JP7796833B2 - system - Google Patents

Description

The technology disclosed herein relates to a system.

Patent document 1 discloses a persona chatbot control method executed 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's character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance.

Japanese Patent Application Laid-Open No. 2022-180282

With conventional technology, it was difficult to identify the cause of a decline in AI performance and make appropriate corrections.

The system according to this embodiment aims to diagnose AI performance and make appropriate corrections.

The system according to the embodiment comprises a collection unit, a diagnosis unit, a correction unit, a forking unit, and a management unit. The collection unit collects data. The diagnosis unit diagnoses the AI's performance based on the data collected by the collection unit. The correction unit inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit. The forking unit reconstructs the AI algorithm corrected by the correction unit. The management unit manages the diagnosis results or correction data.

The system according to the embodiment can diagnose AI performance and make appropriate corrections.

1 is a conceptual diagram illustrating an example of the configuration of a data processing system according to a first embodiment. 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. FIG. 10 is a conceptual diagram illustrating an example of the configuration of a data processing system according to a second embodiment. 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. FIG. 10 is a conceptual diagram illustrating an example of the configuration of a data processing system according to a third embodiment. 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. FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. 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. 1 shows an emotion map onto which multiple emotions are mapped. 1 shows an emotion map onto which multiple emotions are mapped.

Below, an example of an embodiment of a system relating to the technology disclosed herein will be described with reference to the accompanying drawings.

First, let me explain the terminology used in the following explanation.

In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Furthermore, a processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose Computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (Tensor Processing Unit).

In the following embodiments, coded random access memory (RAM) is memory in which information is temporarily stored and is used as work memory by the processor.

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

In the following embodiments, a communication I/F (Interface) is an interface that includes a communication processor, an antenna, and the like. The communication I/F controls communication between multiple computers. Examples of communication standards that can be applied to the communication I/F include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

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 A alone, B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and/or B" also applies when three or more things are expressed connected by "and/or."

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

As shown in FIG. 1, the 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.

The data processing device 12 includes a computer 22, a database 24, and a communication I/F 26. The computer 22 includes a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and 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).

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, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

The reception device 38 includes a touch panel 38A and a microphone 38B, and receives user input. The touch panel 38A detects contact with a pointer (e.g., a pen or a finger) to receive user input via the pointer. The microphone 38B detects the user's voice to receive user input via 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 (see Figure 2) acquires the data indicating the user input.

The output device 40 includes a display 40A and a speaker 40B, and presents data to the user by outputting the data in a form that the user can perceive (e.g., 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.

The communication I/F 44 is connected to the 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.

Figure 2 shows an example of the main functions of the data processing device 12 and smart device 14.

As shown in FIG. 2, in the data processing device 12, specific processing 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 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.

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. The identification processing unit 290 can estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

In the smart device 14, the specific processing is performed by the processor 46. The storage 50 stores the specific processing program 60. The specific processing program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as the control unit 46A in accordance with the specific processing program 60 executed on the RAM 48. The smart device 14 also has a data generation model and emotion identification model similar to the data generation model 58 and emotion identification model 59, and can use these models to perform processing similar to that of the specific processing unit 290.

Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain processing results (prediction results, etc.) using the data generation model 58. The data processing device 12 may also be a server device, or a terminal device owned by a user (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of processing by the data processing system 10 according to the first embodiment will be described.

(Example 1)
An AI diagnosis and treatment system according to an embodiment of the present invention is a system designed to prevent biased AI due to learning from noisy teacher data when a generative AI or a general-purpose AI is engaged in a task. This system diagnoses the AI and, if necessary, "treats" it by hard forking or introducing corrective teacher data. For example, data generated when the AI is engaged in a task is collected and the AI's performance is diagnosed based on the collected data. This diagnosis is achieved by the AI having functions similar to a self-diagnostic program or antivirus software. If the AI is found to be learning in a biased manner based on the diagnosis results, corrective teacher data is introduced. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. For example, if the AI is making biased decisions in a specific task, data related to that task is collected and the AI is diagnosed. If the diagnosis results indicate that the AI is learning in a biased manner, corrective teacher data is introduced to correct the AI's learning. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. This mechanism prevents biased learning when the AI is engaged in a task, ensuring optimal performance. For example, when an AI is engaged in customer service work, it collects customer inquiry data and diagnoses the AI. If the AI's response is biased based on the diagnosis results, corrective training data is input to modify the AI's response. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. In this way, by commercializing AI that diagnoses and treats AI, the AI's performance can be kept at an optimal state at all times. This allows the AI diagnosis and treatment system to always maintain the AI's performance at an optimal state.

An AI diagnosis and treatment system according to an embodiment includes a collection unit, a diagnosis unit, a correction unit, a fork unit, and a management unit. The collection unit collects data generated when the AI performs its tasks. For example, the collection unit can collect customer inquiry data generated when the AI performs customer service tasks. The collection unit can also collect inspection data generated when the AI performs product inspection tasks. The collection unit can also collect marketing data generated when the AI performs marketing tasks. The diagnosis unit diagnoses the performance of the AI based on the data collected by the collection unit. For example, the diagnosis unit can diagnose whether the AI is responding unbalancedly in customer service tasks. The diagnosis unit can also diagnose whether the AI is producing unbalanced inspection results in product inspection tasks. The diagnosis unit can also diagnose whether the AI is formulating a unbalanced marketing strategy in marketing tasks. The correction unit inputs training data to be used for correction based on the results of the diagnosis by the diagnosis unit. For example, if the AI is responding unbalancedly in customer service tasks, the correction unit can input training data to correct that response. Furthermore, if the AI produces biased test results in product inspection work, the correction unit can input training data to correct the test results. Furthermore, if the AI formulates a biased marketing strategy in marketing work, the correction unit can input training data to correct the strategy. The forking unit reconstructs the AI algorithm corrected by the correction unit. For example, if the AI shows a significant bias in customer service work, the forking unit can reconstruct the algorithm. Furthermore, if the AI shows a significant bias in product inspection work, the forking unit can reconstruct the algorithm. Furthermore, if the AI shows a significant bias in marketing work, the forking unit can reconstruct the algorithm. The management unit manages diagnostic results and correction data. For example, the management unit can store diagnostic results and correction data on the cloud so that they can be accessed as needed. Furthermore, the management unit can encrypt and store diagnostic results and correction data to ensure security. Furthermore, the management unit can set access permissions for diagnostic results and correction data and perform access control. This allows the AI diagnosis and treatment system according to the embodiment to always maintain optimal AI performance.

The collection unit collects data generated when the AI is engaged in its work. Specifically, it can collect customer inquiry data generated when the AI is engaged in customer service tasks. For example, this data includes text and voice data when customers make inquiries through the AI chatbot, as well as metadata such as the response content and response time. The collection unit can also collect inspection data generated when the AI is engaged in product inspection tasks. In product inspection tasks, when the AI uses image recognition technology to perform visual inspections of products, image data of the inspection results and log data during the inspection process are collected. The collection unit can also collect marketing data generated when the AI is engaged in marketing tasks. Marketing data includes customer purchase history, website browsing history, and customer response data on social media analyzed by the AI. This data is collected in real time and sent to a central database. The collection unit can centrally manage this data and link it to other systems and departments as needed. For example, the collected data can be stored on a cloud server and accessed by the diagnostics department or orthodontics department. Furthermore, the frequency and accuracy of data collection can be adjusted to accommodate specific situations and conditions. This allows the collection unit to collect data efficiently and effectively, improving overall system performance.

The diagnostics department diagnoses the AI's performance based on the data collected by the collection department. Specifically, it can diagnose whether the AI is responding biasedly in customer service tasks. For example, if the AI consistently responds differently to a specific customer segment, it can identify the cause and evaluate the fairness of the responses. The diagnostics department can also diagnose whether the AI is producing biased test results in product inspection tasks. In product inspection tasks, if the AI is applying overly strict standards to a specific product category, it can evaluate the appropriateness of those standards. Furthermore, the diagnostics department can diagnose whether the AI is formulating a biased marketing strategy in marketing tasks. In marketing tasks, if the AI is allocating excessive resources to a specific customer segment, it can evaluate the effectiveness of that strategy. To perform these diagnoses, the diagnostics department performs detailed analysis of the AI's output results and uses statistical methods and machine learning algorithms to detect bias. Furthermore, the diagnostics department can evaluate the AI's performance and identify areas for improvement by comparing it with past data and benchmark data. This allows the diagnostics department to quickly and accurately evaluate the AI's performance and identify necessary improvements.

The correction unit inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit. Specifically, if the AI is responding unbalancedly in customer service operations, training data can be input to correct that response. For example, if the AI is consistently responding differently to a specific customer segment, a balanced dataset can be prepared and the AI can be retrained to correct that response. Furthermore, if the AI is producing biased inspection results in product inspection operations, the correction unit can input training data to correct those inspection results. In product inspection operations, if the AI is applying overly strict standards to a specific product category, a dataset with appropriate standards can be prepared and the AI can be retrained to correct those standards. Furthermore, if the AI is formulating a biased marketing strategy in marketing operations, the correction unit can input training data to correct that strategy. In marketing operations, if the AI is allocating excessive resources to a specific customer segment, a balanced dataset can be prepared and the AI can be retrained to correct that strategy. To perform these correction tasks efficiently, the correction unit selects appropriate datasets and runs appropriate learning processes on the AI. This allows the correction unit to optimize the AI's performance and provide unbiased and fair results.

The forking department reconstructs the AI algorithm corrected by the correction department. Specifically, if the AI shows significant bias in customer service tasks, it can reconstruct the algorithm. For example, if the AI consistently treats a particular customer group differently, it will review the algorithm and redesign it to provide a fairer response. The forking department can also reconstruct an algorithm if the AI shows significant bias in product inspection tasks. In product inspection tasks, if the AI applies overly strict standards to a particular product category, it will review the algorithm and redesign it to have appropriate standards. Furthermore, the forking department can reconstruct an algorithm if the AI shows significant bias in marketing tasks. In marketing tasks, if the AI allocates excessive resources to a particular customer group, it will review the algorithm and redesign it to create a more balanced strategy. To perform these reconstruction tasks efficiently, the forking department uses appropriate algorithm design methods to optimize the AI's performance. This allows the forking department to reconstruct the AI algorithm and provide unbiased and fair results.

The management department manages diagnostic results and correction data. Specifically, it can store diagnostic results and correction data on the cloud, making them accessible as needed. For example, diagnostic results and correction data can be stored in cloud storage, allowing the diagnostic department and correction department to access them at any time. The management department can also encrypt and store diagnostic results and correction data to ensure security. This prevents unauthorized access and leaks of data. Furthermore, the management department can set permissions for accessing diagnostic results and correction data and perform access control. For example, access rights can be granted only to specific users or departments, ensuring data security. The management department implements appropriate data management tools and security measures to perform these management tasks efficiently. This allows the management department to manage diagnostic results and correction data safely and efficiently, improving the reliability and safety of the entire system.

The collection unit can collect data generated when the AI is engaged in work. For example, the collection unit can collect customer inquiry data generated when the AI is engaged in customer service work. The collection unit can also collect inspection data generated when the AI is engaged in product inspection work. Furthermore, the collection unit can collect marketing data generated when the AI is engaged in marketing work. In this way, by collecting data generated when the AI is engaged in work, it is possible to accurately diagnose the performance of the AI. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input data generated when the AI is engaged in work into the AI and have the AI collect the data.

The diagnostic unit has a self-diagnostic program and can periodically check performance. For example, the diagnostic unit can periodically check whether the AI is responding in a biased manner in customer service operations. The diagnostic unit can also periodically check whether the AI is producing biased test results in product inspection operations. Furthermore, the diagnostic unit can also periodically check whether the AI is formulating a biased marketing strategy in marketing operations. In this way, by periodically checking performance, the state of the AI can be constantly understood. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, for example, or may be performed without using AI. For example, the diagnostic unit can input AI performance data into the AI and have the AI perform a performance check.

The correction unit can input training data generated based on past data. For example, if the AI is responding biasedly in customer service work, the correction unit can input training data to correct that response. Furthermore, if the AI is producing biased inspection results in product inspection work, the correction unit can input training data to correct the inspection results. Furthermore, if the AI is formulating a biased marketing strategy in marketing work, the correction unit can input training data to correct that strategy. In this way, by inputting training data generated based on past data, it is possible to correct bias in the AI. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input past data into the AI and have the AI generate and input training data.

The forking unit can perform a hard fork if the AI shows significant bias. For example, if the AI shows significant bias in customer service operations, the forking unit can perform a hard fork to rebuild the algorithm. Furthermore, if the AI shows significant bias in product inspection operations, the forking unit can perform a hard fork to rebuild the algorithm. Furthermore, if the AI shows significant bias in marketing operations, the forking unit can perform a hard fork to rebuild the algorithm. In this way, by performing a hard fork when the AI shows significant bias, the performance of the AI can be optimized. Some or all of the above-mentioned processing in the forking unit may be performed using AI, for example, or may be performed without using AI. For example, the forking unit can input the AI's algorithm into the AI and leave the execution of the hard fork to the AI.

The management unit can store diagnostic results or correction data on the cloud and make it accessible as needed. The management unit can, for example, store diagnostic results or correction data on the cloud and make it accessible as needed. The management unit can also encrypt and store diagnostic results and correction data to ensure security. Furthermore, the management unit can set access permissions for diagnostic results and correction data and perform access control. By storing diagnostic results and correction data on the cloud, it can be accessed as needed. Some or all of the above-mentioned processing in the management unit can be performed using, for example, AI, or without AI. For example, the management unit can input data stored on the cloud into AI and have the AI manage the data.

The collection unit can analyze the AI's past performance data and select an efficient data collection method. For example, the collection unit can concentrate data collection during specific time periods based on past performance data. The collection unit can also analyze past performance data and select an effective data collection method for specific tasks. Furthermore, the collection unit can optimize the frequency and timing of data collection based on past performance data. This enables efficient data collection by selecting the optimal data collection method based on past performance data. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input past performance data into AI and have the AI select the data collection method.

When collecting data, the collection unit can filter the data based on the AI's current work status and areas of interest. For example, the collection unit can collect only data related to the work the AI is currently working on. The collection unit can also prioritize the collection of highly relevant data based on the AI's areas of interest. Furthermore, the collection unit can filter unnecessary data according to the AI's work status and collect data efficiently. This enables efficient data collection by filtering data based on the AI's work status and areas of interest. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input data on the AI's work status and areas of interest into the AI and have the AI perform data filtering.

When collecting data, the collection unit can prioritize collecting highly relevant data by taking into account the AI's geographical location information. For example, if the AI is operating in a specific area, the collection unit can prioritize collecting data related to that area. The collection unit can also filter and collect highly relevant data based on the AI's geographical location information. Furthermore, if the AI is moving, the collection unit can collect data related to its current location in real time. In this way, by collecting data while taking into account the AI's geographical location information, highly relevant data can be collected efficiently. Some or all of the above-mentioned processing in the collection unit may be performed using AI, or may be performed without using AI. For example, the collection unit can input the AI's geographical location information to the AI and have the AI collect data.

When collecting data, the collection unit can analyze the AI's social media activity and collect relevant data. For example, the collection unit can collect data related to topics in which the AI has shown interest on social media. The collection unit can also prioritize the collection of highly relevant data based on the AI's social media activity. Furthermore, the collection unit can collect data related to accounts that the AI follows on social media. In this way, by analyzing the AI's social media activity and collecting data, highly relevant data can be collected efficiently. Some or all of the above-mentioned processing in the collection unit may be performed using AI, or may be performed without using AI, for example. For example, the collection unit can input the AI's social media activity data into the AI and have the AI collect the data.

During diagnosis, the diagnostic unit can adjust the level of detail of the diagnosis based on the importance of the AI. For example, the diagnostic unit can perform a detailed diagnosis for AI with high importance. The diagnostic unit can also perform a brief diagnosis for AI with low importance. Furthermore, the diagnostic unit can adjust the frequency and level of detail of the diagnosis according to the importance of the AI. This allows for efficient diagnosis by adjusting the level of detail of the diagnosis based on the importance of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI importance data to the AI and have the AI adjust the level of detail of the diagnosis.

When making a diagnosis, the diagnostic unit can apply different diagnostic algorithms depending on the AI category. For example, the diagnostic unit can apply a diagnostic algorithm that emphasizes customer satisfaction to customer service AI. The diagnostic unit can also apply a diagnostic algorithm that emphasizes analysis accuracy to data analysis AI. Furthermore, the diagnostic unit can apply a diagnostic algorithm that emphasizes recognition accuracy to image recognition AI. This allows for appropriate diagnosis by applying a diagnostic algorithm depending on the AI category. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI category data into the AI and have the AI apply the diagnostic algorithm.

When diagnosing, the diagnostic unit can determine the priority of diagnosis based on the operation period of the AI. For example, the diagnostic unit can prioritize diagnosis of newly introduced AI. The diagnostic unit can also periodically diagnose AI that has been in operation for a long time. Furthermore, the diagnostic unit can adjust the frequency and priority of diagnosis according to the operation period. This enables efficient diagnosis by determining the priority of diagnosis based on the operation period of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, for example, or may be performed without using AI. For example, the diagnostic unit can input AI operation period data into the AI and have the AI determine the priority of diagnosis.

The diagnostic unit can adjust the order of diagnostics based on the relevance of the AI during diagnosis. For example, the diagnostic unit can prioritize diagnosing AI related to important tasks. The diagnostic unit can also postpone diagnosing AI related to less relevant tasks. Furthermore, the diagnostic unit can adjust the order and frequency of diagnostics according to the relevance of the AI. This allows for efficient diagnosis by adjusting the order of diagnostics based on the relevance of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI relevance data into the AI and have the AI adjust the order of diagnostics.

When performing correction, the correction unit can analyze the AI's past learning data and select an efficient correction method. For example, the correction unit can select an effective correction method for a specific task based on past learning data. The correction unit can also analyze past learning data and adjust the correction method based on specific patterns. Furthermore, the correction unit can optimize the frequency and timing of correction based on past learning data. This enables efficient correction by selecting the optimal correction method based on past learning data. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input past learning data into the AI and have the AI select the correction method.

When making corrections, the correction unit can customize the means of correction based on the AI's current work situation. For example, the correction unit can provide a correction method related to the work the AI is currently working on. The correction unit can also customize the means of correction according to the AI's work situation. Furthermore, the correction unit can also determine the priority of correction based on the AI's work situation. This enables efficient correction by customizing the means of correction based on the AI's work situation. Some or all of the above-mentioned processing in the correction unit may be performed using AI, or may be performed without using AI, for example. For example, the correction unit can input the AI's work situation data into the AI and have the AI customize the means of correction.

When correcting, the correction unit can select the optimal correction method by taking into account the AI's geographical location information. For example, if the AI is working in a specific area, the correction unit can select a correction method related to that area. The correction unit can also select a highly relevant correction method based on the AI's geographical location information. Furthermore, if the AI is moving, the correction unit can provide a correction method related to the current location. In this way, selecting a correction method by taking into account the AI's geographical location information enables highly relevant correction. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input the AI's geographical location information into the AI and have the AI select a correction method.

When correcting, the correction unit can analyze the AI's social media activity and suggest correction measures. For example, the correction unit can provide correction methods related to topics in which the AI has shown interest on social media. The correction unit can also suggest highly relevant correction methods based on the AI's social media activity. Furthermore, the correction unit can also provide correction methods related to accounts that the AI follows on social media. This enables highly relevant correction by analyzing the AI's social media activity and suggesting correction measures. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input the AI's social media activity data into the AI and have the AI execute the suggested correction measures.

When forking, the forking unit can analyze the AI's past performance data and select the optimal forking method. For example, the forking unit can select an effective forking method for a specific task based on past performance data. The forking unit can also analyze past performance data and adjust the forking method based on specific patterns. Furthermore, the forking unit can optimize the frequency and timing of forking based on past performance data. This enables efficient forking by selecting the optimal forking method based on past performance data. Some or all of the above-mentioned processing in the forking unit may be performed using AI, for example, or may be performed without using AI. For example, the forking unit can input past performance data into the AI and have the AI select the forking method.

At the time of forking, the forking unit can customize the forking means based on the AI's current business situation. For example, the forking unit can provide a forking method related to the business the AI is currently working on. The forking unit can also customize the forking means according to the AI's business situation. Furthermore, the forking unit can also determine forking priorities based on the AI's business situation. This enables efficient forking by customizing the forking means based on the AI's business situation. Some or all of the above-mentioned processing in the forking unit may be performed using the AI, for example, or may be performed without using the AI. For example, the forking unit can input the AI's business situation data into the AI and have the AI customize the forking means.

When forking, the forking unit can select the optimal forking method by taking into account the AI's geographical location information. For example, if the AI is operating in a specific area, the forking unit can select a forking method related to that area. The forking unit can also select a highly relevant forking method based on the AI's geographical location information. Furthermore, if the AI is moving, the forking unit can provide a forking method related to the current location. This makes it possible to select a highly relevant forking method by taking into account the AI's geographical location information. Some or all of the above-mentioned processing in the forking unit may be performed using, or without, the AI, for example. For example, the forking unit can input the AI's geographical location information into the AI and have the AI select the forking method.

At the time of forking, the forking unit can analyze the AI's social media activity and suggest a forking method. For example, the forking unit can provide a forking method related to a topic in which the AI has shown interest on social media. The forking unit can also suggest a highly relevant forking method based on the AI's social media activity. Furthermore, the forking unit can also provide a forking method related to accounts that the AI follows on social media. In this way, by analyzing the AI's social media activity and suggesting a forking method, highly relevant forking becomes possible. Some or all of the above-described processing in the forking unit may be performed using, or without, the AI. For example, the forking unit can input the AI's social media activity data into the AI and have the AI execute the suggested forking method.

The management unit can optimize the management algorithm by referencing past management data during management. For example, the management unit can select an effective management algorithm for a specific task based on past management data. The management unit can also analyze past management data and adjust the management algorithm based on specific patterns. Furthermore, the management unit can optimize the frequency and timing of management based on past management data. This enables efficient management by optimizing the management algorithm based on past management data. Some or all of the above-mentioned processing in the management unit may be performed using, or without, AI, for example. For example, the management unit can input past management data into AI and have it optimize the management algorithm.

During management, the management unit can weight the management data based on the operation period of the AI. For example, the management unit can prioritize weighting of management data for newly introduced AI. The management unit can also periodically weight management data for AI that has been operating for a long period of time. Furthermore, the management unit can adjust the weighting of management data according to the operation period. In this way, weighting of management data based on the operation period of the AI enables efficient management. Some or all of the above-mentioned processing in the management unit may be performed using AI, for example, or may be performed without using AI. For example, the management unit can input AI operation period data into the AI and have the AI perform weighting of the management data.

The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows:

The AI diagnosis and treatment system can further include a prediction unit. The prediction unit can predict the future performance of the AI based on the data collected by the collection unit. For example, the prediction unit can predict how the AI will respond in customer service tasks in the future. The prediction unit can also predict what test results the AI will produce in product inspection tasks in the future. Furthermore, the prediction unit can predict what marketing strategies the AI will formulate in marketing tasks in the future. In this way, the prediction unit can predict the future performance of the AI, allowing measures to be taken in advance.

When collecting AI performance data, the collection unit can evaluate the reliability of the data. For example, the collection unit can check the source and generation method of the data and exclude unreliable data. The collection unit can also check the consistency and integrity of the data and filter out abnormal data. Furthermore, the collection unit can evaluate the recency of the data and exclude old data. This allows the collection unit to collect only highly reliable data, thereby improving the accuracy of AI performance diagnosis.

The diagnostic unit can be equipped with an anomaly detection function when diagnosing AI performance. For example, the diagnostic unit can detect an anomaly if the AI deviates from its normal performance. The diagnostic unit can also issue an alert if the AI exhibits abnormal performance in a specific task. Furthermore, the diagnostic unit can perform a detailed analysis of AI performance based on the results of anomaly detection. This allows the diagnostic unit to detect AI anomalies early and take prompt action.

The correction unit can be equipped with a real-time feedback function when correcting the AI's performance. For example, the correction unit can provide real-time feedback to the AI as it performs its tasks, allowing it to make immediate corrections. The correction unit can also receive real-time feedback from the AI during customer service tasks and correct its responses. Furthermore, the correction unit can receive real-time feedback from the AI during product inspection tasks and correct the inspection results. This allows the correction unit to quickly correct the AI's performance by providing feedback in real time.

The fork unit can be equipped with a version management function when reconstructing an AI algorithm. For example, the fork unit can manage different versions of the AI algorithm and revert to a previous version if necessary. The fork unit can also record the change history of the AI algorithm and track the impact of changes. Furthermore, the fork unit can compare versions of the AI algorithm and select the optimal version. In this way, the fork unit can maintain stable performance by managing versions of the AI algorithm.

The management unit can be equipped with a data visualization function when managing diagnostic results and correction data. For example, the management unit can display diagnostic results and correction data in graphs and charts to make them easier to understand visually. The management unit can also visualize data trends and patterns to understand fluctuations in AI performance. Furthermore, the management unit can provide data filtering and sorting functions to quickly obtain necessary information. In this way, the management unit can make AI performance management more efficient by visualizing the data.

The processing flow for Example 1 is briefly explained below.

Step 1: The collection unit collects data generated when the AI is engaged in its work. For example, it collects customer inquiry data generated when the AI is engaged in customer service work, inspection data generated when the AI is engaged in product inspection work, and marketing data generated when the AI is engaged in marketing work.
Step 2: The diagnostic unit diagnoses the performance of the AI based on the data collected by the collection unit. For example, it diagnoses whether the AI is responding in a biased manner in customer service, producing biased test results in product inspection, or formulating a biased marketing strategy in marketing.
Step 3: The correction unit inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit. For example, if the AI is responding unbalancedly in customer service operations, training data is input to correct that response; if the AI is producing unbalanced test results in product inspection operations, training data is input to correct those test results; and if the AI is formulating a biased marketing strategy in marketing operations, training data is input to correct that strategy.
Step 4: The forking unit reconstructs the AI algorithm corrected by the correction unit. For example, if the AI shows significant bias in customer service, product inspection, or marketing tasks, the algorithm is reconstructed.
Step 5: The management department manages the diagnostic results and correction data. For example, the management department stores the diagnostic results and correction data on the cloud and makes them accessible as needed. The management department also encrypts and stores the diagnostic results and correction data to ensure security. Furthermore, the management department sets permissions to access the diagnostic results and correction data and performs access control.

(Example 2)
An AI diagnosis and treatment system according to an embodiment of the present invention is a system designed to prevent biased AI due to learning from noisy teacher data when a generative AI or a general-purpose AI is engaged in a task. This system diagnoses the AI and, if necessary, "treats" it by hard forking or introducing corrective teacher data. For example, data generated when the AI is engaged in a task is collected and the AI's performance is diagnosed based on the collected data. This diagnosis is achieved by the AI having functions similar to a self-diagnostic program or antivirus software. If the AI is found to be learning in a biased manner based on the diagnosis results, corrective teacher data is introduced. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. For example, if the AI is making biased decisions in a specific task, data related to that task is collected and the AI is diagnosed. If the diagnosis results indicate that the AI is learning in a biased manner, corrective teacher data is introduced to correct the AI's learning. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. This mechanism prevents biased learning when the AI is engaged in a task, ensuring optimal performance. For example, when an AI is engaged in customer service work, it collects customer inquiry data and diagnoses the AI. If the AI's response is biased based on the diagnosis results, corrective training data is input to modify the AI's response. Furthermore, if necessary, a hard fork is performed to optimize the AI's performance. In this way, by commercializing AI that diagnoses and treats AI, the AI's performance can be kept at an optimal state at all times. This allows the AI diagnosis and treatment system to always maintain the AI's performance at an optimal state.

An AI diagnosis and treatment system according to an embodiment includes a collection unit, a diagnosis unit, a correction unit, a fork unit, and a management unit. The collection unit collects data generated when the AI performs its tasks. For example, the collection unit can collect customer inquiry data generated when the AI performs customer service tasks. The collection unit can also collect inspection data generated when the AI performs product inspection tasks. The collection unit can also collect marketing data generated when the AI performs marketing tasks. The diagnosis unit diagnoses the performance of the AI based on the data collected by the collection unit. For example, the diagnosis unit can diagnose whether the AI is responding unbalancedly in customer service tasks. The diagnosis unit can also diagnose whether the AI is producing unbalanced inspection results in product inspection tasks. The diagnosis unit can also diagnose whether the AI is formulating a unbalanced marketing strategy in marketing tasks. The correction unit inputs training data to be used for correction based on the results of the diagnosis by the diagnosis unit. For example, if the AI is responding unbalancedly in customer service tasks, the correction unit can input training data to correct that response. Furthermore, if the AI produces biased test results in product inspection work, the correction unit can input training data to correct the test results. Furthermore, if the AI formulates a biased marketing strategy in marketing work, the correction unit can input training data to correct the strategy. The forking unit reconstructs the AI algorithm corrected by the correction unit. For example, if the AI shows a significant bias in customer service work, the forking unit can reconstruct the algorithm. Furthermore, if the AI shows a significant bias in product inspection work, the forking unit can reconstruct the algorithm. Furthermore, if the AI shows a significant bias in marketing work, the forking unit can reconstruct the algorithm. The management unit manages diagnostic results and correction data. For example, the management unit can store diagnostic results and correction data on the cloud so that they can be accessed as needed. Furthermore, the management unit can encrypt and store diagnostic results and correction data to ensure security. Furthermore, the management unit can set access permissions for diagnostic results and correction data and perform access control. This allows the AI diagnosis and treatment system according to the embodiment to always maintain optimal AI performance.

The collection unit collects data generated when the AI is engaged in its work. Specifically, it can collect customer inquiry data generated when the AI is engaged in customer service tasks. For example, this data includes text and voice data when customers make inquiries through the AI chatbot, as well as metadata such as the response content and response time. The collection unit can also collect inspection data generated when the AI is engaged in product inspection tasks. In product inspection tasks, when the AI uses image recognition technology to perform visual inspections of products, image data of the inspection results and log data during the inspection process are collected. The collection unit can also collect marketing data generated when the AI is engaged in marketing tasks. Marketing data includes customer purchase history, website browsing history, and customer response data on social media analyzed by the AI. This data is collected in real time and sent to a central database. The collection unit can centrally manage this data and link it to other systems and departments as needed. For example, the collected data can be stored on a cloud server and accessed by the diagnostics department or orthodontics department. Furthermore, the frequency and accuracy of data collection can be adjusted to accommodate specific situations and conditions. This allows the collection unit to collect data efficiently and effectively, improving overall system performance.

The diagnostics department diagnoses the AI's performance based on the data collected by the collection department. Specifically, it can diagnose whether the AI is responding biasedly in customer service tasks. For example, if the AI consistently responds differently to a specific customer segment, it can identify the cause and evaluate the fairness of the responses. The diagnostics department can also diagnose whether the AI is producing biased test results in product inspection tasks. In product inspection tasks, if the AI is applying overly strict standards to a specific product category, it can evaluate the appropriateness of those standards. Furthermore, the diagnostics department can diagnose whether the AI is formulating a biased marketing strategy in marketing tasks. In marketing tasks, if the AI is allocating excessive resources to a specific customer segment, it can evaluate the effectiveness of that strategy. To perform these diagnoses, the diagnostics department performs detailed analysis of the AI's output results and uses statistical methods and machine learning algorithms to detect bias. Furthermore, the diagnostics department can evaluate the AI's performance and identify areas for improvement by comparing it with past data and benchmark data. This allows the diagnostics department to quickly and accurately evaluate the AI's performance and identify necessary improvements.

The correction unit inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit. Specifically, if the AI is responding unbalancedly in customer service operations, training data can be input to correct that response. For example, if the AI is consistently responding differently to a specific customer segment, a balanced dataset can be prepared and the AI can be retrained to correct that response. Furthermore, if the AI is producing biased inspection results in product inspection operations, the correction unit can input training data to correct those inspection results. In product inspection operations, if the AI is applying overly strict standards to a specific product category, a dataset with appropriate standards can be prepared and the AI can be retrained to correct those standards. Furthermore, if the AI is formulating a biased marketing strategy in marketing operations, the correction unit can input training data to correct that strategy. In marketing operations, if the AI is allocating excessive resources to a specific customer segment, a balanced dataset can be prepared and the AI can be retrained to correct that strategy. To perform these correction tasks efficiently, the correction unit selects appropriate datasets and runs appropriate learning processes on the AI. This allows the correction unit to optimize the AI's performance and provide unbiased and fair results.

The forking department reconstructs the AI algorithm corrected by the correction department. Specifically, if the AI shows significant bias in customer service tasks, it can reconstruct the algorithm. For example, if the AI consistently treats a particular customer group differently, it will review the algorithm and redesign it to provide a fairer response. The forking department can also reconstruct an algorithm if the AI shows significant bias in product inspection tasks. In product inspection tasks, if the AI applies overly strict standards to a particular product category, it will review the algorithm and redesign it to have appropriate standards. Furthermore, the forking department can reconstruct an algorithm if the AI shows significant bias in marketing tasks. In marketing tasks, if the AI allocates excessive resources to a particular customer group, it will review the algorithm and redesign it to create a more balanced strategy. To perform these reconstruction tasks efficiently, the forking department uses appropriate algorithm design methods to optimize the AI's performance. This allows the forking department to reconstruct the AI algorithm and provide unbiased and fair results.

The management department manages diagnostic results and correction data. Specifically, it can store diagnostic results and correction data on the cloud, making them accessible as needed. For example, diagnostic results and correction data can be stored in cloud storage, allowing the diagnostic department and correction department to access them at any time. The management department can also encrypt and store diagnostic results and correction data to ensure security. This prevents unauthorized access and leaks of data. Furthermore, the management department can set permissions for accessing diagnostic results and correction data and perform access control. For example, access rights can be granted only to specific users or departments, ensuring data security. The management department implements appropriate data management tools and security measures to perform these management tasks efficiently. This allows the management department to manage diagnostic results and correction data safely and efficiently, improving the reliability and safety of the entire system.

The collection unit can collect data generated when the AI is engaged in work. For example, the collection unit can collect customer inquiry data generated when the AI is engaged in customer service work. The collection unit can also collect inspection data generated when the AI is engaged in product inspection work. Furthermore, the collection unit can collect marketing data generated when the AI is engaged in marketing work. In this way, by collecting data generated when the AI is engaged in work, it is possible to accurately diagnose the performance of the AI. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input data generated when the AI is engaged in work into the AI and have the AI collect the data.

The diagnostic unit has a self-diagnostic program and can periodically check performance. For example, the diagnostic unit can periodically check whether the AI is responding in a biased manner in customer service operations. The diagnostic unit can also periodically check whether the AI is producing biased test results in product inspection operations. Furthermore, the diagnostic unit can also periodically check whether the AI is formulating a biased marketing strategy in marketing operations. In this way, by periodically checking performance, the state of the AI can be constantly understood. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, for example, or may be performed without using AI. For example, the diagnostic unit can input AI performance data into the AI and have the AI perform a performance check.

The correction unit can input training data generated based on past data. For example, if the AI is responding biasedly in customer service work, the correction unit can input training data to correct that response. Furthermore, if the AI is producing biased inspection results in product inspection work, the correction unit can input training data to correct the inspection results. Furthermore, if the AI is formulating a biased marketing strategy in marketing work, the correction unit can input training data to correct that strategy. In this way, by inputting training data generated based on past data, it is possible to correct bias in the AI. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input past data into the AI and have the AI generate and input training data.

The forking unit can perform a hard fork if the AI shows significant bias. For example, if the AI shows significant bias in customer service operations, the forking unit can perform a hard fork to rebuild the algorithm. Furthermore, if the AI shows significant bias in product inspection operations, the forking unit can perform a hard fork to rebuild the algorithm. Furthermore, if the AI shows significant bias in marketing operations, the forking unit can perform a hard fork to rebuild the algorithm. In this way, by performing a hard fork when the AI shows significant bias, the performance of the AI can be optimized. Some or all of the above-mentioned processing in the forking unit may be performed using AI, for example, or may be performed without using AI. For example, the forking unit can input the AI's algorithm into the AI and leave the execution of the hard fork to the AI.

The management unit can store diagnostic results or correction data on the cloud and make it accessible as needed. The management unit can, for example, store diagnostic results or correction data on the cloud and make it accessible as needed. The management unit can also encrypt and store diagnostic results and correction data to ensure security. Furthermore, the management unit can set access permissions for diagnostic results and correction data and perform access control. By storing diagnostic results and correction data on the cloud, it can be accessed as needed. Some or all of the above-mentioned processing in the management unit can be performed using, for example, AI, or without AI. For example, the management unit can input data stored on the cloud into AI and have the AI manage the data.

The collection unit can estimate the user's emotions and adjust the timing of data collection based on the estimated user emotions. For example, if the user is feeling stressed, the collection unit can reduce the frequency of data collection to reduce the user's burden. The collection unit can also increase the frequency of data collection to collect more detailed data if the user is relaxed. Furthermore, if the user is in a hurry, the collection unit can shorten the timing of data collection to collect data more quickly. This reduces the user's burden by adjusting the timing of data collection based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-described processing in the collection unit may be performed using AI, or may be performed without AI. For example, the collection unit can input the user's emotion data into AI and have the AI adjust the timing of data collection.

The collection unit can analyze the AI's past performance data and select an efficient data collection method. For example, the collection unit can concentrate data collection during specific time periods based on past performance data. The collection unit can also analyze past performance data and select an effective data collection method for specific tasks. Furthermore, the collection unit can optimize the frequency and timing of data collection based on past performance data. This enables efficient data collection by selecting the optimal data collection method based on past performance data. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input past performance data into AI and have the AI select the data collection method.

When collecting data, the collection unit can filter the data based on the AI's current work status and areas of interest. For example, the collection unit can collect only data related to the work the AI is currently working on. The collection unit can also prioritize the collection of highly relevant data based on the AI's areas of interest. Furthermore, the collection unit can filter unnecessary data according to the AI's work status and collect data efficiently. This enables efficient data collection by filtering data based on the AI's work status and areas of interest. Some or all of the above-mentioned processing in the collection unit may be performed using AI, for example, or may be performed without using AI. For example, the collection unit can input data on the AI's work status and areas of interest into the AI and have the AI perform data filtering.

The collection unit can estimate the user's emotions and determine the priority of data to be collected based on the estimated user emotions. For example, if the user is feeling stressed, the collection unit can postpone the collection of less important data. The collection unit can also prioritize the collection of detailed data if the user is relaxed. Furthermore, if the user is in a hurry, the collection unit can prioritize the collection of more important data. In this way, by determining the priority of data based on the user's emotions, important data can be collected preferentially. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the collection unit may be performed using AI, or may be performed without AI. For example, the collection unit can input the user's emotion data into AI and have the AI determine the priority of the data.

When collecting data, the collection unit can prioritize collecting highly relevant data by taking into account the AI's geographical location information. For example, if the AI is operating in a specific area, the collection unit can prioritize collecting data related to that area. The collection unit can also filter and collect highly relevant data based on the AI's geographical location information. Furthermore, if the AI is moving, the collection unit can collect data related to its current location in real time. In this way, by collecting data while taking into account the AI's geographical location information, highly relevant data can be collected efficiently. Some or all of the above-mentioned processing in the collection unit may be performed using AI, or may be performed without using AI. For example, the collection unit can input the AI's geographical location information to the AI and have the AI collect data.

When collecting data, the collection unit can analyze the AI's social media activity and collect relevant data. For example, the collection unit can collect data related to topics in which the AI has shown interest on social media. The collection unit can also prioritize the collection of highly relevant data based on the AI's social media activity. Furthermore, the collection unit can collect data related to accounts that the AI follows on social media. In this way, by analyzing the AI's social media activity and collecting data, highly relevant data can be collected efficiently. Some or all of the above-mentioned processing in the collection unit may be performed using AI, or may be performed without using AI, for example. For example, the collection unit can input the AI's social media activity data into the AI and have the AI collect the data.

The diagnosis unit can estimate the user's emotions and adjust the way the diagnosis is presented based on the estimated user emotions. For example, if the user is nervous, the diagnosis unit can provide a simple, highly visible diagnosis result. The diagnosis unit can also provide a detailed diagnosis result if the user is relaxed. Furthermore, if the user is in a hurry, the diagnosis unit can provide a concise diagnosis result that focuses on the main points. By adjusting the way the diagnosis is presented based on the user's emotions, it is possible to provide a diagnosis result that is easy for the user to understand. Emotion estimation is achieved using an emotion estimation function, for example, an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the diagnosis unit may be performed using AI, or may be performed without AI. For example, the diagnosis unit can input the user's emotion data into AI and have the AI adjust the way the diagnosis is presented.

The diagnostic unit can adjust the level of detail of the diagnosis based on the importance of the AI during diagnosis. For example, the diagnostic unit can perform a detailed diagnosis for AI with high importance. The diagnostic unit can also perform a brief diagnosis for AI with low importance. Furthermore, the diagnostic unit can adjust the frequency and level of detail of the diagnosis according to the importance of the AI. This allows for efficient diagnosis by adjusting the level of detail of the diagnosis based on the importance of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI importance data to the AI and have the AI adjust the level of detail of the diagnosis.

When making a diagnosis, the diagnostic unit can apply different diagnostic algorithms depending on the AI category. For example, the diagnostic unit can apply a diagnostic algorithm that emphasizes customer satisfaction to customer service AI. The diagnostic unit can also apply a diagnostic algorithm that emphasizes analysis accuracy to data analysis AI. Furthermore, the diagnostic unit can apply a diagnostic algorithm that emphasizes recognition accuracy to image recognition AI. This allows for appropriate diagnosis by applying a diagnostic algorithm depending on the AI category. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI category data into the AI and have the AI apply the diagnostic algorithm.

The diagnosis unit can estimate the user's emotions and adjust the length of the diagnosis based on the estimated user emotions. For example, if the user is in a hurry, the diagnosis unit can provide a short, to-the-point diagnosis. Furthermore, if the user is relaxed, the diagnosis unit can provide a longer diagnosis with detailed explanations. Furthermore, if the user is excited, the diagnosis unit can provide a diagnosis with visually stimulating effects. By adjusting the length of the diagnosis based on the user's emotions, it is possible to provide a diagnosis result that is appropriate for the user. Emotion estimation is achieved using an emotion estimation function, for example, an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-described processing in the diagnosis unit may be performed using AI, or without AI. For example, the diagnosis unit can input the user's emotion data into AI and have the AI adjust the length of the diagnosis.

When diagnosing, the diagnostic unit can determine the priority of diagnosis based on the operation period of the AI. For example, the diagnostic unit can prioritize diagnosis of newly introduced AI. The diagnostic unit can also periodically diagnose AI that has been in operation for a long time. Furthermore, the diagnostic unit can adjust the frequency and priority of diagnosis according to the operation period. This enables efficient diagnosis by determining the priority of diagnosis based on the operation period of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, for example, or may be performed without using AI. For example, the diagnostic unit can input AI operation period data into the AI and have the AI determine the priority of diagnosis.

The diagnostic unit can adjust the order of diagnostics based on the relevance of the AI during diagnosis. For example, the diagnostic unit can prioritize diagnosing AI related to important tasks. The diagnostic unit can also postpone diagnosing AI related to less relevant tasks. Furthermore, the diagnostic unit can adjust the order and frequency of diagnostics according to the relevance of the AI. This allows for efficient diagnosis by adjusting the order of diagnostics based on the relevance of the AI. Some or all of the above-mentioned processing in the diagnostic unit may be performed using AI, or may be performed without using AI, for example. For example, the diagnostic unit can input AI relevance data into the AI and have the AI adjust the order of diagnostics.

The correction unit can estimate the user's emotions and adjust the correction method based on the estimated user emotions. For example, if the user is nervous, the correction unit can provide a simple, highly visible correction method. Furthermore, if the user is relaxed, the correction unit can provide a detailed correction method. Furthermore, if the user is in a hurry, the correction unit can provide a concise correction method that focuses on the main points. By adjusting the correction method based on the user's emotions, it is possible to provide an appropriate correction method for the user. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-described processing in the correction unit may be performed using AI, or may be performed without AI. For example, the correction unit can input the user's emotion data into AI and have the AI adjust the correction method.

When performing correction, the correction unit can analyze the AI's past learning data and select an efficient correction method. For example, the correction unit can select an effective correction method for a specific task based on past learning data. The correction unit can also analyze past learning data and adjust the correction method based on specific patterns. Furthermore, the correction unit can optimize the frequency and timing of correction based on past learning data. This enables efficient correction by selecting the optimal correction method based on past learning data. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input past learning data into the AI and have the AI select the correction method.

When making corrections, the correction unit can customize the means of correction based on the AI's current work situation. For example, the correction unit can provide a correction method related to the work the AI is currently working on. The correction unit can also customize the means of correction according to the AI's work situation. Furthermore, the correction unit can also determine the priority of correction based on the AI's work situation. This enables efficient correction by customizing the means of correction based on the AI's work situation. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input the AI's work situation data into the AI and have the AI customize the means of correction.

The correction unit can estimate the user's emotions and determine the priority of corrections based on the estimated user emotions. For example, if the user is feeling stressed, the correction unit can postpone corrections of less importance. The correction unit can also prioritize detailed corrections if the user is relaxed. Furthermore, if the user is in a hurry, the correction unit can prioritize corrections of greater importance. In this way, by determining the priority of corrections based on the user's emotions, important corrections can be prioritized. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the correction unit may be performed using AI, or may be performed without AI. For example, the correction unit can input the user's emotion data into AI and have the AI determine the priority of corrections.

When correcting, the correction unit can select the optimal correction method by taking into account the AI's geographical location information. For example, if the AI is working in a specific area, the correction unit can select a correction method related to that area. The correction unit can also select a highly relevant correction method based on the AI's geographical location information. Furthermore, if the AI is moving, the correction unit can provide a correction method related to the current location. In this way, selecting a correction method by taking into account the AI's geographical location information enables highly relevant correction. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input the AI's geographical location information into the AI and have the AI select a correction method.

When correcting, the correction unit can analyze the AI's social media activity and suggest correction measures. For example, the correction unit can provide correction methods related to topics in which the AI has shown interest on social media. The correction unit can also suggest highly relevant correction methods based on the AI's social media activity. Furthermore, the correction unit can also provide correction methods related to accounts that the AI follows on social media. This enables highly relevant correction by analyzing the AI's social media activity and suggesting correction measures. Some or all of the above-mentioned processing in the correction unit may be performed using AI, for example, or may be performed without using AI. For example, the correction unit can input the AI's social media activity data into the AI and have the AI execute the suggested correction measures.

The forking unit can estimate the user's emotions and adjust the forking method based on the estimated user emotions. For example, if the user is nervous, the forking unit can provide a simple, highly visible forking method. The forking unit can also provide a detailed forking method if the user is relaxed. Furthermore, if the user is in a hurry, the forking unit can provide a concise forking method that focuses on the main points. By adjusting the forking method based on the user's emotions, it is possible to provide a forking method that is appropriate for the user. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the forking unit may be performed using AI, or may be performed without AI. For example, the forking unit can input user emotion data into AI and have the AI adjust the forking method.

When forking, the forking unit can analyze the AI's past performance data and select the optimal forking method. For example, the forking unit can select an effective forking method for a specific task based on past performance data. The forking unit can also analyze past performance data and adjust the forking method based on specific patterns. Furthermore, the forking unit can optimize the frequency and timing of forking based on past performance data. This enables efficient forking by selecting the optimal forking method based on past performance data. Some or all of the above-mentioned processing in the forking unit may be performed using AI, for example, or may be performed without using AI. For example, the forking unit can input past performance data into the AI and have the AI select the forking method.

At the time of forking, the forking unit can customize the forking means based on the AI's current business situation. For example, the forking unit can provide a forking method related to the business the AI is currently working on. The forking unit can also customize the forking means according to the AI's business situation. Furthermore, the forking unit can also determine forking priorities based on the AI's business situation. This enables efficient forking by customizing the forking means based on the AI's business situation. Some or all of the above-mentioned processing in the forking unit may be performed using the AI, for example, or may be performed without using the AI. For example, the forking unit can input the AI's business situation data into the AI and have the AI customize the forking means.

The fork unit can estimate the user's emotions and determine the priority of forks based on the estimated user emotions. For example, if the user is feeling stressed, the fork unit can postpone forks with lower importance. The fork unit can also prioritize detailed forks if the user is relaxed. Furthermore, if the user is in a hurry, the fork unit can prioritize forks with higher importance. In this way, by determining the priority of forks based on the user's emotions, important forks can be prioritized. Emotion estimation is realized using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the fork unit may be performed using AI, or may be performed without AI. For example, the fork unit can input user emotion data into AI and have the AI determine the priority of forks.

When forking, the forking unit can select the optimal forking method by taking into account the AI's geographical location information. For example, if the AI is operating in a specific area, the forking unit can select a forking method related to that area. The forking unit can also select a highly relevant forking method based on the AI's geographical location information. Furthermore, if the AI is moving, the forking unit can provide a forking method related to the current location. This makes it possible to select a highly relevant forking method by taking into account the AI's geographical location information. Some or all of the above-mentioned processing in the forking unit may be performed using, or without, the AI, for example. For example, the forking unit can input the AI's geographical location information into the AI and have the AI select the forking method.

At the time of forking, the forking unit can analyze the AI's social media activity and suggest a method for forking. For example, the forking unit can provide a forking method related to a topic in which the AI has shown interest on social media. The forking unit can also suggest a highly relevant forking method based on the AI's social media activity. Furthermore, the forking unit can also provide a forking method related to accounts that the AI follows on social media. In this way, by analyzing the AI's social media activity and suggesting a method for forking, highly relevant forking becomes possible. Some or all of the above-described processing in the forking unit may be performed using, or without, the AI. For example, the forking unit can input the AI's social media activity data into the AI and have the AI execute the suggested forking method.

The management unit can estimate the user's emotions and select management data based on the estimated user emotions. For example, if the user is nervous, the management unit can provide simple, highly visible management data. Furthermore, if the user is relaxed, the management unit can provide detailed management data. Furthermore, if the user is in a hurry, the management unit can provide concise management data that focuses on the main points. By selecting management data based on the user's emotions, appropriate management data can be provided for the user. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-described processing in the management unit may be performed using AI, or may be performed without AI. For example, the management unit can input the user's emotion data into AI and have the AI select the management data.

The management unit can optimize the management algorithm by referencing past management data during management. For example, the management unit can select an effective management algorithm for a specific task based on past management data. The management unit can also analyze past management data and adjust the management algorithm based on specific patterns. Furthermore, the management unit can optimize the frequency and timing of management based on past management data. This enables efficient management by optimizing the management algorithm based on past management data. Some or all of the above-mentioned processing in the management unit may be performed using, or without, AI, for example. For example, the management unit can input past management data into AI and have it optimize the management algorithm.

The management unit can estimate the user's emotions and adjust the frequency of management based on the estimated user emotions. For example, if the user is feeling stressed, the management unit can reduce the frequency of management to reduce the user's burden. The management unit can also increase the frequency of management and collect detailed data if the user is relaxed. Furthermore, if the user is in a hurry, the management unit can shorten the frequency of management and collect data quickly. This reduces the user's burden by adjusting the frequency of management based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) and multimodal generation AI. Some or all of the above-mentioned processing in the management unit may be performed using AI, or may be performed without AI. For example, the management unit can input the user's emotion data into AI and have the AI adjust the frequency of management.

During management, the management unit can weight the management data based on the operation period of the AI. For example, the management unit can prioritize weighting of management data for newly introduced AI. The management unit can also periodically weight management data for AI that has been operating for a long period of time. Furthermore, the management unit can adjust the weighting of management data according to the operation period. In this way, weighting of management data based on the operation period of the AI enables efficient management. Some or all of the above-mentioned processing in the management unit may be performed using AI, for example, or may be performed without using AI. For example, the management unit can input AI operation period data into the AI and have the AI perform weighting of the management data.

The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows:

The AI diagnosis and treatment system can further include a prediction unit. The prediction unit can predict the future performance of the AI based on the data collected by the collection unit. For example, the prediction unit can predict how the AI will respond in customer service tasks in the future. The prediction unit can also predict what test results the AI will produce in product inspection tasks in the future. Furthermore, the prediction unit can predict what marketing strategies the AI will formulate in marketing tasks in the future. In this way, the prediction unit can predict the future performance of the AI, allowing measures to be taken in advance.

When collecting AI performance data, the collection unit can evaluate the reliability of the data. For example, the collection unit can check the source and generation method of the data and exclude unreliable data. The collection unit can also check the consistency and integrity of the data and filter out abnormal data. Furthermore, the collection unit can evaluate the recency of the data and exclude old data. This allows the collection unit to collect only highly reliable data, thereby improving the accuracy of AI performance diagnosis.

The diagnostic unit can be equipped with an anomaly detection function when diagnosing AI performance. For example, the diagnostic unit can detect an anomaly if the AI deviates from its normal performance. The diagnostic unit can also issue an alert if the AI exhibits abnormal performance in a specific task. Furthermore, the diagnostic unit can perform a detailed analysis of AI performance based on the results of anomaly detection. This allows the diagnostic unit to detect AI anomalies early and take prompt action.

The correction unit can be equipped with a real-time feedback function when correcting the AI's performance. For example, the correction unit can provide real-time feedback to the AI as it performs its tasks, allowing it to make immediate corrections. The correction unit can also receive real-time feedback from the AI during customer service tasks and correct its responses. Furthermore, the correction unit can receive real-time feedback from the AI during product inspection tasks and correct the inspection results. This allows the correction unit to quickly correct the AI's performance by providing feedback in real time.

The fork unit can be equipped with a version management function when reconstructing an AI algorithm. For example, the fork unit can manage different versions of the AI algorithm and revert to a previous version if necessary. The fork unit can also record the change history of the AI algorithm and track the impact of changes. Furthermore, the fork unit can compare versions of the AI algorithm and select the optimal version. In this way, the fork unit can maintain stable performance by managing versions of the AI algorithm.

The management unit can be equipped with a data visualization function when managing diagnostic results and correction data. For example, the management unit can display diagnostic results and correction data in graphs and charts to make them easier to understand visually. The management unit can also visualize data trends and patterns to understand fluctuations in AI performance. Furthermore, the management unit can provide data filtering and sorting functions to quickly obtain necessary information. In this way, the management unit can make AI performance management more efficient by visualizing data.

The collection unit can estimate the user's emotions and customize the data collection method based on the estimated user emotions. For example, if the user is feeling stressed, the collection unit can simplify the data collection method to reduce the user's burden. Alternatively, if the user is relaxed, the collection unit can adopt a more detailed data collection method to collect more information. Furthermore, if the user is in a hurry, the collection unit can select a method to collect data quickly. In this way, the collection unit can customize the data collection method based on the user's emotions, reducing the user's burden and achieving efficient data collection.

The diagnosis unit can estimate the user's emotions and adjust the timing of the diagnosis based on the estimated user emotions. For example, if the user is feeling stressed, the diagnosis unit can delay the timing of the diagnosis, thereby reducing the burden on the user. Furthermore, if the user is relaxed, the diagnosis unit can advance the timing of the diagnosis and provide diagnostic results quickly. Furthermore, if the user is in a hurry, the diagnosis unit can adjust the timing of the diagnosis and perform a quick diagnosis. In this way, the diagnosis unit can adjust the timing of the diagnosis based on the user's emotions, thereby reducing the burden on the user and achieving efficient diagnosis.

The correction unit can estimate the user's emotions and adjust the correction feedback method based on the estimated user's emotions. For example, if the user is nervous, the correction unit can provide simple, highly visible feedback. If the user is relaxed, the correction unit can also provide detailed feedback. Furthermore, if the user is in a hurry, the correction unit can provide concise feedback that hits the main points. In this way, the correction unit can provide appropriate feedback for the user by adjusting the feedback method based on the user's emotions.

The forking unit can estimate the user's emotions and adjust the timing of forking based on the estimated user emotions. For example, if the user is feeling stressed, the forking unit can delay the timing of forking to reduce the burden on the user. Also, if the user is relaxed, the forking unit can advance the timing of forking to perform the fork quickly. Furthermore, if the user is in a hurry, the forking unit can adjust the timing of forking to perform the fork quickly. In this way, the forking unit can adjust the timing of forking based on the user's emotions, thereby reducing the burden on the user and achieving efficient forking.

The processing flow for Example 2 is briefly explained below.

Step 1: The collection unit collects data generated when the AI is engaged in its work. For example, it collects customer inquiry data generated when the AI is engaged in customer service work, inspection data generated when the AI is engaged in product inspection work, and marketing data generated when the AI is engaged in marketing work.
Step 2: The diagnostic unit diagnoses the performance of the AI based on the data collected by the collection unit. For example, it diagnoses whether the AI is responding in a biased manner in customer service, producing biased test results in product inspection, or formulating a biased marketing strategy in marketing.
Step 3: The correction unit inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit. For example, if the AI is responding unbalancedly in customer service operations, training data is input to correct that response; if the AI is producing unbalanced test results in product inspection operations, training data is input to correct those test results; and if the AI is formulating a biased marketing strategy in marketing operations, training data is input to correct that strategy.
Step 4: The forking unit reconstructs the AI algorithm corrected by the correction unit. For example, if the AI shows significant bias in customer service, product inspection, or marketing tasks, the algorithm is reconstructed.
Step 5: The management department manages the diagnostic results and correction data. For example, the management department stores the diagnostic results and correction data on the cloud and makes them accessible as needed. The management department also encrypts and stores the diagnostic results and correction data to ensure security. Furthermore, the management department sets permissions to access the diagnostic results and correction data and performs access control.

The specific processing unit 290 transmits the results 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 results of the specific processing. The microphone 38B acquires audio indicating the user input regarding the results 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.

The data generation model 58 is what is known as generative AI (artificial intelligence). An example of a data generation model 58 is generative AI such as ChatGPT (registered trademark) (Internet search <URL: https://openai.com/blog/chatgpt>). The data generation model 58 is obtained by performing deep learning on a neural network. A prompt containing an instruction is input to the data generation model 58, and inference data such as voice data indicating speech, text data indicating text, and image data indicating an image (e.g., still image data or video data) is also input. The data generation model 58 performs inference on the input inference data in accordance with the instructions indicated by the prompt and outputs the inference results in one or more data formats, such as voice data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and/or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not include instructions. In this case, the data generation model 58 can output inference results from prompts that do not include instructions. The data processing device 12 or the like includes multiple types of data generation models 58, and the data generation model 58 includes AI other than the generation AI. Examples of AI other than the generation AI include linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), and naive Bayes, and can perform various processes, but are not limited to these examples. The AI may also be an AI agent. When the processing of each of the above-mentioned parts is performed by AI, the processing may be performed in part or in whole by AI, but is not limited to these examples. In addition, processing performed by AI, including generation AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generation AI.

Furthermore, the processing by the data processing system 10 described above is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart device 14 or an external device, etc., and the smart device 14 acquires or collects information necessary for processing from the data processing device 12 or an external device, etc.

Each of the multiple elements, including the collection unit, diagnosis unit, correction unit, fork unit, and management unit, described above, is realized, for example, by at least one of the smart device 14 and the data processing device 12. For example, the collection unit collects data using the camera 42 and microphone 38B of the smart device 14 and transmits the collected data to the data processing device 12 via the control unit 46A. The diagnosis unit, realized, for example, by the specific processing unit 290 of the data processing device 12, diagnoses AI performance based on the collected data. The correction unit, realized, for example, by the specific processing unit 290 of the data processing device 12, inputs training data for correction based on the diagnosis results. The fork unit, realized, for example, by the specific processing unit 290 of the data processing device 12, reconstructs the corrected AI algorithm. The management unit, realized, for example, by the specific processing unit 290 of the data processing device 12, manages the diagnosis results and correction data. The correspondence between each unit and the device or control unit is not limited to the above example and various modifications are possible.

Second Embodiment
FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

As shown in FIG. 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

The data processing device 12 includes a computer 22, a database 24, and a communication I/F 26. The computer 22 includes a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and 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 and/or a LAN.

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, RAM 48, and storage 50. The processor 46, RAM 48, and 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.

The microphone 238 receives instructions and other information from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, 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.

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 user's surroundings (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

The communication I/F 44 is connected to the 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 manner.

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

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 the specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

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. The identification processing unit 290 can estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

In the smart glasses 214, specific processing is performed by the processor 46. A specific processing program 60 is stored in the storage 50. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as the control unit 46A in accordance with the specific processing program 60 executed on the RAM 48. The smart glasses 214 also have a data generation model and emotion identification model similar to the data generation model 58 and emotion identification model 59, and can use these models to perform processing similar to that of the specific processing unit 290.

Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain processing results (such as prediction results) using the data generation model 58. The data processing device 12 may also be a server device, or a terminal device owned by a user (for example, a mobile phone, robot, home appliance, etc.).

The specific processing unit 290 transmits the results 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 results of the specific processing. The microphone 238 acquires audio indicating the user input regarding the results 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.

The data generation model 58 is what is known as generative AI. An example of a data generation model 58 is generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt containing an instruction is input to the data generation model 58, and inference data such as voice data indicating speech, text data indicating text, and image data indicating an image (e.g., still image data or video data) is 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 one or more data formats, such as voice data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and/or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not include instructions. In this case, the data generation model 58 can output inference results from prompts that do not include instructions. The data processing device 12 or the like includes multiple types of data generation models 58, and the data generation model 58 includes AI other than the generation AI. Examples of AI other than the generation AI include linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), and naive Bayes, and can perform various processes, but are not limited to these examples. The AI may also be an AI agent. When the processing of each of the above-mentioned parts is performed by AI, the processing may be performed in part or in whole by AI, but is not limited to these examples. In addition, processing performed by AI, including generation AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generation AI.

The data processing system 210 according to the second embodiment performs processing similar to that of the data processing system 10 according to the first embodiment. Processing by the data processing system 210 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart glasses 214 or an external device, etc., and the smart glasses 214 acquires or collects information necessary for processing from the data processing device 12 or an external device, etc.

Each of the multiple elements, including the collection unit, diagnosis unit, correction unit, fork unit, and management unit, described above, is realized, for example, by at least one of the smart glasses 214 and the data processing device 12. For example, the collection unit collects data using the camera 42 and microphone 238 of the smart glasses 214 and transmits the collected data to the data processing device 12 via the control unit 46A. The diagnosis unit, realized, for example, by the specific processing unit 290 of the data processing device 12, diagnoses AI performance based on the collected data. The correction unit, realized, for example, by the specific processing unit 290 of the data processing device 12, inputs training data for correction based on the diagnosis results. The fork unit, realized, for example, by the specific processing unit 290 of the data processing device 12, reconstructs the corrected AI algorithm. The management unit, realized, for example, by the specific processing unit 290 of the data processing device 12, manages the diagnosis results and correction data. The correspondence between each unit and the device or control unit is not limited to the above example and various modifications are possible.

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

As shown in FIG. 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

The data processing device 12 includes a computer 22, a database 24, and a communication I/F 26. The computer 22 includes a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and 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 and/or a LAN.

The headset 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, RAM 48, and storage 50. The processor 46, RAM 48, and 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.

The microphone 238 receives instructions and other information from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, 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.

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 user's surroundings (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

The communication I/F 44 is connected to the 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 manner.

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

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 the specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

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. The identification processing unit 290 can estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

In the headset type terminal 314, the specific processing is performed by the processor 46. A specific program 60 is stored in the storage 50. The processor 46 reads the specific program 60 from the storage 50 and executes the read specific program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as the control unit 46A in accordance with the specific program 60 executed on the RAM 48. The headset type terminal 314 also has a data generation model and emotion identification model similar to the data generation model 58 and emotion identification model 59, and can use these models to perform processing similar to that of the identification processing unit 290.

Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain processing results (such as prediction results) using the data generation model 58. The data processing device 12 may also be a server device, or a terminal device owned by a user (for example, a mobile phone, robot, home appliance, etc.).

The specific processing unit 290 transmits the results of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the results of the specific processing. The microphone 238 acquires audio indicating the user input regarding the results 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.

The data generation model 58 is what is known as generative AI. An example of a data generation model 58 is generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt containing an instruction is input to the data generation model 58, and inference data such as voice data indicating speech, text data indicating text, and image data indicating an image (e.g., still image data or video data) is 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 one or more data formats, such as voice data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and/or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not include instructions. In this case, the data generation model 58 can output inference results from prompts that do not include instructions. The data processing device 12 or the like includes multiple types of data generation models 58, and the data generation model 58 includes AI other than the generation AI. Examples of AI other than the generation AI include linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), and naive Bayes, and can perform various processes, but are not limited to these examples. The AI may also be an AI agent. When the processing of each of the above-mentioned parts is performed by AI, the processing may be performed in part or in whole by AI, but is not limited to these examples. In addition, processing performed by AI, including generation AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generation AI.

The data processing system 310 according to the third embodiment performs the same processing as the data processing system 10 according to the first embodiment. Processing by the data processing system 310 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the headset terminal 314, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the headset terminal 314. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the headset terminal 314 or an external device, and the headset terminal 314 acquires or collects information required for processing from the data processing device 12 or an external device.

Each of the multiple elements, including the collection unit, diagnosis unit, correction unit, fork unit, and management unit, described above, is realized, for example, by at least one of the headset terminal 314 and the data processing device 12. For example, the collection unit collects data using the camera 42 and microphone 238 of the headset terminal 314 and transmits the collected data to the data processing device 12 via the control unit 46A. The diagnosis unit, realized, for example, by the specific processing unit 290 of the data processing device 12, diagnoses AI performance based on the collected data. The correction unit, realized, for example, by the specific processing unit 290 of the data processing device 12, inputs training data for correction based on the diagnosis results. The fork unit, realized, for example, by the specific processing unit 290 of the data processing device 12, reconstructs the corrected AI algorithm. The management unit, realized, for example, by the specific processing unit 290 of the data processing device 12, manages the diagnosis results and correction data. The correspondence between each unit and the device or control unit is not limited to the above example and various modifications are possible.

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

As shown in FIG. 7, the 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.

The data processing device 12 includes a computer 22, a database 24, and a communication I/F 26. The computer 22 includes a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and 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 and/or a LAN.

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, RAM 48, and storage 50. The processor 46, RAM 48, and 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.

The microphone 238 receives instructions and other information from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, 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.

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 image sensor or CCD image sensor, and captures images of the user's surroundings (e.g., an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

The communication I/F 44 is connected to the 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 manner.

The control object 443 includes a display device, LEDs in the eyes, and motors that drive 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.

Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, in the data processing device 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32.

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 the specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

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. The identification processing unit 290 can estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

In robot 414, specific processing is performed by processor 46. Specific program 60 is stored in storage 50. Processor 46 reads specific program 60 from storage 50 and executes the read specific program 60 on RAM 48. The specific processing is realized by processor 46 operating as control unit 46A in accordance with specific program 60 executed on RAM 48. Robot 414 also has a data generation model and emotion identification model similar to data generation model 58 and emotion identification model 59, and can use these models to perform processing similar to that of identification processing unit 290.

Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain processing results (such as prediction results) using the data generation model 58. The data processing device 12 may also be a server device, or a terminal device owned by a user (for example, a mobile phone, robot, home appliance, etc.).

The specific processing unit 290 transmits the results 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 results of the specific processing. The microphone 238 acquires audio indicating the user input regarding the results 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.

The data generation model 58 is what is known as generative AI. An example of a data generation model 58 is generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt containing an instruction is input to the data generation model 58, and inference data such as voice data indicating speech, text data indicating text, and image data indicating an image (e.g., still image data or video data) is 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 one or more data formats, such as voice data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and/or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not include instructions. In this case, the data generation model 58 can output inference results from prompts that do not include instructions. The data processing device 12 or the like includes multiple types of data generation models 58, and the data generation model 58 includes AI other than the generation AI. Examples of AI other than the generation AI include linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), and naive Bayes, and can perform various processes, but are not limited to these examples. The AI may also be an AI agent. When the processing of each of the above-mentioned parts is performed by AI, the processing may be performed in part or in whole by AI, but is not limited to these examples. In addition, processing performed by AI, including generation AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generation AI.

The data processing system 410 according to the fourth embodiment performs processing similar to that of the data processing system 10 according to the first embodiment. Processing by the data processing system 410 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or an external device, etc., and the robot 414 acquires or collects information required for processing from the data processing device 12 or an external device, etc.

Each of the multiple elements, including the collection unit, diagnosis unit, correction unit, fork unit, and management unit, described above, is realized, for example, by at least one of the robot 414 and the data processing device 12. For example, the collection unit collects data using the camera 42 and microphone 238 of the robot 414 and transmits the collected data to the data processing device 12 by the control unit 46A. The diagnosis unit, realized, for example, by the specific processing unit 290 of the data processing device 12, diagnoses the AI performance based on the collected data. The correction unit, realized, for example, by the specific processing unit 290 of the data processing device 12, inputs training data for correction based on the diagnosis results. The fork unit, realized, for example, by the specific processing unit 290 of the data processing device 12, reconstructs the corrected AI algorithm. The management unit, realized, for example, by the specific processing unit 290 of the data processing device 12, manages the diagnosis results and correction data. The correspondence between each unit and the device or control unit is not limited to the above example and various modifications are possible.

The emotion identification model 59, which serves 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 Figure 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.

Figure 9 shows an emotion map 400 on which multiple emotions are mapped. In 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 emotions and mental states. Emotions that are generally generated from reactions that occur in the brain are arranged on the left side of the concentric circles. Emotions that are generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions that are generally generated from reactions that occur in the brain and are 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, emotion map 400 maps multiple emotions based on the structure by which emotions are generated, with emotions that tend to occur simultaneously being mapped close together.

These emotions are distributed in the 3 o'clock direction on emotion map 400, and usually fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness takes precedence over internal sensations, resulting in a sense of calm.

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 out you go on emotion map 400, the more visible (expressed in behavior) the emotion becomes.

Here, human emotions are based on various balances such as posture and blood sugar levels, and when these balances deviate from the ideal, it indicates discomfort, and when they approach the ideal, it indicates pleasure. Emotions can also be created for robots, cars, motorcycles, etc., based on various balances such as posture and remaining battery life, and when these balances deviate from the ideal, it indicates discomfort, and when they approach the ideal, it indicates pleasure. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on speech 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 is lined with emotions belonging to an area called "reaction," where sensation is dominant. The right half of the emotion map is lined with emotions belonging to an area called "situation," where situational awareness is dominant.

The emotion map defines two emotions that encourage learning. One is the negative emotion around the middle of "repentance" or "reflection" on the situation side. In other words, this is 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 the positive emotion around "desire" on the response side. In other words, this is when the robot experiences positive feelings such as "I want more" or "I want to know more."

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 close to each other have similar values, as in the emotion map 900 shown in Figure 10. Figure 10 shows an example in which multiple emotions, such as "relieved," "calm," and "reassuring," have similar emotion values.

In the above embodiment, an example was given in which a specific process is performed by a single computer 22, but the technology disclosed herein is not limited to this, and distributed processing of the specific process may also be performed by multiple computers, including computer 22.

In the above embodiment, an example was described in which the specific processing program 56 is stored in the storage 32, 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.

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.

It is not necessary to store the entire 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 the entire specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

The following types of processors can be used as hardware resources for executing specific processes. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource for executing specific processes by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which are processors with a circuit configuration designed specifically for executing specific processes. All processors have built-in or connected memory, and all use the memory to execute specific processes.

The hardware resource that executes the specific processing may be composed of one of these various processors, or may be composed of 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). The hardware resource that executes the specific processing may also be a single processor.

As an example of a configuration using a single processor, first, there is a configuration in which one processor is configured using a combination of one or more CPUs and software, and this processor functions as a hardware resource that executes specific processing. Second, there is a configuration in which a processor is used to realize the functions of an entire system, including multiple hardware resources that execute specific processing, on a single IC chip, as typified by SoC (System-on-a-chip). In this way, specific processing is realized using one or more of the various processors listed above as hardware resources.

More specifically, the hardware structure of these various processors can be an electrical circuit that combines circuit elements such as semiconductor devices. Furthermore, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps can be added, or the processing order can be rearranged, all within the scope of the spirit of the invention.

Furthermore, in the above examples, the first to fourth embodiments have been described separately, but some or all of these embodiments may be combined. Furthermore, the smart device 14, smart glasses 214, headset terminal 314, and robot 414 are only examples, and they may be combined, or other devices may be used. Furthermore, in the above examples, the first and second embodiments have been described separately, but these may also be combined.

The above-described written content 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 explanation 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 substitutions may be made to the above-described written content and illustrations, as long as they do not deviate from the spirit 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 written content 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.

All publications, patent applications, and technical standards mentioned in this specification are incorporated herein by reference to the same extent as if each individual publication, patent application, and technical standard was specifically and individually indicated to be incorporated by reference.

(Appendix 1)
a collection unit that collects data;
A diagnosis unit that diagnoses AI performance based on the data collected by the collection unit;
a correction unit that inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit;
A fork unit that reconstructs the AI algorithm corrected by the correction unit;
a management unit that manages the diagnostic results or correction data.
(Appendix 2)
The collecting unit
The system described in Appendix 1, characterized in that it collects data generated when the AI engages in work.
(Appendix 3)
The diagnostic unit
2. The system of claim 1, having a self-diagnostic program to periodically check performance.
(Appendix 4)
The correction unit is
The system according to claim 1, wherein training data generated based on past data is input.
(Appendix 5)
The fork portion is
2. The system of claim 1, wherein a hard fork is performed if the AI shows significant bias.
(Appendix 6)
The management unit
2. The system of claim 1, wherein the diagnostic results or correction data are stored on the cloud and can be accessed as needed.
(Appendix 7)
The collecting unit
2. The system of claim 1, further comprising: estimating a user's emotion; and adjusting timing of data collection based on the estimated user's emotion.
(Appendix 8)
The collecting unit
2. The system of claim 1, further comprising: analyzing past performance data of the AI and selecting an efficient data collection method.
(Appendix 9)
The collecting unit
The system of claim 1, wherein filtering is performed based on the AI's current business status and areas of interest when collecting data.
(Appendix 10)
The collecting unit
The system according to claim 1, further comprising: estimating a user's emotion; and determining a priority of data to be collected based on the estimated user's emotion.
(Appendix 11)
The collecting unit
The system according to claim 1, characterized in that when collecting data, highly relevant data is collected preferentially taking into account the geographical location information of the AI.
(Appendix 12)
The collecting unit
10. The system of claim 1, wherein, during data collection, the system analyzes the social media activity of the AI and collects relevant data.
(Appendix 13)
The diagnostic unit
2. The system of claim 1, further comprising: estimating a user's emotion; and adjusting a manner in which the diagnosis is presented based on the estimated user's emotion.
(Appendix 14)
The diagnostic unit
The system of claim 1, wherein the level of detail of the diagnosis is adjusted based on the importance of the AI during diagnosis.
(Appendix 15)
The diagnostic unit
The system of claim 1, wherein different diagnostic algorithms are applied depending on the AI category during diagnosis.
(Appendix 16)
The diagnostic unit
2. The system of claim 1, further comprising: estimating a user's emotion; and adjusting the length of the diagnosis based on the estimated user's emotion.
(Appendix 17)
The diagnostic unit
The system according to claim 1, wherein during diagnosis, the system determines the priority of the diagnosis based on the time when the AI is operating.
(Appendix 18)
The diagnostic unit
2. The system of claim 1, wherein, during diagnosis, the order of diagnosis is adjusted based on the relevance of the AI.
(Appendix 19)
The correction unit is
2. The system of claim 1, further comprising: estimating a user's emotion; and adjusting a correction method based on the estimated user's emotion.
(Appendix 20)
The correction unit is
The system according to claim 1, wherein, during correction, the system analyzes past learning data of the AI to select an efficient correction method.
(Appendix 21)
The correction unit is
The system of claim 1, wherein the correction measures are customized based on the current operational status of the AI during correction.
(Appendix 22)
The correction unit is
The system according to claim 1, further comprising: estimating a user's emotion; and determining a priority of correction based on the estimated user's emotion.
(Appendix 23)
The correction unit is
The system according to claim 1, wherein the system selects the optimal correction method in consideration of the geographical location information of the AI during correction.
(Appendix 24)
The correction unit is
10. The system of claim 1, wherein, during remediation, the system analyzes the AI's social media activity to suggest remedial measures.
(Appendix 25)
The fork portion is
2. The system of claim 1, further comprising: estimating a user's sentiment; and adjusting a forking strategy based on the estimated user's sentiment.
(Appendix 26)
The fork portion is
The system of claim 1, wherein at the time of forking, the system analyzes past performance data of the AI to select the optimal forking method.
(Appendix 27)
The fork portion is
The system of claim 1, wherein at the time of forking, the forking method is customized based on the current business situation of the AI.
(Appendix 28)
The fork portion is
2. The system of claim 1, further comprising: estimating a user's sentiment; and determining a fork priority based on the estimated user's sentiment.
(Appendix 29)
The fork portion is
The system described in Appendix 1, characterized in that when forking, the optimal forking method is selected taking into account the AI's geographical location information.
(Appendix 30)
The fork portion is
2. The system of claim 1, wherein, at the time of forking, the system analyzes the social media activity of the AI to suggest a means of forking.
(Appendix 31)
The management unit
The system according to claim 1, further comprising: estimating a user's emotion; and selecting management data based on the estimated user's emotion.
(Appendix 32)
The management unit
The system according to claim 1, wherein the management algorithm is optimized by referring to past management data during management.
(Appendix 33)
The management unit
2. The system of claim 1, further comprising: estimating a user's emotion; and adjusting the frequency of management based on the estimated user's emotion.
(Appendix 34)
The management unit
The system according to claim 1, wherein, during management, the management data is weighted based on the operation period of the AI.

10, 210, 310, 410 Data processing system 12 Data processing device 14 Smart device 214 Smart glasses 314 Headset type terminal 414 Robot

Claims (9)

a collection unit that collects data;
A diagnosis unit that diagnoses AI performance based on the data collected by the collection unit;
a correction unit that inputs training data to be corrected based on the results of the diagnosis by the diagnosis unit;
A fork unit that reconstructs the AI algorithm corrected by the correction unit;
a management unit that manages the diagnosis results or correction data of the diagnosis unit ,
The collecting unit
Inferring user emotions and adjusting the timing of data collection based on the estimated user emotions
A system characterized by: The collecting unit
The system according to claim 1, wherein the system collects data generated when the AI performs its work. The diagnostic unit
10. The system of claim 1, further comprising a self-diagnostic program for periodically checking performance. The correction unit is
The system according to claim 1, wherein training data generated based on past data is input. The fork portion is
The system of claim 1, wherein a hard fork is performed if the AI shows a significant bias. The management unit
The system of claim 1, wherein the diagnostic results or correction data are stored on the cloud and can be accessed as needed. The collecting unit
The system according to claim 1, further comprising: analyzing past performance data of the AI and selecting an efficient data collection method. The collecting unit
The system of claim 1, wherein filtering is performed based on the AI's current business situation and areas of interest when collecting data. The collecting unit
The system according to claim 1 , further comprising: estimating a user's emotion; and determining a priority of data to be collected based on the estimated user's emotion.