshold is removed.

[0664] Emotion engine integration

[0665] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or Google Cloud Speech-to-Text API). The collected emotion data is sent to the emotion engine for analysis.

[0666] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[0667] Specific working example:

[0668] Facial expressions are captured using a laptop camera and analyzed using the OpenCV library.

[0669] Convert audio recorded by the microphone into text using the Google Cloud Speech-to-Text API.

[0670] The emotion engine identifies the user's stress level from the tone and rate of the recorded voice and sends that information to the generative AI.

[0671] Security Monitoring and Diagnostics

[0672] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[0673] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers the necessary countermeasures.

[0674] Specific working example:

[0675] The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[0676] The server analyzes the logs for abnormal traffic and finds that frequent requests are coming from a specific IP address, which is then added to a blacklist.

[0677] Feedback and Corrections

[0678] The server feeds detected misinformation and security risks back to the AI ​​generator, which then self-corrects based on this feedback to generate more reliable responses.

[0679] The emotion engine feeds the user's emotional data back to the generative AI, adapting it to provide responses that correspond to the user's emotions.

[0680] Specific working example:

[0681] The server analyzes the output text of the generated AI, detects that it contains incorrect information, and feeds back the error log to the generated AI.

[0682] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, which then adjusts the response it provides to be more relaxing.

[0683] Automatic recovery

[0684] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[0685] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[0686] Specific working example:

[0687] The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the generative AI's learning model.

[0688] The device uses Splunk to monitor the retraining logs in real time to check for any unusual activity.

[0689] Specific examples

[0690] 1. Specific examples of data collection and analysis

[0691] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[0692] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[0693] 2. Specific examples of emotion engines

[0694] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[0695] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[0696] 3. Specific examples of security monitoring

[0697] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[0698] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[0699] 4. Examples of feedback and correction

[0700] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[0701] If the emotion engine detects that the user is feeling stressed, it feeds that data back to the generative AI, which then adapts its responses to help the user relax.

[0702] 5. Specific examples of automatic recovery

[0703] The server retrains the generated AI's learning model to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[0704] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[0705] Prompt Sentence Examples

[0706] "Analyze the latest medical research data using NLP processing and evaluate its reliability."

[0707] "Determine whether the user is stressed and tailor your response accordingly."

[0708] "Detect abnormal network traffic and diagnose DDoS attacks."

[0709] "Detect misinformation and correct it based on reliable data."

[0710] "Retrain your generative AI learning models and restore your systems based on reliable data."

[0711] This system will reduce the risk of misinformation and cyberattacks from generative AI, and will also enable responses based on user emotions, improving reliability and user experience.

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

[0713] Step 1: Data collection

[0714] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases, and the collected data is annotated with metadata such as date, origin, and category.

[0715] Input: A list of URLs for the specified data sources

[0716] Specific example of operation: The server crawls the URL "https: / / example-academic-database.com" and downloads text data and metadata.

[0717] Output: Collected dataset (text data and metadata)

[0718] Step 2: Data analysis

[0719] The data collected by the server is analyzed using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow). The data is verified for misinformation and bias, and unreliable data is eliminated.

[0720] Input: Collected dataset

[0721] Example of how it works: The server uses SpaCy to perform grammatical analysis on the collected data, calculates a reliability score for the data using a TensorFlow model, and deletes data that falls below a threshold.

[0722] Output: A reliable dataset

[0723] Step 3: Collecting emotion data

[0724] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or the Google Cloud Speech-to-Text API).

[0725] Input: User interaction

[0726] Specific example: Facial expressions are captured using a laptop camera and analyzed using the OpenCV library. Audio recorded by a microphone is converted to text using the Google Cloud Speech-to-Text API.

[0727] Output: Collected emotional data (voice tone, facial expression analysis data, text)

[0728] Step 4: Analyze the sentiment data

[0729] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[0730] Input: Collected emotion data

[0731] Example of how it works: The emotion engine identifies the user's stress level from the tone and speed of the recorded voice and sends that information to the generative AI.

[0732] Output: Feedback data to the generation AI (user emotion analysis results)

[0733] Step 5: Security Monitoring

[0734] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[0735] Input: Real-time data from the generative AI system (network traffic, system logs)

[0736] Specific example of operation: The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[0737] Output: Security alerts and anomaly data

[0738] Step 6: Analysis of abnormal data and feedback

[0739] The server analyzes the abnormal data received from the device, diagnoses the type of problem and the extent of its impact, and, if necessary, provides feedback to the generating AI to correct the problem.

[0740] Input: Security alerts and anomaly data

[0741] Specific operation example: The server analyzes the log of abnormal traffic and finds that frequent requests are coming from a specific IP address. The server then adds that IP address to the blacklist.

[0742] Output: Feedback data and system correction instructions for the generative AI

[0743] Step 7: Retrain the learning model

[0744] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[0745] Input: New high-confidence dataset, inaccurate data detection results

[0746] Specific example of operation: The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the learning model of the generated AI.

[0747] Output: Retrained learning model

[0748] Step 8: Monitoring During Retraining

[0749] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[0750] Input: System state data during retraining

[0751] Example of how it works: The device uses Splunk to monitor the retraining logs in real time and check for any abnormal activity.

[0752] Output: Security status and monitoring reports

[0753] The above are the specific processing steps of the program for this system. By explaining in detail the specific operations performed at each step and their inputs and outputs, the specific method of implementing the invention will become clearer.

[0754] (Application example 2)

[0755] Next, a description will be given of Application 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."

[0756] Conventional generative AI systems have issues with insufficient resistance to learning misinformation and cyberattacks, and lack of appropriate countermeasures. Furthermore, they do not take user emotions into consideration and are unable to respond adaptively, which has the problem of not improving the user experience. The present invention aims to solve these issues and provide a system that achieves the reliability and safety of generative AI and an improved user experience.

[0757] 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.

[0758] In this invention, the server includes means for collecting data used by the generating AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is in operation, means for detecting misinformation and security risks and feeding that information back to the generating AI, means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for analyzing and recognizing passenger emotions and adjusting the generating AI's operation based on that information, and means for monitoring the network traffic of the autonomous vehicle, detecting abnormal activity, and taking immediate measures. This improves the reliability and safety of the generating AI and enables adaptive responses based on user emotions.

[0759] "Generative AI" is an artificial intelligence system that automatically generates information based on data and interacts with users.

[0760] A "data collection means" is a machine or software device that collects data for use in generative AI training.

[0761] "Data analysis tools" are algorithms and software tools used to analyze collected data and detect misinformation and bias.

[0762] "Security diagnostic means" is a function for checking the security of the system in real time while the generating AI is operating.

[0763] "Feedback measures" are mechanisms for detecting misinformation and security risks and returning that information to the generating AI.

[0764] "Retraining measures" are methods for retraining the generative AI's learning model to remove the effects of inaccurate data or cyber attacks.

[0765] "Emotion analysis means" refers to equipment or software that analyzes passenger emotions and can identify emotions based on voice, facial expressions, etc.

[0766] The "behavior adjustment means" is a function for adjusting the behavior of the generating AI based on the analyzed emotional data.

[0767] "Network traffic monitoring means" refers to devices or software that monitor communication data from autonomous vehicles, detect abnormal activity, and take immediate action.

[0768] MODE FOR CARRYING OUT THE INVENTION

[0769] Overall system configuration

[0770] This invention builds a system to realize the reliability and safety of self-driving vehicles, as well as to improve the user experience. The system is mainly composed of a server, a terminal, a generative AI, and an emotion engine working together. The roles and specific operations of each are explained below.

[0771] Hardware and software used

[0772] Hardware:

[0773] On-board computer (high performance computing unit)

[0774] Smartphone (Android or iOS)

[0775] In-vehicle camera (high resolution camera)

[0776] software:

[0777] Emotion recognition engine (commonly used API examples: Microsoft Azure Emotion API, Amazon Rekognition)

[0778] Machine learning platform (commonly used framework examples: TensorFlow, PyTorch)

[0779] Natural language processing (general software examples: SpaCy, NLTK)

[0780] Network monitoring tools (common tools examples: Wireshark, Snort)

[0781] System operation explanation

[0782] Data collection and analysis

[0783] The server collects data for the generative AI to use in its training. It uses a web crawler to gather the latest map and traffic information. The collected data is then analyzed through natural language processing (NLP) algorithms to verify its reliability. The server compares the data with reliable data sources such as Google Maps and OpenStreetMap to eliminate misinformation.

[0784] Emotion Recognition and Behavior Regulation

[0785] The devices (in this case, the onboard camera and smartphone) collect passenger facial expressions and vocal tones in real time. Based on this, the emotion recognition engine analyzes the passenger's emotions. The results of this analysis are fed back to the generative AI and used as a basis for adjusting the behavior of the autonomous vehicle.

[0786] For example, if the emotion recognition engine determines that a passenger is feeling stressed, the generative AI will adjust the vehicle's behavior, such as slowing down the vehicle's speed or driving more smoothly.

[0787] Security Monitoring

[0788] The terminal also monitors the autonomous vehicle's network traffic in real time. For example, if an abnormally high number of data requests is coming from a specific IP address, it could be detected as a sign of a cyber attack. This information is then sent to a server, which can then take appropriate measures.

[0789] Feedback and Corrections

[0790] The server feeds misinformation and detected security risks back to the Generator AI, which then self-corrects based on this feedback to generate more reliable responses. The retraining process uses newly collected, reliable data to keep the Generator AI's model up to date.

[0791] Examples and prompts

[0792] Specific examples

[0793] For example, if an onboard camera analyzes a passenger's facial expression and determines that the passenger is anxious, the autonomous vehicle will slow down. Also, if the latest map data contains incorrect information, that information will be immediately fed back to the generating AI, and the route of the autonomous vehicle will be corrected.

[0794] Prompt Sentence Examples

[0795] Prompt: Generate reliable routes based on up-to-date map and traffic data. Adjust responses based on real-time feedback, including sentiment data.

[0796] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0797] Step 1:

[0798] The server uses data collection methods to collect data for the generative AI to use for learning. It uses a web crawler to obtain the latest map data and traffic information from multiple reliable data sources (e.g., Google Maps, OpenStreetMap). The input is the URL of the data source and the data category to be collected, and the output is the collected raw data. The data is accompanied by metadata such as the acquisition date and time, data source, and data category.

[0799] Step 2:

[0800] The server processes the collected data using analytical tools to verify misinformation and bias. It uses natural language processing (NLP) algorithms to perform text analysis of the data and detect inaccurate or biased information. The input is the collected raw data, and the output is verified, reliable data. Specific operations include keyword extraction, theme classification, and reliability assessment.

[0801] Step 3:

[0802] The device uses emotion recognition means to collect passengers' facial expressions and vocal tones in real time. Input data is acquired using the onboard camera and microphone, and analyzed by an emotion recognition engine (API example: Microsoft Azure Emotion API). The input is facial expression data and voice data, and the output is analyzed emotion data. Specific operations include facial recognition, tone analysis, and emotion classification.

[0803] Step 4:

[0804] The terminal uses an action adjustment means based on the analyzed emotional data to adjust the behavior of the generating AI. If it determines that the passenger is feeling stressed, it feeds that information back to the generating AI, which then adjusts the vehicle's speed and driving pattern. The input is emotional data, and the output is adjusted driving instructions. Specifically, it generates and sends speed adjustment commands.

[0805] Step 5:

[0806] The terminal uses security monitoring means to monitor the autonomous vehicle's network traffic in real time. If it detects an abnormal traffic pattern (e.g., a high frequency of data requests), it immediately reports it to the server. The input is network traffic data, and the output is an anomaly detection report. Specifically, it monitors and analyzes data packets and stores abnormal traffic logs.

[0807] Step 6:

[0808] The server uses a feedback mechanism to feed back misinformation and detected security risks to the generative AI. It then reanalyzes the data using an NLP algorithm or machine learning model to update the generative AI's learning model. The input is the feedback information and new data, and the output is an updated generative AI model. Specific operations include retraining the model and evaluating its accuracy.

[0809] Step 7:

[0810] The server uses a retraining method to retrain the generative AI's learning model. It uses new data that has been cleaned of inaccurate data and the effects of cyberattacks to retrain the learning model so that it can generate reliable responses. The input is new, reliable data, and the output is a retrained generative AI model. Specific operations include preparing and cleaning the data and training the model.

[0811] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio 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 audio data.

[0812] 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> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with 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.

[0813] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0814] [Third embodiment]

[0815] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0816] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0817] 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 WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0818] 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.

[0819] 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 audio in accordance with instructions from the processor 46.

[0820] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and 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 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0821] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0822] Fig. 6 shows an example of the 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.

[0823] 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.

[0824] 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.

[0825] In the headset type terminal 314, a 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.

[0826] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0827] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0828] System Overview

[0829] 1. Data Collection and Analysis

[0830] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0831] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to verify whether it contains misinformation or bias. Unreliable data is filtered out, and reliable data is used as training material for the generative AI.

[0832] 2. Security Monitoring and Diagnostics

[0833] The device constantly monitors network traffic and system logs while the generated AI is running.

[0834] If the device detects any abnormal patterns or signs of a cyber attack, it will send that information to a server.

[0835] The server diagnoses the problem based on the information received from the device and notifies the user of the results. For example, if a high frequency of access from a specific IP address is detected, it will diagnose that there is a high possibility of a DDoS attack.

[0836] 3. Feedback and Corrections

[0837] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects based on this feedback, for example by preventing unreliable data from being reused.

[0838] Users receive notifications from the server, review reported issues, and can manually correct the generated AI's operations and settings if necessary.

[0839] 4. Automatic recovery

[0840] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data and cyber attacks.

[0841] The device will continue to monitor the generation AI during the recovery process to prevent any further abnormalities.

[0842] Specific examples

[0843] 1. Specific examples of data collection and analysis

[0844] The server crawls news sites to collect the latest medical information, adds metadata such as the acquisition date and time, data source, and data category to the collected data, and stores it in a data store.

[0845] The server performs NLP processing on the collected data to evaluate the reliability of the medical information. For example, if the same information is collected from multiple reliable sources, the data is assigned a high reliability score.

[0846] 2. Specific examples of security monitoring

[0847] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack.

[0848] The device sends this information to a server, which identifies any abnormal activity and diagnoses whether it is a DDoS attack.

[0849] 3. Examples of feedback and correction

[0850] The server detects misinformation generated by the AI ​​and feeds that information back to the AI. For example, if incorrect medical information is output, the server instructs the AI ​​to correct it based on a reliable data source.

[0851] Users receive feedback and change how the generating AI operates to prevent further misinformation from being generated.

[0852] 4. Specific examples of automatic recovery

[0853] The server retrains the generation AI's learning model and filters out data affected by the DDoS attack.

[0854] The device continuously monitors the operation of the generation AI during the recovery process to ensure that recurrence is prevented.

[0855] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[0856] The processing flow will be explained below.

[0857] Step 1:

[0858] The server launches a web crawler to collect data from specified data sources (news sites, academic paper databases, etc.). Specifically, the crawler accesses each data source, retrieves data via HTML or API, and stores it in a database.

[0859] Step 2:

[0860] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. For example, it tokenizes the acquired text data and applies a reliability evaluation algorithm to detect misinformation and biased data. It calculates a reliability score and adds low-quality data to a warning list.

[0861] Step 3:

[0862] The device monitors the system environment in real time while the generative AI is running, monitoring network traffic and analyzing system logs, and applying anomaly detection algorithms to detect abnormal activity and patterns.

[0863] Step 4:

[0864] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, the device reports a detailed log and the details of the detection to the server.

[0865] Step 5:

[0866] The server analyzes the abnormal data received from the terminal and determines the type of problem and the scope of its impact. For example, it analyzes abnormal access patterns, evaluates whether this is a DDoS attack, and identifies the affected services and data range.

[0867] Step 6:

[0868] The server generates a report based on the diagnostic results and notifies the user of that information. For example, if a DDoS attack is occurring, a report will be sent to the user containing details of the attack, the extent of its impact, and recommended countermeasures.

[0869] Step 7:

[0870] The user checks the notification from the server and manually adjusts the operation and settings of the generated AI as necessary, for example by blocking the IP address from which the unauthorized access originated.

[0871] Step 8:

[0872] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, specifically by filtering out unreliable data and updating the learning model with new, reliable data.

[0873] Step 9:

[0874] The device monitors the operation of the generation AI during the recovery process to ensure that abnormalities do not occur again. For example, it monitors data access and analysis processes during retraining in real time, and immediately reports any abnormalities to the server.

[0875] This series of processes reduces the risk of misinformation and cyberattacks from the generative AI, improving reliability and security.

[0876] Example 1

[0877] Next, a description will be given of 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."

[0878] If the data used by generative AI for training contains false information, there is a risk that false information will be generated. Furthermore, if generative AI becomes the target of a cyberattack, there is a high possibility that the data and the system will be adversely affected. It is necessary to prevent such problems and ensure the reliability and security of generative AI.

[0879] 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.

[0880] In this invention, the server includes means for collecting data used by the generative AI for learning, means for analyzing the reliability of the collected data using natural language processing and machine learning algorithms and verifying misinformation and bias, means for monitoring network traffic and system logs in real time while the generative AI is operating and detecting abnormal patterns and signs of cyberattacks, means for feeding back detected misinformation and security risks to the generative AI and adjusting or retraining the model, and means for retraining the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. This reduces the risk of the generative AI learning misinformation and makes it possible to protect against cyberattacks.

[0881] "Generative AI" is a system that uses artificial intelligence (AI) technology to generate new content and information from data.

[0882] "Data collection methods" are the processes and tools used to obtain the required information from designated data sources.

[0883] "Natural language processing (NLP)" is a technology that enables computers to understand, interpret, and generate natural language.

[0884] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications.

[0885] A "reliability score" is an index used to evaluate the accuracy and reliability of collected data.

[0886] "Network traffic" is the flow of data sent and received over a network.

[0887] A "system log" is a file or database that records system activity and events.

[0888] An "anomalous pattern" refers to a behavior or data trend that deviates from normal behavior.

[0889] A "cyber attack" is a malicious attack on a computer system or network.

[0890] "Feedback means" is a process that returns information to a system or model based on detected information, allowing for improvements or corrections.

[0891] "Retraining" is the process of retraining an existing learning model using new data to improve its performance.

[0892] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[0893] Data collection and analysis

[0894] The server uses a web crawler to collect data from specified data sources (e.g., news sites or academic paper databases). The collected data is then accompanied by metadata such as the acquisition date and time, data source, and author. The server then evaluates the reliability of the collected data using natural language processing (NLP) tools (e.g., SpaCy or NLTK) or machine learning algorithms (e.g., Random Forest or SVM). Data with a low reliability score is filtered and not used as training material for the generative AI. For example, if the same information is obtained from multiple reliable sources, that information is assigned a high reliability score.

[0895] Security Monitoring and Diagnostics

[0896] While the generative AI is running, the device monitors network traffic and system logs in real time. It uses an IDS (intrusion detection system), such as Snort or Suricata. If the device detects an abnormal pattern or signs of a cyber attack, it sends that information to a server. The server diagnoses the problem based on the anomaly detection information received from the device. For example, frequent access from a specific IP address may be diagnosed as a DDoS attack. The diagnosis results are then notified to the user.

[0897] Feedback and Corrections

[0898] The server feeds back the results of the analysis and security diagnosis to the generative AI. For example, if the generative AI learns false information, it will provide feedback so that it can make corrections based on a reliable data source. The user receives notifications from the server and checks the reported issues. If necessary, it is also possible to manually correct the settings and operation of the generative AI model.

[0899] Automatic recovery

[0900] The server retrains the affected Generative AI's learning model to remove the effects of misinformation and cyberattacks. The retraining process uses a new data set that removes the effects of inaccurate data and attacks. The device monitors the Generative AI's behavior during this retraining process to ensure that recurrence is prevented.

[0901] Specific examples

[0902] 1. Specific examples of data collection

[0903] The server uses a web crawler to collect the latest news about the new coronavirus from news sites, and adds metadata to the collected data, such as the article title, text, publication date, and author.

[0904] 2. Specific examples of data analysis

[0905] The server analyzes the collected medical information using NLP tools. As part of natural language processing, key topics and sentiment are extracted. Machine learning models are also used to score reliability, with matching information from multiple reliable sources being given a high reliability score. Information with low reliability is filtered out.

[0906] 3. Specific examples of security monitoring

[0907] The device monitors the network traffic of the generated AI, and if more than 100 requests per second come from a specific IP address, it sends this as an anomaly to the server.

[0908] 4. Examples of Problem Diagnosis and Notification

[0909] The server diagnoses the anomaly, determines that "high frequency access from a specific IP address is likely a DDoS attack," and notifies the user. The user receives this information and takes appropriate measures.

[0910] 5. Examples of feedback and correction

[0911] The server provides feedback to the generative AI model that it contains misinformation and provides reliable alternative information. For example, if the model contains incorrect medical information, the server instructs the generative AI to refer to accurate information from an official medical institution.

[0912] 6. Specific examples of automatic recovery

[0913] The server uses the retraining dataset to retrain the generative AI model, removing data affected by attacks or misinformation, and the device monitors this retraining process to ensure no new anomalies arise.

[0914] Prompt Sentence Examples

[0915] Here are some examples of prompts for generative AI models:

[0916] "Collect medical information from news site A and news site B and rate the reliability of each."

[0917] "Monitor high-frequency access from specific IP addresses and report to the server if any signs of a DDoS attack are detected."

[0918] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

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

[0920] Step 1: Data collection

[0921] The server uses a web crawler to collect data from news sites and academic paper databases. The input is the URL of the specified data source and a search query. The server accesses these URLs and retrieves the HTML content. It then uses an HTML parser to extract the necessary information (article title, body text, publication date, author, etc.). The extracted data is stored in a database along with metadata. The output is the collected dataset.

[0922] Step 2: Data analysis

[0923] The server analyzes the collected data by applying natural language processing (NLP) and machine learning algorithms. The input is the collected dataset. The server uses NLP tools (e.g., SpaCy, NLTK) to tokenize the text and apply sentiment analysis and topic models. The server scores the reliability of the data using machine learning algorithms (e.g., Random Forest, SVM). Data with low reliability is filtered. The output is the analyzed dataset with reliability ratings.

[0924] Step 3: Security monitoring

[0925] The terminal monitors network traffic and system logs in real time while the generative AI is running. The input is the generative AI's network traffic and system logs. The terminal uses an intrusion detection system (e.g., Snort, Suricata) to detect abnormal patterns and signs of cyber attacks. If an anomaly is detected, the information is sent to the server. The output is information about the detected anomaly.

[0926] Step 4: Problem diagnosis and notification

[0927] The server diagnoses the problem based on the anomaly information received from the terminal. The input is the anomaly information sent from the terminal. The server uses an anomaly detection algorithm (e.g., k-means clustering, isolated forest) to identify the type of anomaly and its scope of impact. For example, it determines whether high frequency access is a DDoS attack. The diagnosis result is notified to the user. The output is the diagnosis result indicating the type of problem and the scope of impact, along with a notification message.

[0928] Step 5: Feedback and revisions

[0929] The server feeds detected misinformation and security risks back to the generative AI. The inputs are the results from problem diagnosis and additional accurate data. The generative AI self-corrects based on the feedback, for example, by instructing it to correct misinformation based on reliable data sources. The user receives notifications from the server and manually corrects the settings and operation of the generative AI model as needed. The output is a corrected generative AI model.

[0930] Step 6: Automatic Recovery

[0931] The server retrains the generative AI's learning model to remove the effects of misinformation and cyberattacks. The input is the dataset used for retraining. The server uses a machine learning algorithm to train new data. During this process, it filters out inaccurate data or data affected by attacks. The device monitors the generative AI's behavior during the retraining process to prevent recurrence. The output is a retrained generative AI model.

[0932] (Application example 1)

[0933] Next, a description will be given of Application 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."

[0934] In content distribution services using generative AI, the training data may contain misinformation or misleading biased information. Furthermore, generative AI systems are vulnerable to cyberattacks, increasing the risk of security vulnerabilities. Ensuring that users receive reliable information and can use services safely is a key challenge.

[0935] 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.

[0936] In this invention, the server includes means for collecting data used by the generation AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing security diagnosis in real time while the generation AI is operating, means for detecting misinformation and security risks and feeding that information back to the generation AI, means for retraining the generation AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for evaluating reliability scores on user terminals and scoring collected content, means for performing natural language processing on the collected data and extracting features, and means for monitoring network traffic on user terminals and detecting abnormal behavior. This allows users to obtain reliable information and ensures the safety of the generation AI system.

[0937] "Generative AI" is artificial intelligence that learns from data and automatically generates content such as text and images.

[0938] A "data collection tool" is a method or device for collecting data from a specified source.

[0939] "Data analysis means" refers to methods or devices for analyzing collected data and detecting misinformation or bias.

[0940] "Security diagnostic means" refers to a method or device for evaluating and diagnosing the security status of a system in real time while the generating AI is operating.

[0941] "Feedback means" refers to a method or device for detecting misinformation or security risks and sending that information back to the generating AI to instruct it to make corrections.

[0942] "Retraining means" refers to methods or devices for updating the learning model of generative AI and eliminating the effects of inaccurate data or cyber attacks.

[0943] A "trust score evaluation means" is a method or device for evaluating collected content and quantifying its trust.

[0944] "Natural language processing" is a technology that allows computers to understand, analyze, and process human language.

[0945] A "feature extraction means" is a method or device for extracting useful information or patterns from data.

[0946] A "network traffic monitor" is a method or device for monitoring communication flows and detecting anomalous patterns or unauthorized behavior.

[0947] This invention is a system that provides safe and reliable information in a content distribution service using generative AI. This system operates in cooperation with three parties: a server, a terminal, and a user.

[0948] 1. Data Collection and Analysis

[0949] The server collects data from specified sources (such as news sites or academic paper databases) using tools such as web crawlers and analyzes the collected data using natural language processing (NLP) techniques. It calculates a reliability score and filters out unreliable data. Specifically, it accesses websites using the requests module and extracts text using BeautifulSoup.

[0950] 2. Security Monitoring and Diagnostics

[0951] The user device monitors network traffic and system logs in real time while the AI ​​generator is running. If any abnormal behavior or patterns are detected, the information is sent to the server, which then uses the information to diagnose the problem. The logging library is used to record network traffic.

[0952] 3. Feedback and Corrections

[0953] The server detects misinformation and security risks and provides feedback to the generation AI. For example, if a user reports a specific article as misinformation, the generation AI's learning data is corrected based on that information. The user receives a notification from the server and can manually correct the generation AI's settings as needed.

[0954] 4. Automatic recovery

[0955] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data or cyberattacks. The retrained model replaces the old model and begins working normally again.

[0956] Specific examples

[0957] For example, to ensure the reliability of medical information, the server collects medical news articles and rates their reliability. The same information collected from multiple reliable sources is given a high reliability score, and less reliable information is filtered out.

[0958] Prompt Sentence Examples

[0959] "Collect news articles from the URL 'https: / / example.com / news1' and calculate their confidence scores. Adjust the generative AI based on user feedback and perform automatic retraining."

[0960] This system allows users to obtain highly reliable information and use generative AI systems in a secure environment.

[0961] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0962] Step 1:

[0963] The server collects data from specified sources by accessing URLs using the requests module and extracting text from HTML using BeautifulSoup, taking a list of URLs to collect as input and producing raw text data as output.

[0964] Step 2:

[0965] The server analyzes the collected text data. It tokenizes the text data using natural language processing (NLP) techniques and extracts features using TfidfVectorizer. The input is the output data from step 1, and the output is a feature vector. It then calculates a reliability score and excludes data that is deemed unreliable.

[0966] Step 3:

[0967] The terminal monitors network traffic and system logs while the generation AI is running. It detects abnormal patterns and frequent access and sends this information to the server. The input is real-time network traffic data, and the output is log information when an abnormality is detected. Specifically, the logging library is used to record logs.

[0968] Step 4:

[0969] The server analyzes the received anomaly log and determines the extent of the impact of the security risk that occurred. It then feeds the results back to the user and the generating AI, correcting the generating AI model as needed. The input is the anomaly log information from Step 3, and the output is the feedback information.

[0970] Step 5:

[0971] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. Specifically, it retrains the model using a new dataset that removes misinformation and security risks. The input is the corrected dataset, and the output is an updated generative AI model.

[0972] Step 6:

[0973] The user checks the feedback information provided by the server and modifies the operation and settings of the generative AI system. At this time, the input is the feedback information from the server, and the output is the modified settings of the generative AI system.

[0974] Through these steps, content distribution services using generative AI are operated with high reliability and security.

[0975] Furthermore, 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.

[0976] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[0977] System Overview

[0978] 1. Data Collection and Analysis

[0979] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[0980] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to check for misinformation and bias, analyze trends, and eliminate unreliable data.

[0981] 2. Emotion engine integration

[0982] The device collects emotional data through interactions with the user, extracted from voice tone, facial expressions, text input, and more.

[0983] The emotion engine analyzes the collected emotion data and provides the results as feedback, adjusting the generative AI's responses and output based on the user's emotions.

[0984] 3. Security Monitoring and Diagnostics

[0985] The device monitors the system environment in real time while the generative AI is running, analyzing network traffic and system logs and reporting any abnormalities detected to the server.

[0986] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers countermeasures.

[0987] 4. Feedback and Corrections

[0988] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects and generates highly reliable responses.

[0989] The emotion engine feeds user emotional data back to the generative AI, which then adapts to provide responses based on the user's emotions.

[0990] 5. Automatic Recovery

[0991] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, more reliable data for the retraining process.

[0992] The device will continue to monitor the generating AI during the recovery process and strengthen measures to prevent recurrence.

[0993] Specific examples

[0994] 1. Specific examples of data collection and analysis

[0995] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[0996] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[0997] 2. Specific examples of emotion engines

[0998] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[0999] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[1000] 3. Specific examples of security monitoring

[1001] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[1002] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[1003] 4. Examples of feedback and correction

[1004] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[1005] The user checks the notification from the server and manually corrects the operation of the generating AI to prevent further misinformation from being generated.

[1006] 5. Specific examples of automatic recovery

[1007] The server retrains the learning model of the generated AI to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[1008] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[1009] This system not only reduces the risk of misinformation and cyberattacks from generative AI, but also enables adaptive responses based on user emotions, improving reliability and user experience.

[1010] The processing flow will be explained below.

[1011] Step 1:

[1012] The server collects data from specified data sources (news sites, academic paper databases, etc.) by launching a web crawler, accessing each data source, acquiring data, and storing it in a database with metadata (acquisition date and time, data source, data category).

[1013] Step 2:

[1014] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. This involves tokenizing and contextual analysis to check for misinformation and bias. A reliability score is calculated, and any low scores are added to a warning list.

[1015] Step 3:

[1016] The device monitors the system environment in real time while the generative AI is running, specifically monitoring network traffic and system logs to detect abnormal patterns and signs of cyberattacks using an anomaly detection algorithm.

[1017] Step 4:

[1018] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, a warning will be reported to the server along with a detailed log of the situation.

[1019] Step 5:

[1020] The server analyzes the abnormal data received from the device to determine the type of problem and the extent of its impact. If a DDoS attack is suspected, it identifies the extent of the impact and takes necessary measures.

[1021] Step 6:

[1022] The server generates a report based on the diagnosis results and notifies the user of that information. For example, it generates a report that a DDoS attack has occurred, including details of the attack, the extent of its impact, and recommended countermeasures, and sends it to the user.

[1023] Step 7:

[1024] The user receives notifications from the server and manually adjusts the AI's operations and settings as needed, for example by blocking the IP address of the attacking source.

[1025] Step 8:

[1026] The server retrains the generative AI's learning model, which includes a retraining process to remove the effects of inaccurate data or cyberattacks, and updates the old model with new, more reliable data.

[1027] Step 9:

[1028] The terminal continues to monitor the generation AI in real time during the recovery process to ensure that the abnormality does not recur. Active monitoring is applied as a preventative measure, and if an abnormality is detected again, it is immediately reported to the server.

[1029] Step 10:

[1030] The device collects emotional data through interactions with the user, specifically analyzing the tone of voice during conversations, facial expressions, and text input to quantify the user's emotional state.

[1031] Step 11:

[1032] The emotion engine analyzes the collected emotional data and feeds the analysis results back to the generative AI. If the user is feeling stressed, they provide that information to the generative AI, which then adapts its response.

[1033] Step 12:

[1034] The emotion engine adjusts the generative AI's responses based on the user's emotions: for example, if the user is determined to be relaxed, the generative AI will use a friendly, calming tone.

[1035] This series of processes enables generative AI to reduce misinformation, protect against cyberattacks, and respond in a way that adapts to the user's emotions.

[1036] Example 2

[1037] Next, a description will be given of 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."

[1038] Generative AI relies on its performance to generate information from a wide range of data, but if that data contains misinformation or bias, the quality of the generated information will be reduced. Generative AI systems are also at risk of cyberattacks, making real-time security monitoring essential. Furthermore, there is a demand for improved user experience, making it necessary to respond based on user emotions.

[1039] 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.

[1040] In this invention, the server includes means for collecting data, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is running, means for collecting emotional data using interactions with users, means for analyzing the emotional data and adjusting the generating AI's response, means for detecting misinformation and security risks and feeding that information back to the generating AI, and means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyberattacks. This provides a system that reduces the risk of generating misinformation and cyberattacks by the generating AI and enables responses based on the user's emotions.

[1041] "Data" refers to the information and materials that generative AI uses for learning, and is collected from news sites, academic paper databases, etc.

[1042] "Collection means" refers to the technology or tools used to automatically obtain the required data from the designated data sources.

[1043] "Analytical means" refers to techniques or tools used to analyze collected data and verify whether it contains misinformation or bias.

[1044] "Generative AI" refers to artificial intelligence systems that generate new information or responses based on input data.

[1045] "Means for performing security diagnosis" refers to technologies or tools that monitor the environment in which the generative AI is operating in real time and detect cyber attacks and system anomalies.

[1046] A "user" is someone who interacts with the system and uses its services.

[1047] "Interaction" refers to the exchange of information and operations between a user and a system.

[1048] "Emotional data" is data that indicates the user's emotional state and is extracted from voice tone, facial expressions, text input, and the like.

[1049] "Means for collecting emotional data" refers to techniques or tools for capturing a user's emotional state from voice tone, facial expressions, text input, etc.

[1050] "Means for analyzing emotional data" refers to technology or tools that analyze the user's emotional state based on collected emotional data and provide feedback on the results to the generating AI.

[1051] "Feedback means" refers to technologies or tools that provide the generative AI with information such as detected misinformation, security risks, and user emotional data, allowing it to adjust or modify its response.

[1052] "Means for retraining learning models" refers to techniques or tools for rebuilding the learning models of generative AI using new, reliable data to remove the effects of inaccurate data or cyberattacks.

[1053] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[1054] System Overview

[1055] Data collection and analysis

[1056] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases. The collected data is accompanied by metadata such as date, time, source, and category. The server then analyzes the collected data using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow) to verify whether it contains misinformation or bias. Unreliable data is then removed.

[1057] Specific working example:

[1058] The server crawls the URL "https: / / example-academic-database.com" and downloads the text data and metadata.

[1059] The server uses SpaCy to perform grammatical analysis on the collected data.

[1060] A TensorFlow model is used to calculate a reliability score for the data, and data below a threshold is removed.

[1061] Emotion engine integration

[1062] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or Google Cloud Speech-to-Text API). The collected emotion data is sent to the emotion engine for analysis.

[1063] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[1064] Specific working example:

[1065] Facial expressions are captured using a laptop camera and analyzed using the OpenCV library.

[1066] Convert audio recorded by the microphone into text using the Google Cloud Speech-to-Text API.

[1067] The emotion engine identifies the user's stress level from the tone and rate of the recorded voice and sends that information to the generative AI.

[1068] Security Monitoring and Diagnostics

[1069] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[1070] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers the necessary countermeasures.

[1071] Specific working example:

[1072] The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[1073] The server analyzes the logs for abnormal traffic and finds that frequent requests are coming from a specific IP address, which is then added to a blacklist.

[1074] Feedback and Corrections

[1075] The server feeds detected misinformation and security risks back to the AI ​​generator, which then self-corrects based on this feedback to generate more reliable responses.

[1076] The emotion engine feeds the user's emotional data back to the generative AI, adapting it to provide responses that correspond to the user's emotions.

[1077] Specific working example:

[1078] The server analyzes the output text of the generated AI, detects that it contains incorrect information, and feeds back the error log to the generated AI.

[1079] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, which then adjusts the response it provides to be more relaxing.

[1080] Automatic recovery

[1081] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[1082] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[1083] Specific working example:

[1084] The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the generative AI's learning model.

[1085] The device uses Splunk to monitor the retraining logs in real time to check for any unusual activity.

[1086] Specific examples

[1087] 1. Specific examples of data collection and analysis

[1088] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[1089] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[1090] 2. Specific examples of emotion engines

[1091] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[1092] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[1093] 3. Specific examples of security monitoring

[1094] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[1095] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[1096] 4. Examples of feedback and correction

[1097] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[1098] If the emotion engine detects that the user is feeling stressed, it feeds that data back to the generative AI, which then adapts its responses to help the user relax.

[1099] 5. Specific examples of automatic recovery

[1100] The server retrains the generated AI's learning model to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[1101] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[1102] Prompt Sentence Examples

[1103] "Analyze the latest medical research data using NLP processing and evaluate its reliability."

[1104] "Determine whether the user is stressed and tailor your response accordingly."

[1105] "Detect abnormal network traffic and diagnose DDoS attacks."

[1106] "Detect misinformation and correct it based on reliable data."

[1107] "Retrain your generative AI learning models and restore your systems based on reliable data."

[1108] This system will reduce the risk of misinformation and cyberattacks from generative AI, and will also enable responses based on user emotions, improving reliability and user experience.

[1109] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1110] Step 1: Data collection

[1111] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases, and the collected data is annotated with metadata such as date, origin, and category.

[1112] Input: A list of URLs for the specified data sources

[1113] Specific example of operation: The server crawls the URL "https: / / example-academic-database.com" and downloads text data and metadata.

[1114] Output: Collected dataset (text data and metadata)

[1115] Step 2: Data analysis

[1116] The data collected by the server is analyzed using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow). The data is verified for misinformation and bias, and unreliable data is eliminated.

[1117] Input: Collected dataset

[1118] Example of how it works: The server uses SpaCy to perform grammatical analysis on the collected data, calculates a reliability score for the data using a TensorFlow model, and deletes data that falls below a threshold.

[1119] Output: A reliable dataset

[1120] Step 3: Collecting emotion data

[1121] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or the Google Cloud Speech-to-Text API).

[1122] Input: User interaction

[1123] Specific example: Facial expressions are captured using a laptop camera and analyzed using the OpenCV library. Audio recorded by a microphone is converted to text using the Google Cloud Speech-to-Text API.

[1124] Output: Collected emotional data (voice tone, facial expression analysis data, text)

[1125] Step 4: Analyze the sentiment data

[1126] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[1127] Input: Collected emotion data

[1128] Example of how it works: The emotion engine identifies the user's stress level from the tone and speed of the recorded voice and sends that information to the generative AI.

[1129] Output: Feedback data to the generation AI (user emotion analysis results)

[1130] Step 5: Security Monitoring

[1131] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[1132] Input: Real-time data from the generative AI system (network traffic, system logs)

[1133] Specific example of operation: The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[1134] Output: Security alerts and anomaly data

[1135] Step 6: Analysis of abnormal data and feedback

[1136] The server analyzes the abnormal data received from the device, diagnoses the type of problem and the extent of its impact, and, if necessary, provides feedback to the generating AI to correct the problem.

[1137] Input: Security alerts and anomaly data

[1138] Specific operation example: The server analyzes the log of abnormal traffic and finds that frequent requests are coming from a specific IP address. The server then adds that IP address to the blacklist.

[1139] Output: Feedback data and system correction instructions for the generative AI

[1140] Step 7: Retrain the learning model

[1141] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[1142] Input: New high-confidence dataset, inaccurate data detection results

[1143] Specific example of operation: The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the learning model of the generated AI.

[1144] Output: Retrained learning model

[1145] Step 8: Monitoring During Retraining

[1146] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[1147] Input: System state data during retraining

[1148] Example of how it works: The device uses Splunk to monitor the retraining logs in real time and check for any abnormal activity.

[1149] Output: Security status and monitoring reports

[1150] The above are the specific processing steps of the program for this system. By explaining in detail the specific operations performed at each step and their inputs and outputs, the specific method of implementing the invention will become clearer.

[1151] (Application example 2)

[1152] Next, a description will be given of Application 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."

[1153] Conventional generative AI systems have issues with insufficient resistance to learning misinformation and cyberattacks, and lack of appropriate countermeasures. Furthermore, they do not take user emotions into consideration and are unable to respond adaptively, which has the problem of not improving the user experience. The present invention aims to solve these issues and provide a system that achieves the reliability and safety of generative AI and an improved user experience.

[1154] 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.

[1155] In this invention, the server includes means for collecting data used by the generating AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is in operation, means for detecting misinformation and security risks and feeding that information back to the generating AI, means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for analyzing and recognizing passenger emotions and adjusting the generating AI's operation based on that information, and means for monitoring the network traffic of the autonomous vehicle, detecting abnormal activity, and taking immediate measures. This improves the reliability and safety of the generating AI and enables adaptive responses based on user emotions.

[1156] "Generative AI" is an artificial intelligence system that automatically generates information based on data and interacts with users.

[1157] A "data collection means" is a machine or software device that collects data for use in generative AI training.

[1158] "Data analysis tools" are algorithms and software tools used to analyze collected data and detect misinformation and bias.

[1159] "Security diagnostic means" is a function for checking the security of the system in real time while the generating AI is operating.

[1160] "Feedback measures" are mechanisms for detecting misinformation and security risks and returning that information to the generating AI.

[1161] "Retraining measures" are methods for retraining the generative AI's learning model to remove the effects of inaccurate data or cyber attacks.

[1162] "Emotion analysis means" refers to equipment or software that analyzes passenger emotions and can identify emotions based on voice, facial expressions, etc.

[1163] The "behavior adjustment means" is a function for adjusting the behavior of the generating AI based on the analyzed emotional data.

[1164] "Network traffic monitoring means" refers to devices or software that monitor communication data from autonomous vehicles, detect abnormal activity, and take immediate action.

[1165] MODE FOR CARRYING OUT THE INVENTION

[1166] Overall system configuration

[1167] This invention builds a system to realize the reliability and safety of self-driving vehicles, as well as to improve the user experience. The system is mainly composed of a server, a terminal, a generative AI, and an emotion engine working together. The roles and specific operations of each are explained below.

[1168] Hardware and software used

[1169] Hardware:

[1170] On-board computer (high performance computing unit)

[1171] Smartphone (Android or iOS)

[1172] In-vehicle camera (high resolution camera)

[1173] software:

[1174] Emotion recognition engine (commonly used API examples: Microsoft Azure Emotion API, Amazon Rekognition)

[1175] Machine learning platform (commonly used framework examples: TensorFlow, PyTorch)

[1176] Natural language processing (general software examples: SpaCy, NLTK)

[1177] Network monitoring tools (common tools examples: Wireshark, Snort)

[1178] System operation explanation

[1179] Data collection and analysis

[1180] The server collects data for the generative AI to use in its training. It uses a web crawler to gather the latest map and traffic information. The collected data is then analyzed through natural language processing (NLP) algorithms to verify its reliability. The server compares the data with reliable data sources such as Google Maps and OpenStreetMap to eliminate misinformation.

[1181] Emotion Recognition and Behavior Regulation

[1182] The devices (in this case, the onboard camera and smartphone) collect passenger facial expressions and vocal tones in real time. Based on this, the emotion recognition engine analyzes the passenger's emotions. The results of this analysis are fed back to the generative AI and used as a basis for adjusting the behavior of the autonomous vehicle.

[1183] For example, if the emotion recognition engine determines that a passenger is feeling stressed, the generative AI will adjust the vehicle's behavior, such as slowing down the vehicle's speed or driving more smoothly.

[1184] Security Monitoring

[1185] The terminal also monitors the autonomous vehicle's network traffic in real time. For example, if an abnormally high number of data requests is coming from a specific IP address, it could be detected as a sign of a cyber attack. This information is then sent to a server, which can then take appropriate measures.

[1186] Feedback and Corrections

[1187] The server feeds misinformation and detected security risks back to the Generator AI, which then self-corrects based on this feedback to generate more reliable responses. The retraining process uses newly collected, reliable data to keep the Generator AI's model up to date.

[1188] Examples and prompts

[1189] Specific examples

[1190] For example, if an onboard camera analyzes a passenger's facial expression and determines that the passenger is anxious, the autonomous vehicle will slow down. Also, if the latest map data contains incorrect information, that information will be immediately fed back to the generating AI, and the route of the autonomous vehicle will be corrected.

[1191] Prompt Sentence Examples

[1192] Prompt: Generate reliable routes based on up-to-date map and traffic data. Adjust responses based on real-time feedback, including sentiment data.

[1193] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1194] Step 1:

[1195] The server uses data collection methods to collect data for the generative AI to use for learning. It uses a web crawler to obtain the latest map data and traffic information from multiple reliable data sources (e.g., Google Maps, OpenStreetMap). The input is the URL of the data source and the data category to be collected, and the output is the collected raw data. The data is accompanied by metadata such as the acquisition date and time, data source, and data category.

[1196] Step 2:

[1197] The server processes the collected data using analytical tools to verify misinformation and bias. It uses natural language processing (NLP) algorithms to perform text analysis of the data and detect inaccurate or biased information. The input is the collected raw data, and the output is verified, reliable data. Specific operations include keyword extraction, theme classification, and reliability assessment.

[1198] Step 3:

[1199] The device uses emotion recognition means to collect passengers' facial expressions and vocal tones in real time. Input data is acquired using the onboard camera and microphone, and analyzed by an emotion recognition engine (API example: Microsoft Azure Emotion API). The input is facial expression data and voice data, and the output is analyzed emotion data. Specific operations include facial recognition, tone analysis, and emotion classification.

[1200] Step 4:

[1201] The terminal uses an action adjustment means based on the analyzed emotional data to adjust the behavior of the generating AI. If it determines that the passenger is feeling stressed, it feeds that information back to the generating AI, which then adjusts the vehicle's speed and driving pattern. The input is emotional data, and the output is adjusted driving instructions. Specifically, it generates and sends speed adjustment commands.

[1202] Step 5:

[1203] The terminal uses security monitoring means to monitor the autonomous vehicle's network traffic in real time. If it detects an abnormal traffic pattern (e.g., a high frequency of data requests), it immediately reports it to the server. The input is network traffic data, and the output is an anomaly detection report. Specifically, it monitors and analyzes data packets and stores abnormal traffic logs.

[1204] Step 6:

[1205] The server uses a feedback mechanism to feed back misinformation and detected security risks to the generative AI. It then reanalyzes the data using an NLP algorithm or machine learning model to update the generative AI's learning model. The input is the feedback information and new data, and the output is an updated generative AI model. Specific operations include retraining the model and evaluating its accuracy.

[1206] Step 7:

[1207] The server uses a retraining method to retrain the generative AI's learning model. It uses new data that has been cleaned of inaccurate data and the effects of cyberattacks to retrain the learning model so that it can generate reliable responses. The input is new, reliable data, and the output is a retrained generative AI model. Specific operations include preparing and cleaning the data and training the model.

[1208] 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 audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio 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 audio data.

[1209] 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> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with 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.

[1210] 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.

[1211] [Fourth embodiment]

[1212] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1213] 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.

[1214] 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 WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1215] 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. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1216] 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 audio in accordance with instructions from the processor 46.

[1217] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and 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 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1218] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1219] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the 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.

[1220] Fig. 8 shows an example of the 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.

[1221] 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.

[1222] 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.

[1223] In the robot 414, the processor 46 performs the reception output process. 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.

[1224] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. 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."

[1225] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[1226] System Overview

[1227] 1. Data Collection and Analysis

[1228] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[1229] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to verify whether it contains misinformation or bias. Unreliable data is filtered out, and reliable data is used as training material for the generative AI.

[1230] 2. Security Monitoring and Diagnostics

[1231] The device constantly monitors network traffic and system logs while the generated AI is running.

[1232] If the device detects any abnormal patterns or signs of a cyber attack, it will send that information to a server.

[1233] The server diagnoses the problem based on the information received from the device and notifies the user of the results. For example, if a high frequency of access from a specific IP address is detected, it will diagnose that there is a high possibility of a DDoS attack.

[1234] 3. Feedback and Corrections

[1235] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects based on this feedback, for example by preventing unreliable data from being reused.

[1236] Users receive notifications from the server, review reported issues, and can manually correct the generated AI's operations and settings if necessary.

[1237] 4. Automatic recovery

[1238] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data and cyber attacks.

[1239] The device will continue to monitor the generation AI during the recovery process to prevent any further abnormalities.

[1240] Specific examples

[1241] 1. Specific examples of data collection and analysis

[1242] The server crawls news sites to collect the latest medical information, adds metadata such as the acquisition date and time, data source, and data category to the collected data, and stores it in a data store.

[1243] The server performs NLP processing on the collected data to evaluate the reliability of the medical information. For example, if the same information is collected from multiple reliable sources, the data is assigned a high reliability score.

[1244] 2. Specific examples of security monitoring

[1245] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack.

[1246] The device sends this information to a server, which identifies any abnormal activity and diagnoses whether it is a DDoS attack.

[1247] 3. Examples of feedback and correction

[1248] The server detects misinformation generated by the AI ​​and feeds that information back to the AI. For example, if incorrect medical information is output, the server instructs the AI ​​to correct it based on a reliable data source.

[1249] Users receive feedback and change how the generating AI operates to prevent further misinformation from being generated.

[1250] 4. Specific examples of automatic recovery

[1251] The server retrains the generation AI's learning model and filters out data affected by the DDoS attack.

[1252] The device continuously monitors the operation of the generation AI during the recovery process to ensure that recurrence is prevented.

[1253] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[1254] The processing flow will be explained below.

[1255] Step 1:

[1256] The server launches a web crawler to collect data from specified data sources (news sites, academic paper databases, etc.). Specifically, the crawler accesses each data source, retrieves data via HTML or API, and stores it in a database.

[1257] Step 2:

[1258] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. For example, it tokenizes the acquired text data and applies a reliability evaluation algorithm to detect misinformation and biased data. It calculates a reliability score and adds low-quality data to a warning list.

[1259] Step 3:

[1260] The device monitors the system environment in real time while the generative AI is running, monitoring network traffic and analyzing system logs, and applying anomaly detection algorithms to detect abnormal activity and patterns.

[1261] Step 4:

[1262] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, the device reports a detailed log and the details of the detection to the server.

[1263] Step 5:

[1264] The server analyzes the abnormal data received from the terminal and determines the type of problem and the scope of its impact. For example, it analyzes abnormal access patterns, evaluates whether this is a DDoS attack, and identifies the affected services and data range.

[1265] Step 6:

[1266] The server generates a report based on the diagnostic results and notifies the user of that information. For example, if a DDoS attack is occurring, a report will be sent to the user containing details of the attack, the extent of its impact, and recommended countermeasures.

[1267] Step 7:

[1268] The user checks the notification from the server and manually adjusts the operation and settings of the generated AI as necessary, for example by blocking the IP address from which the unauthorized access originated.

[1269] Step 8:

[1270] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, specifically by filtering out unreliable data and updating the learning model with new, reliable data.

[1271] Step 9:

[1272] The device monitors the operation of the generation AI during the recovery process to ensure that abnormalities do not occur again. For example, it monitors data access and analysis processes during retraining in real time, and immediately reports any abnormalities to the server.

[1273] This series of processes reduces the risk of misinformation and cyberattacks from the generative AI, improving reliability and security.

[1274] Example 1

[1275] Next, a description will be given of 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."

[1276] If the data used by generative AI for training contains false information, there is a risk that false information will be generated. Furthermore, if generative AI becomes the target of a cyberattack, there is a high possibility that the data and the system will be adversely affected. It is necessary to prevent such problems and ensure the reliability and security of generative AI.

[1277] 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.

[1278] In this invention, the server includes means for collecting data used by the generative AI for learning, means for analyzing the reliability of the collected data using natural language processing and machine learning algorithms and verifying misinformation and bias, means for monitoring network traffic and system logs in real time while the generative AI is operating and detecting abnormal patterns and signs of cyberattacks, means for feeding back detected misinformation and security risks to the generative AI and adjusting or retraining the model, and means for retraining the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. This reduces the risk of the generative AI learning misinformation and makes it possible to protect against cyberattacks.

[1279] "Generative AI" is a system that uses artificial intelligence (AI) technology to generate new content and information from data.

[1280] "Data collection methods" are the processes and tools used to obtain the required information from designated data sources.

[1281] "Natural language processing (NLP)" is a technology that enables computers to understand, interpret, and generate natural language.

[1282] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications.

[1283] A "reliability score" is an index used to evaluate the accuracy and reliability of collected data.

[1284] "Network traffic" is the flow of data sent and received over a network.

[1285] A "system log" is a file or database that records system activity and events.

[1286] An "anomalous pattern" refers to a behavior or data trend that deviates from normal behavior.

[1287] A "cyber attack" is a malicious attack on a computer system or network.

[1288] "Feedback means" is a process that returns information to a system or model based on detected information, allowing for improvements or corrections.

[1289] "Retraining" is the process of retraining an existing learning model using new data to improve its performance.

[1290] This invention is a system that prevents generative AI from learning false information and protects it from cyber attacks. This system works in cooperation with a server, a terminal, and a user.

[1291] Data collection and analysis

[1292] The server uses a web crawler to collect data from specified data sources (e.g., news sites or academic paper databases). The collected data is then accompanied by metadata such as the acquisition date and time, data source, and author. The server then evaluates the reliability of the collected data using natural language processing (NLP) tools (e.g., SpaCy or NLTK) or machine learning algorithms (e.g., Random Forest or SVM). Data with a low reliability score is filtered and not used as training material for the generative AI. For example, if the same information is obtained from multiple reliable sources, that information is assigned a high reliability score.

[1293] Security Monitoring and Diagnostics

[1294] While the generative AI is running, the device monitors network traffic and system logs in real time. It uses an IDS (intrusion detection system), such as Snort or Suricata. If the device detects an abnormal pattern or signs of a cyber attack, it sends that information to a server. The server diagnoses the problem based on the anomaly detection information received from the device. For example, frequent access from a specific IP address may be diagnosed as a DDoS attack. The diagnosis results are then notified to the user.

[1295] Feedback and Corrections

[1296] The server feeds back the results of the analysis and security diagnosis to the generative AI. For example, if the generative AI learns false information, it will provide feedback so that it can make corrections based on a reliable data source. The user receives notifications from the server and checks the reported issues. If necessary, it is also possible to manually correct the settings and operation of the generative AI model.

[1297] Automatic recovery

[1298] The server retrains the affected Generative AI's learning model to remove the effects of misinformation and cyberattacks. The retraining process uses a new data set that removes the effects of inaccurate data and attacks. The device monitors the Generative AI's behavior during this retraining process to ensure that recurrence is prevented.

[1299] Specific examples

[1300] 1. Specific examples of data collection

[1301] The server uses a web crawler to collect the latest news about the new coronavirus from news sites, and adds metadata to the collected data, such as the article title, text, publication date, and author.

[1302] 2. Specific examples of data analysis

[1303] The server analyzes the collected medical information using NLP tools. As part of natural language processing, key topics and sentiment are extracted. Machine learning models are also used to score reliability, with matching information from multiple reliable sources being given a high reliability score. Information with low reliability is filtered out.

[1304] 3. Specific examples of security monitoring

[1305] The device monitors the network traffic of the generated AI, and if more than 100 requests per second come from a specific IP address, it sends this as an anomaly to the server.

[1306] 4. Examples of Problem Diagnosis and Notification

[1307] The server diagnoses the anomaly, determines that "high frequency access from a specific IP address is likely a DDoS attack," and notifies the user. The user receives this information and takes appropriate measures.

[1308] 5. Examples of feedback and correction

[1309] The server provides feedback to the generative AI model that it contains misinformation and provides reliable alternative information. For example, if the model contains incorrect medical information, the server instructs the generative AI to refer to accurate information from an official medical institution.

[1310] 6. Specific examples of automatic recovery

[1311] The server uses the retraining dataset to retrain the generative AI model, removing data affected by attacks or misinformation, and the device monitors this retraining process to ensure no new anomalies arise.

[1312] Prompt Sentence Examples

[1313] Here are some examples of prompts for generative AI models:

[1314] "Collect medical information from news site A and news site B and rate the reliability of each."

[1315] "Monitor high-frequency access from specific IP addresses and report to the server if any signs of a DDoS attack are detected."

[1316] This system minimizes the risk of misinformation and cyberattacks from generative AI, thereby increasing reliability and security.

[1317] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1318] Step 1: Data collection

[1319] The server uses a web crawler to collect data from news sites and academic paper databases. The input is the URL of the specified data source and a search query. The server accesses these URLs and retrieves the HTML content. It then uses an HTML parser to extract the necessary information (article title, body text, publication date, author, etc.). The extracted data is stored in a database along with metadata. The output is the collected dataset.

[1320] Step 2: Data analysis

[1321] The server analyzes the collected data by applying natural language processing (NLP) and machine learning algorithms. The input is the collected dataset. The server uses NLP tools (e.g., SpaCy, NLTK) to tokenize the text and apply sentiment analysis and topic models. The server scores the reliability of the data using machine learning algorithms (e.g., Random Forest, SVM). Data with low reliability is filtered. The output is the analyzed dataset with reliability ratings.

[1322] Step 3: Security monitoring

[1323] The terminal monitors network traffic and system logs in real time while the generative AI is running. The input is the generative AI's network traffic and system logs. The terminal uses an intrusion detection system (e.g., Snort, Suricata) to detect abnormal patterns and signs of cyber attacks. If an anomaly is detected, the information is sent to the server. The output is information about the detected anomaly.

[1324] Step 4: Problem diagnosis and notification

[1325] The server diagnoses the problem based on the anomaly information received from the terminal. The input is the anomaly information sent from the terminal. The server uses an anomaly detection algorithm (e.g., k-means clustering, isolated forest) to identify the type of anomaly and its scope of impact. For example, it determines whether high frequency access is a DDoS attack. The diagnosis result is notified to the user. The output is the diagnosis result indicating the type of problem and the scope of impact, along with a notification message.

[1326] Step 5: Feedback and revisions

[1327] The server feeds detected misinformation and security risks back to the generative AI. The inputs are the results from problem diagnosis and additional accurate data. The generative AI self-corrects based on the feedback, for example, by instructing it to correct misinformation based on reliable data sources. The user receives notifications from the server and manually corrects the settings and operation of the generative AI model as needed. The output is a corrected generative AI model.

[1328] Step 6: Automatic Recovery

[1329] The server retrains the generative AI's learning model to remove the effects of misinformation and cyberattacks. The input is the dataset used for retraining. The server uses a machine learning algorithm to train new data. During this process, it filters out inaccurate data or data affected by attacks. The device monitors the generative AI's behavior during the retraining process to prevent recurrence. The output is a retrained generative AI model.

[1330] (Application example 1)

[1331] Next, a description will be given of Application 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."

[1332] In content distribution services using generative AI, the training data may contain misinformation or misleading biased information. Furthermore, generative AI systems are vulnerable to cyberattacks, increasing the risk of security vulnerabilities. Ensuring that users receive reliable information and can use services safely is a key challenge.

[1333] 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.

[1334] In this invention, the server includes means for collecting data used by the generation AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing security diagnosis in real time while the generation AI is operating, means for detecting misinformation and security risks and feeding that information back to the generation AI, means for retraining the generation AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for evaluating reliability scores on user terminals and scoring collected content, means for performing natural language processing on the collected data and extracting features, and means for monitoring network traffic on user terminals and detecting abnormal behavior. This allows users to obtain reliable information and ensures the safety of the generation AI system.

[1335] "Generative AI" is artificial intelligence that learns from data and automatically generates content such as text and images.

[1336] A "data collection tool" is a method or device for collecting data from a specified source.

[1337] "Data analysis means" refers to methods or devices for analyzing collected data and detecting misinformation or bias.

[1338] "Security diagnostic means" refers to a method or device for evaluating and diagnosing the security status of a system in real time while the generating AI is operating.

[1339] "Feedback means" refers to a method or device for detecting misinformation or security risks and sending that information back to the generating AI to instruct it to make corrections.

[1340] "Retraining means" refers to methods or devices for updating the learning model of generative AI and eliminating the effects of inaccurate data or cyber attacks.

[1341] A "trust score evaluation means" is a method or device for evaluating collected content and quantifying its trust.

[1342] "Natural language processing" is a technology that allows computers to understand, analyze, and process human language.

[1343] A "feature extraction means" is a method or device for extracting useful information or patterns from data.

[1344] A "network traffic monitor" is a method or device for monitoring communication flows and detecting anomalous patterns or unauthorized behavior.

[1345] This invention is a system that provides safe and reliable information in a content distribution service using generative AI. This system operates in cooperation with three parties: a server, a terminal, and a user.

[1346] 1. Data Collection and Analysis

[1347] The server collects data from specified sources (such as news sites or academic paper databases) using tools such as web crawlers and analyzes the collected data using natural language processing (NLP) techniques. It calculates a reliability score and filters out unreliable data. Specifically, it accesses websites using the requests module and extracts text using BeautifulSoup.

[1348] 2. Security Monitoring and Diagnostics

[1349] The user device monitors network traffic and system logs in real time while the AI ​​generator is running. If any abnormal behavior or patterns are detected, the information is sent to the server, which then uses the information to diagnose the problem. The logging library is used to record network traffic.

[1350] 3. Feedback and Corrections

[1351] The server detects misinformation and security risks and provides feedback to the generation AI. For example, if a user reports a specific article as misinformation, the generation AI's learning data is corrected based on that information. The user receives a notification from the server and can manually correct the generation AI's settings as needed.

[1352] 4. Automatic recovery

[1353] The server retrains the generative AI's learning model to eliminate the effects of inaccurate data or cyberattacks. The retrained model replaces the old model and begins working normally again.

[1354] Specific examples

[1355] For example, to ensure the reliability of medical information, the server collects medical news articles and rates their reliability. The same information collected from multiple reliable sources is given a high reliability score, and less reliable information is filtered out.

[1356] Prompt Sentence Examples

[1357] "Collect news articles from the URL 'https: / / example.com / news1' and calculate their confidence scores. Adjust the generative AI based on user feedback and perform automatic retraining."

[1358] This system allows users to obtain highly reliable information and use generative AI systems in a secure environment.

[1359] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1360] Step 1:

[1361] The server collects data from specified sources by accessing URLs using the requests module and extracting text from HTML using BeautifulSoup, taking a list of URLs to collect as input and producing raw text data as output.

[1362] Step 2:

[1363] The server analyzes the collected text data. It tokenizes the text data using natural language processing (NLP) techniques and extracts features using TfidfVectorizer. The input is the output data from step 1, and the output is a feature vector. It then calculates a reliability score and excludes data that is deemed unreliable.

[1364] Step 3:

[1365] The terminal monitors network traffic and system logs while the generation AI is running. It detects abnormal patterns and frequent access and sends this information to the server. The input is real-time network traffic data, and the output is log information when an abnormality is detected. Specifically, the logging library is used to record logs.

[1366] Step 4:

[1367] The server analyzes the received anomaly log and determines the extent of the impact of the security risk that occurred. It then feeds the results back to the user and the generating AI, correcting the generating AI model as needed. The input is the anomaly log information from Step 3, and the output is the feedback information.

[1368] Step 5:

[1369] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks. Specifically, it retrains the model using a new dataset that removes misinformation and security risks. The input is the corrected dataset, and the output is an updated generative AI model.

[1370] Step 6:

[1371] The user checks the feedback information provided by the server and modifies the operation and settings of the generative AI system. At this time, the input is the feedback information from the server, and the output is the modified settings of the generative AI system.

[1372] Through these steps, content distribution services using generative AI are operated with high reliability and security.

[1373] Furthermore, 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.

[1374] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[1375] System Overview

[1376] 1. Data Collection and Analysis

[1377] The server collects data from specified data sources (news sites, academic paper databases, etc.) through a web crawler.

[1378] The server analyzes the collected data and uses natural language processing (NLP) and machine learning algorithms to check for misinformation and bias, analyze trends, and eliminate unreliable data.

[1379] 2. Emotion engine integration

[1380] The device collects emotional data through interactions with the user, extracted from voice tone, facial expressions, text input, and more.

[1381] The emotion engine analyzes the collected emotion data and provides the results as feedback, adjusting the generative AI's responses and output based on the user's emotions.

[1382] 3. Security Monitoring and Diagnostics

[1383] The device monitors the system environment in real time while the generative AI is running, analyzing network traffic and system logs and reporting any abnormalities detected to the server.

[1384] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers countermeasures.

[1385] 4. Feedback and Corrections

[1386] The server then feeds back any detected misinformation or security risks to the AI ​​generator, which then self-corrects and generates highly reliable responses.

[1387] The emotion engine feeds user emotional data back to the generative AI, which then adapts to provide responses based on the user's emotions.

[1388] 5. Automatic Recovery

[1389] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, more reliable data for the retraining process.

[1390] The device will continue to monitor the generating AI during the recovery process and strengthen measures to prevent recurrence.

[1391] Specific examples

[1392] 1. Specific examples of data collection and analysis

[1393] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[1394] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[1395] 2. Specific examples of emotion engines

[1396] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[1397] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[1398] 3. Specific examples of security monitoring

[1399] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[1400] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[1401] 4. Examples of feedback and correction

[1402] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[1403] The user checks the notification from the server and manually corrects the operation of the generating AI to prevent further misinformation from being generated.

[1404] 5. Specific examples of automatic recovery

[1405] The server retrains the learning model of the generated AI to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[1406] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[1407] This system not only reduces the risk of misinformation and cyberattacks from generative AI, but also enables adaptive responses based on user emotions, improving reliability and user experience.

[1408] The processing flow will be explained below.

[1409] Step 1:

[1410] The server collects data from specified data sources (news sites, academic paper databases, etc.) by launching a web crawler, accessing each data source, acquiring data, and storing it in a database with metadata (acquisition date and time, data source, data category).

[1411] Step 2:

[1412] The server analyzes the collected data using natural language processing (NLP) and machine learning algorithms. This involves tokenizing and contextual analysis to check for misinformation and bias. A reliability score is calculated, and any low scores are added to a warning list.

[1413] Step 3:

[1414] The device monitors the system environment in real time while the generative AI is running, specifically monitoring network traffic and system logs to detect abnormal patterns and signs of cyberattacks using an anomaly detection algorithm.

[1415] Step 4:

[1416] If an abnormality is detected, the device immediately sends the information to the server. For example, if a high frequency of requests or unauthorized data access is detected, a warning will be reported to the server along with a detailed log of the situation.

[1417] Step 5:

[1418] The server analyzes the abnormal data received from the device to determine the type of problem and the extent of its impact. If a DDoS attack is suspected, it identifies the extent of the impact and takes necessary measures.

[1419] Step 6:

[1420] The server generates a report based on the diagnosis results and notifies the user of that information. For example, it generates a report that a DDoS attack has occurred, including details of the attack, the extent of its impact, and recommended countermeasures, and sends it to the user.

[1421] Step 7:

[1422] The user receives notifications from the server and manually adjusts the AI's operations and settings as needed, for example by blocking the IP address of the attacking source.

[1423] Step 8:

[1424] The server retrains the generative AI's learning model, which includes a retraining process to remove the effects of inaccurate data or cyberattacks, and updates the old model with new, more reliable data.

[1425] Step 9:

[1426] The terminal continues to monitor the generation AI in real time during the recovery process to ensure that the abnormality does not recur. Active monitoring is applied as a preventative measure, and if an abnormality is detected again, it is immediately reported to the server.

[1427] Step 10:

[1428] The device collects emotional data through interactions with the user, specifically analyzing the tone of voice during conversations, facial expressions, and text input to quantify the user's emotional state.

[1429] Step 11:

[1430] The emotion engine analyzes the collected emotional data and feeds the analysis results back to the generative AI. If the user is feeling stressed, they provide that information to the generative AI, which then adapts its response.

[1431] Step 12:

[1432] The emotion engine adjusts the generative AI's responses based on the user's emotions: for example, if the user is determined to be relaxed, the generative AI will use a friendly, calming tone.

[1433] This series of processes enables generative AI to reduce misinformation, protect against cyberattacks, and respond in a way that adapts to the user's emotions.

[1434] Example 2

[1435] Next, a description will be given of Example 2. 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."

[1436] Generative AI relies on its performance to generate information from a wide range of data, but if that data contains misinformation or bias, the quality of the generated information will be reduced. Generative AI systems are also at risk of cyberattacks, making real-time security monitoring essential. Furthermore, there is a demand for improved user experience, making it necessary to respond based on user emotions.

[1437] 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.

[1438] In this invention, the server includes means for collecting data, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is running, means for collecting emotional data using interactions with users, means for analyzing the emotional data and adjusting the generating AI's response, means for detecting misinformation and security risks and feeding that information back to the generating AI, and means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyberattacks. This provides a system that reduces the risk of generating misinformation and cyberattacks by the generating AI and enables responses based on the user's emotions.

[1439] "Data" refers to the information and materials that generative AI uses for learning, and is collected from news sites, academic paper databases, etc.

[1440] "Collection means" refers to the technology or tools used to automatically obtain the required data from the designated data sources.

[1441] "Analytical means" refers to techniques or tools used to analyze collected data and verify whether it contains misinformation or bias.

[1442] "Generative AI" refers to artificial intelligence systems that generate new information or responses based on input data.

[1443] "Means for performing security diagnosis" refers to technologies or tools that monitor the environment in which the generative AI is operating in real time and detect cyber attacks and system anomalies.

[1444] A "user" is someone who interacts with the system and uses its services.

[1445] "Interaction" refers to the exchange of information and operations between a user and a system.

[1446] "Emotional data" is data that indicates the user's emotional state and is extracted from voice tone, facial expressions, text input, and the like.

[1447] "Means for collecting emotional data" refers to techniques or tools for capturing a user's emotional state from voice tone, facial expressions, text input, etc.

[1448] "Means for analyzing emotional data" refers to technology or tools that analyze the user's emotional state based on collected emotional data and provide feedback on the results to the generating AI.

[1449] "Feedback means" refers to technologies or tools that provide the generative AI with information such as detected misinformation, security risks, and user emotional data, allowing it to adjust or modify its response.

[1450] "Means for retraining learning models" refers to techniques or tools for rebuilding the learning models of generative AI using new, reliable data to remove the effects of inaccurate data or cyberattacks.

[1451] This invention is a system that prevents generative AI from learning false information and protects it from cyberattacks. This system incorporates an emotion engine that recognizes the user's emotions, making it possible to improve the user experience. The system works in cooperation with five parties: generative AI, emotion engine, server, terminal, and user.

[1452] System Overview

[1453] Data collection and analysis

[1454] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases. The collected data is accompanied by metadata such as date, time, source, and category. The server then analyzes the collected data using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow) to verify whether it contains misinformation or bias. Unreliable data is then removed.

[1455] Specific working example:

[1456] The server crawls the URL "https: / / example-academic-database.com" and downloads the text data and metadata.

[1457] The server uses SpaCy to perform grammatical analysis on the collected data.

[1458] A TensorFlow model is used to calculate a reliability score for the data, and data below a threshold is removed.

[1459] Emotion engine integration

[1460] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or Google Cloud Speech-to-Text API). The collected emotion data is sent to the emotion engine for analysis.

[1461] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[1462] Specific working example:

[1463] Facial expressions are captured using a laptop camera and analyzed using the OpenCV library.

[1464] Convert audio recorded by the microphone into text using the Google Cloud Speech-to-Text API.

[1465] The emotion engine identifies the user's stress level from the tone and rate of the recorded voice and sends that information to the generative AI.

[1466] Security Monitoring and Diagnostics

[1467] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[1468] The server analyzes the abnormal data received from the device and diagnoses the type of problem and the extent of its impact. For example, if a DDoS attack is suspected, it identifies the extent of the impact and considers the necessary countermeasures.

[1469] Specific working example:

[1470] The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[1471] The server analyzes the logs for abnormal traffic and finds that frequent requests are coming from a specific IP address, which is then added to a blacklist.

[1472] Feedback and Corrections

[1473] The server feeds detected misinformation and security risks back to the AI ​​generator, which then self-corrects based on this feedback to generate more reliable responses.

[1474] The emotion engine feeds the user's emotional data back to the generative AI, adapting it to provide responses that correspond to the user's emotions.

[1475] Specific working example:

[1476] The server analyzes the output text of the generated AI, detects that it contains incorrect information, and feeds back the error log to the generated AI.

[1477] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, which then adjusts the response it provides to be more relaxing.

[1478] Automatic recovery

[1479] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[1480] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[1481] Specific working example:

[1482] The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the generative AI's learning model.

[1483] The device uses Splunk to monitor the retraining logs in real time to check for any unusual activity.

[1484] Specific examples

[1485] 1. Specific examples of data collection and analysis

[1486] The server crawls medical academic paper databases to collect the latest research data, and metadata such as the acquisition date, data source, and data category are added to the collected data.

[1487] The server performs NLP processing on the collected data to assess the reliability of the research data. For example, if research results are similarly reported by multiple reputable research institutions, a high reliability score is assigned.

[1488] 2. Specific examples of emotion engines

[1489] The device acquires emotional data through conversations with the user, identifying emotions from the user's tone of voice, speed, and facial expressions during conversations, and sends this information to the emotion engine.

[1490] If the emotion engine determines that the user is feeling stressed, it feeds that information back to the generation AI, adjusting the response it provides to be more relaxing.

[1491] 3. Specific examples of security monitoring

[1492] The device monitors the network traffic of the generated AI in real time, and if an abnormally high frequency of data requests occurs from a specific IP address, it detects signs of a cyber attack and reports this to the server.

[1493] The server analyzes detailed logs of abnormal activity, determines the extent of the impact of the DDoS attack, and takes measures to minimize damage.

[1494] 4. Examples of feedback and correction

[1495] The server detects misinformation from the AI ​​generator and feeds that information back to the AI ​​generator. For example, if incorrect medical information is detected, the server instructs the AI ​​generator to correct it based on a reliable data source.

[1496] If the emotion engine detects that the user is feeling stressed, it feeds that data back to the generative AI, which then adapts its responses to help the user relax.

[1497] 5. Specific examples of automatic recovery

[1498] The server retrains the generated AI's learning model to eliminate the effects of DDoS attacks and unreliable data, using new, reliable data for retraining.

[1499] The device continuously monitors the behavior of the generated AI while it is being retrained to ensure that abnormalities do not recur.

[1500] Prompt Sentence Examples

[1501] "Analyze the latest medical research data using NLP processing and evaluate its reliability."

[1502] "Determine whether the user is stressed and tailor your response accordingly."

[1503] "Detect abnormal network traffic and diagnose DDoS attacks."

[1504] "Detect misinformation and correct it based on reliable data."

[1505] "Retrain your generative AI learning models and restore your systems based on reliable data."

[1506] This system will reduce the risk of misinformation and cyberattacks from generative AI, and will also enable responses based on user emotions, improving reliability and user experience.

[1507] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1508] Step 1: Data collection

[1509] The server uses a web crawler (e.g., Apache Nutch) to collect data from news sites and academic paper databases, and the collected data is annotated with metadata such as date, origin, and category.

[1510] Input: A list of URLs for the specified data sources

[1511] Specific example of operation: The server crawls the URL "https: / / example-academic-database.com" and downloads text data and metadata.

[1512] Output: Collected dataset (text data and metadata)

[1513] Step 2: Data analysis

[1514] The data collected by the server is analyzed using natural language processing (NLP) tools (e.g., SpaCy) and machine learning algorithms (e.g., TensorFlow). The data is verified for misinformation and bias, and unreliable data is eliminated.

[1515] Input: Collected dataset

[1516] Example of how it works: The server uses SpaCy to perform grammatical analysis on the collected data, calculates a reliability score for the data using a TensorFlow model, and deletes data that falls below a threshold.

[1517] Output: A reliable dataset

[1518] Step 3: Collecting emotion data

[1519] The device collects emotion data through user interaction, using information such as voice tone, facial expressions, and text input (e.g., Python's OpenCV library or the Google Cloud Speech-to-Text API).

[1520] Input: User interaction

[1521] Specific example: Facial expressions are captured using a laptop camera and analyzed using the OpenCV library. Audio recorded by a microphone is converted to text using the Google Cloud Speech-to-Text API.

[1522] Output: Collected emotional data (voice tone, facial expression analysis data, text)

[1523] Step 4: Analyze the sentiment data

[1524] The emotion engine analyzes the collected data and feeds the results back to the generative AI, which then generates a personalized response based on the user's emotions.

[1525] Input: Collected emotion data

[1526] Example of how it works: The emotion engine identifies the user's stress level from the tone and speed of the recorded voice and sends that information to the generative AI.

[1527] Output: Feedback data to the generation AI (user emotion analysis results)

[1528] Step 5: Security Monitoring

[1529] The device monitors the operating environment of the generative AI in real time, analyzing network traffic and system logs to detect anomalies and reporting them to the server (for example, using Splunk or Wireshark).

[1530] Input: Real-time data from the generative AI system (network traffic, system logs)

[1531] Specific example of operation: The device uses Wireshark to monitor network traffic and generates an alert if an abnormally high frequency of data requests occurs.

[1532] Output: Security alerts and anomaly data

[1533] Step 6: Analysis of abnormal data and feedback

[1534] The server analyzes the abnormal data received from the device, diagnoses the type of problem and the extent of its impact, and, if necessary, provides feedback to the generating AI to correct the problem.

[1535] Input: Security alerts and anomaly data

[1536] Specific operation example: The server analyzes the log of abnormal traffic and finds that frequent requests are coming from a specific IP address. The server then adds that IP address to the blacklist.

[1537] Output: Feedback data and system correction instructions for the generative AI

[1538] Step 7: Retrain the learning model

[1539] The server retrains the generative AI's learning model to remove the effects of inaccurate data and cyberattacks, using new, reliable data.

[1540] Input: New high-confidence dataset, inaccurate data detection results

[1541] Specific example of operation: The server rebuilds the TensorFlow model using a new, highly reliable dataset to retrain the learning model of the generated AI.

[1542] Output: Retrained learning model

[1543] Step 8: Monitoring During Retraining

[1544] The terminal will monitor the system during the generation AI recovery process to prevent the abnormality from recurring.

[1545] Input: System state data during retraining

[1546] Example of how it works: The device uses Splunk to monitor the retraining logs in real time and check for any abnormal activity.

[1547] Output: Security status and monitoring reports

[1548] The above are the specific processing steps of the program for this system. By explaining in detail the specific operations performed at each step and their inputs and outputs, the specific method of implementing the invention will become clearer.

[1549] (Application example 2)

[1550] Next, a description will be given of Application Example 2. 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."

[1551] Conventional generative AI systems have issues with insufficient resistance to learning misinformation and cyberattacks, and lack of appropriate countermeasures. Furthermore, they do not take user emotions into consideration and are unable to respond adaptively, which has the problem of not improving the user experience. The present invention aims to solve these issues and provide a system that achieves the reliability and safety of generative AI and an improved user experience.

[1552] 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.

[1553] In this invention, the server includes means for collecting data used by the generating AI for learning, means for analyzing the collected data and verifying misinformation and bias, means for performing real-time security diagnosis while the generating AI is in operation, means for detecting misinformation and security risks and feeding that information back to the generating AI, means for retraining the generating AI's learning model and eliminating the effects of inaccurate data and cyber attacks, means for analyzing and recognizing passenger emotions and adjusting the generating AI's operation based on that information, and means for monitoring the network traffic of the autonomous vehicle, detecting abnormal activity, and taking immediate measures. This improves the reliability and safety of the generating AI and enables adaptive responses based on user emotions.

[1554] "Generative AI" is an artificial intelligence system that automatically generates information based on data and interacts with users.

[1555] A "data collection means" is a machine or software device that collects data for use in generative AI training.

[1556] "Data analysis tools" are algorithms and software tools used to analyze collected data and detect misinformation and bias.

[1557] "Security diagnostic means" is a function for checking the security of the system in real time while the generating AI is operating.

[1558] "Feedback measures" are mechanisms for detecting misinformation and security risks and returning that information to the generating AI.

[1559] "Retraining measures" are methods for retraining the generative AI's learning model to remove the effects of inaccurate data or cyber attacks.

[1560] "Emotion analysis means" refers to equipment or software that analyzes passenger emotions and can identify emotions based on voice, facial expressions, etc.

[1561] The "behavior adjustment means" is a function for adjusting the behavior of the generating AI based on the analyzed emotional data.

[1562] "Network traffic monitoring means" refers to devices or software that monitor communication data from autonomous vehicles, detect abnormal activity, and take immediate action.

[1563] MODE FOR CARRYING OUT THE INVENTION

[1564] Overall system configuration

[1565] This invention builds a system to realize the reliability and safety of self-driving vehicles, as well as to improve the user experience. The system is mainly composed of a server, a terminal, a generative AI, and an emotion engine working together. The roles and specific operations of each are explained below.

[1566] Hardware and software used

[1567] Hardware:

[1568] On-board computer (high performance computing unit)

[1569] Smartphone (Android or iOS)

[1570] In-vehicle camera (high resolution camera)

[1571] software:

[1572] Emotion recognition engine (commonly used API examples: Microsoft Azure Emotion API, Amazon Rekognition)

[1573] Machine learning platform (commonly used framework examples: TensorFlow, PyTorch)

[1574] Natural language processing (general software examples: SpaCy, NLTK)

[1575] Network monitoring tools (common tools examples: Wireshark, Snort)

[1576] System operation explanation

[1577] Data collection and analysis

[1578] The server collects data for the generative AI to use in its training. It uses a web crawler to gather the latest map and traffic information. The collected data is then analyzed through natural language processing (NLP) algorithms to verify its reliability. The server compares the data with reliable data sources such as Google Maps and OpenStreetMap to eliminate misinformation.

[1579] Emotion Recognition and Behavior Regulation

[1580] The devices (in this case, the onboard camera and smartphone) collect passenger facial expressions and vocal tones in real time. Based on this, the emotion recognition engine analyzes the passenger's emotions. The results of this analysis are fed back to the generative AI and used as a basis for adjusting the behavior of the autonomous vehicle.

[1581] For example, if the emotion recognition engine determines that a passenger is feeling stressed, the generative AI will adjust the vehicle's behavior, such as slowing down the vehicle's speed or driving more smoothly.

[1582] Security Monitoring

[1583] The terminal also monitors the autonomous vehicle's network traffic in real time. For example, if an abnormally high number of data requests is coming from a specific IP address, it could be detected as a sign of a cyber attack. This information is then sent to a server, which can then take appropriate measures.

[1584] Feedback and Corrections

[1585] The server feeds misinformation and detected security risks back to the Generator AI, which then self-corrects based on this feedback to generate more reliable responses. The retraining process uses newly collected, reliable data to keep the Generator AI's model up to date.

[1586] Examples and prompts

[1587] Specific examples

[1588] For example, if an onboard camera analyzes a passenger's facial expression and determines that the passenger is anxious, the autonomous vehicle will slow down. Also, if the latest map data contains incorrect information, that information will be immediately fed back to the generating AI, and the route of the autonomous vehicle will be corrected.

[1589] Prompt Sentence Examples

[1590] Prompt: Generate reliable routes based on up-to-date map and traffic data. Adjust responses based on real-time feedback, including sentiment data.

[1591] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1592] Step 1:

[1593] The server uses data collection methods to collect data for the generative AI to use for learning. It uses a web crawler to obtain the latest map data and traffic information from multiple reliable data sources (e.g., Google Maps, OpenStreetMap). The input is the URL of the data source and the data category to be collected, and the output is the collected raw data. The data is accompanied by metadata such as the acquisition date and time, data source, and data category.

[1594] Step 2:

[1595] The server processes the collected data using analytical tools to verify misinformation and bias. It uses natural language processing (NLP) algorithms to perform text analysis of the data and detect inaccurate or biased information. The input is the collected raw data, and the output is verified, reliable data. Specific operations include keyword extraction, theme classification, and reliability assessment.

[1596] Step 3:

[1597] The device uses emotion recognition means to collect passengers' facial expressions and vocal tones in real time. Input data is acquired using the onboard camera and microphone, and analyzed by an emotion recognition engine (API example: Microsoft Azure Emotion API). The input is facial expression data and voice data, and the output is analyzed emotion data. Specific operations include facial recognition, tone analysis, and emotion classification.

[1598] Step 4:

[1599] The terminal uses an action adjustment means based on the analyzed emotional data to adjust the behavior of the generating AI. If it determines that the passenger is feeling stressed, it feeds that information back to the generating AI, which then adjusts the vehicle's speed and driving pattern. The input is emotional data, and the output is adjusted driving instructions. Specifically, it generates and sends speed adjustment commands.

[1600] Step 5:

[1601] The terminal uses security monitoring means to monitor the autonomous vehicle's network traffic in real time. If it detects an abnormal traffic pattern (e.g., a high frequency of data requests), it immediately reports it to the server. The input is network traffic data, and the output is an anomaly detection report. Specifically, it monitors and analyzes data packets and stores abnormal traffic logs.

[1602] Step 6:

[1603] The server uses a feedback mechanism to feed back misinformation and detected security risks to the generative AI. It then reanalyzes the data using an NLP algorithm or machine learning model to update the generative AI's learning model. The input is the feedback information and new data, and the output is an updated generative AI model. Specific operations include retraining the model and evaluating its accuracy.

[1604] Step 7:

[1605] The server uses a retraining method to retrain the generative AI's learning model. It uses new data that has been cleaned of inaccurate data and the effects of cyberattacks to retrain the learning model so that it can generate reliable responses. The input is new, reliable data, and the output is a retrained generative AI model. Specific operations include preparing and cleaning the data and training the model.

[1606] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding 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.

[1607] 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> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with 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.

[1608] 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 robot 414.

[1609] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1610] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1611] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1612] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1613] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1614] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1615] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1616] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1617] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1618] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1619] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1620] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1621] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1622] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1623] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1624] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1625] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1626] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1627] The following is further disclosed regarding the above embodiment.

[1628] (Claim 1)

[1629] A means of collecting data for the generative AI to use in training;

[1630] A means to analyze the collected data and verify misinformation and bias;

[1631] A means of conducting real-time security diagnosis while the generating AI is running;

[1632] A means to detect misinformation and security risks and feed that information back to the generating AI,

[1633] Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks;

[1634] A system including:

[1635] (Claim 2)

[1636] 10. The system of claim 1, further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

[1637] (Claim 3)

[1638] 10. The system of claim 1, further comprising means for monitoring for anomalous network traffic or patterns and storing detected anomalies as a log.

[1639] "Example 1"

[1640] (Claim 1)

[1641] A means of collecting data for the generative AI to use in training;

[1642] The reliability of the collected data will be analyzed using natural language processing and machine learning algorithms to verify misinformation and bias.

[1643] A means of monitoring network traffic and system logs in real time while the generative AI is running to detect abnormal patterns and signs of cyber attacks;

[1644] A means to feed back detected misinformation and security risks to the generative AI and adjust and retrain the model, and

[1645] Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks;

[1646] A system including:

[1647] (Claim 2)

[1648] 10. The system of claim 1, further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

[1649] (Claim 3)

[1650] 10. The system of claim 1, further comprising means for monitoring for anomalous network traffic or patterns and storing detected anomalies as a log.

[1651] "Application Example 1"

[1652] (Claim 1)

[1653] A means of collecting data for the generative AI to use in training;

[1654] A means to analyze the collected data and verify misinformation and bias;

[1655] A means of conducting real-time security diagnosis while the generating AI is running;

[1656] A means to detect misinformation and security risks and feed that information back to the generating AI,

[1657] Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks;

[1658] A means for performing a reliability score evaluation on a user terminal and scoring the collected content;

[1659] A means for applying natural language processing to the collected data and extracting features;

[1660] means for monitoring network traffic of a user terminal and detecting abnormal behavior;

[1661] A system including:

[1662] (Claim 2)

[1663] 10. The system of claim 1, further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

[1664] (Claim 3)

[1665] 10. The system of claim 1, further comprising means for monitoring for anomalous network traffic or patterns and storing detected anomalies as a log.

[1666] "Example 2: Combining Emotion Engines"

[1667] (Claim 1)

[1668] a means of collecting data;

[1669] A means to analyze the collected data and verify misinformation and bias;

[1670] A means of conducting real-time security diagnosis while the generating AI is running;

[1671] means for collecting emotion data using interaction with a user;

[1672] A means of analyzing the emotional data and adjusting the response of the generative AI;

[1673] A means to detect misinformation and security risks and feed that information back to the generating AI,

[1674] Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks;

[1675] A system including:

[1676] (Claim 2)

[1677] 10. The system of claim 1, further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

[1678] (Claim 3)

[1679] 10. The system of claim 1, further comprising means for monitoring for anomalous network traffic or patterns and storing detected anomalies as a log.

[1680] "Application example 2 when combining emotion engines"

[1681] (Claim 1)

[1682] A means of collecting data for the generative AI to use in training;

[1683] A means to analyze the collected data and verify misinformation and bias;

[1684] A means of conducting real-time security diagnosis while the generating AI is running;

[1685] A means to detect misinformation and security risks and feed that information back to the generating AI,

[1686] Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks;

[1687] A means to analyze and recognize passenger emotions and adjust the behavior of the generating AI based on that information;

[1688] A means to monitor the network traffic of autonomous vehicles, detect abnormal activity and take immediate action; and

[1689] A system including:

[1690] (Claim 2)

[1691] 10. The system of claim 1, further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

[1692] (Claim 3)

[1693] 10. The system of claim 1, further comprising means for monitoring for anomalous network traffic or patterns and storing detected anomalies as a log. [Explanation of symbols]

[1694] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting data for the generative AI to use in training; A means to analyze the collected data and verify misinformation and bias; A means of conducting real-time security diagnosis while the generating AI is running; A means to detect misinformation and security risks and feed that information back to the generating AI, Retraining generative AI learning models to eliminate the effects of inaccurate data and cyberattacks; A system including:

2. The system of claim 1 , further comprising means for analyzing security alert and anomaly data to assess the type and scope of the problem.

3. The system of claim 1 further comprising means for monitoring for abnormal network traffic or patterns and storing detected abnormalities as a log.

Images