Information processing systems, information processing methods, information processing programs, and AI agents

Abstract

This invention provides an information processing system, information processing method, information processing program, and AI agent that can improve the execution accuracy and processing speed of a given task by evaluating and weighting multiple tasks that constitute a given task, and by reviewing the decomposition method and processing order of the given task. [Solution] love Information processing system A The I agent comprises a task decomposition means capable of decomposing a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means capable of executing the plurality of subtasks in a predetermined processing order, and a task evaluation means that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks. Furthermore, the task decomposition means can decompose the predetermined task into a plurality of subtasks according to the processing burden, the task execution means can execute the plurality of subtasks in order of increasing processing burden, the task evaluation means assigns weights to each of the plurality of subtasks based on the execution results and evaluation indicators of the plurality of subtasks, and if there is a subtask whose weighting does not meet a predetermined standard, the task decomposition means decomposes or deletes that subtask.

Description

Technical Field

[0001] The present invention relates to an information processing system, an information processing method, an information processing program using an AI agent, and an AI agent.

Background Art

[0002] Conventionally, there has been proposed a management system using a generative AI that visualizes and classifies tasks across the company by AI and appropriately assigns them to each employee, where the AI monitors the progress of tasks in real time and reassigns tasks as necessary (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, although the conventional system can assign tasks by AI, it is for the purpose of company management, so there is a problem that the scope of application is limited.

[0005] In view of such circumstances, the present invention provides an information processing system, an information processing method, an information processing program, and an AI agent that can improve the execution accuracy and processing speed of a predetermined task by evaluating and weighting a plurality of tasks constituting the predetermined task and reviewing the decomposition method and processing order of the predetermined task.

Means for Solving the Problems

[0006] The information processing system according to the present invention is an information processing system configured to have an AI agent, wherein the AI ​​agent comprises: task decomposition means capable of decomposing a predetermined task into a plurality of subtasks using a predetermined decomposition method; task execution means capable of executing the plurality of subtasks in a predetermined processing order; and task evaluation means that evaluate the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the processing burden, the task execution means can execute the plurality of subtasks in order of increasing processing burden, the task evaluation means assigns weights to each of the plurality of subtasks based on the execution results and evaluation indicators of the plurality of subtasks, and the task decomposition means decomposes or deletes any subtasks whose weighting does not meet a predetermined standard. This is an information processing system characterized by the following:

[0007] Book The information processing method according to the invention is an information processing method performed using an AI agent which is a computer, wherein the AI ​​agent includes a task decomposition step in which a predetermined task can be decomposed into a plurality of subtasks using a predetermined decomposition method, a task execution step in which the plurality of subtasks can be executed in a predetermined processing order, and a task that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks. evaluation The system is configured to perform at least the steps and , and to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks. The task decomposition step allows the predetermined task to be decomposed into a plurality of subtasks according to the processing burden; the task execution step allows the plurality of subtasks to be executed in order of increasing processing burden; the task evaluation step weights each of the plurality of subtasks based on the execution results and evaluation indicators of the plurality of subtasks; and if there is a subtask whose weighting does not meet a predetermined standard, the task decomposition step decomposes or deletes that subtask. This is an information processing method characterized by the following features.

[0008] Book The information processing program according to the invention is a program for an information processing system configured to have an AI agent, wherein the AI ​​agent, which is a computer, functions as a task decomposition means capable of decomposing a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means capable of executing the plurality of subtasks in a predetermined processing order, and a task evaluation means that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the processing burden, the task execution means can execute the plurality of subtasks in order of increasing processing burden, the task evaluation means assigns weights to each of the plurality of subtasks based on the execution results and evaluation indicators of the plurality of subtasks, and the task decomposition means decomposes or deletes any subtasks whose weighting does not meet a predetermined standard. This is an information processing program characterized by the following:

[0009] BookThe AI ​​agent according to the invention is an AI agent which is a computer that acts as a substitute for a human and has the ability to learn and make decisions on its own, and the AI ​​agent is configured to function as a task decomposition means that can decompose a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means that can execute the plurality of subtasks in a predetermined processing order, and a task evaluation means that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the processing burden, the task execution means can execute the plurality of subtasks in order of increasing processing burden, the task evaluation means assigns weights to each of the plurality of subtasks based on the execution results and evaluation indicators of the plurality of subtasks, and if there is a subtask whose weighting does not meet a predetermined standard, the task decomposition means decomposes or deletes that subtask. The AI ​​agent is characterized in that the predetermined decomposition method and the predetermined processing order are repeatedly provided as training data, and the agent is configured to learn the relationship between the predetermined decomposition method and the predetermined processing order and the predetermined task. [Effects of the Invention]

[0010] According to the information processing system, information processing method, information processing program, and AI agent of the present invention, it is possible to achieve the excellent effect of improving the execution accuracy and processing speed of a predetermined task by evaluating and weighting multiple tasks that constitute a predetermined task, and by reviewing the decomposition method and processing order of the predetermined task. [Brief explanation of the drawing]

[0011] [Figure 1] This is a schematic diagram showing an overview of the information processing system 10 according to this embodiment. [Figure 2] This is a system configuration diagram showing an example of the configuration of the information processing system 10 according to this embodiment. [Figure 3] (a) A diagram showing an example of how to decompose a task. (b) A diagram showing an example of the processing order of a task. (c) A diagram showing an example of task evaluation and revision of the task decomposition method and processing order. [Modes for carrying out the invention]

[0012] Hereinafter, the information processing system 10 according to an embodiment of the present invention will be described with reference to the drawings.

[0013] <Overview of the System> First, the overview of the information processing system 10 according to the present embodiment will be described with reference to FIG. 1. FIG. 1 is a schematic diagram showing the overview of the information processing system 10 according to the present embodiment.

[0014] The information processing system 10 is an information processing system configured to include an AI agent 12. The AI agent 12 includes a task decomposition means 12a that can decompose a predetermined task into a plurality of subtasks by a predetermined decomposition method, a task execution means 12b that can execute a plurality of subtasks in a predetermined processing order, and a task evaluation means 12c that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks. The information processing system is characterized in that it reviews a predetermined decomposition method and a predetermined processing order based on the evaluation of the plurality of subtasks.

[0015] According to the information processing system 10, by evaluating and weighting a plurality of tasks constituting a predetermined task and reviewing the decomposition method and processing order of the predetermined task, the execution accuracy and processing speed of the predetermined task can be improved.

[0016] Here, the "AI agent" is a system that operates as an agent of a human, has the ability to learn and make judgments on its own, and is a system that integrates various AI technologies for solving complex problems.

[0017] The AI agent executes processing based on a trigger given from the outside (for example, a question (prompt) by a user, a signal input from a sensor), or executes processing based on a trigger given from the inside (for example, activation by an internal scheduler, occurrence of an abnormality).

[0018] For example, when a question (prompt) is given as an external activation trigger (event), tasks such as searching for an answer to the question or generating an answer are executed, and the answer to the question is output. Also, for example, when a task (prompt) is given as an external activation trigger (event), tasks such as searching for a solution to the task or generating a solution are executed, and the solution to the task is output.

[0019] Also, a "prompt" refers to an instruction or question given by a user to the system, and includes, for example, an AI prompt given to a generative AI, a command prompt for giving an instruction by a command sentence (command), etc. The data format of the "prompt" is not particularly limited, and may be any of, for example, text data, voice data, still image data, video data, files, etc.

[0020] Various processes executed by the AI agent can also be executed by "agent-based AI".

[0021] Here, "agent-based AI" refers to a system in which multiple AI agents operate in cooperation. Agent-based AI is used to solve complex problems that cannot be solved by a single AI agent. Each AI agent plays a specialized role and outputs an answer to a question or a solution to a task while exchanging information with each other.

[0022] A "task" is an operation necessary to achieve a certain purpose or objective. Examples of tasks include tasks generated based on external triggers (e.g., user questions (prompts), sensor inputs), tasks generated (defined) by the AI ​​agent itself in response to ambiguous or abstract questions or issues (prompts) entered by the user, and tasks generated based on internal triggers (e.g., activation by an internal scheduler, occurrence of an anomaly). There is no specific type of task, but examples include tasks related to management figures and tasks related to purchase history information.

[0023] "Management numerical information" refers to information used to evaluate a company's performance through management indicators and financial indicators. Examples of "management numerical information" include: (1) sales and profit-related data (e.g., sales data, profit data), (2) sales and product management-related data (e.g., product master data, shelf layout data, sales performance and promotional effectiveness), (3) inventory and procurement-related data (e.g., inventory status, order and procurement data), (4) customer and marketing-related data (e.g., customer ID data (ID-POS), marketing campaign results), (5) store operations and personnel management-related data (e.g., store performance, staff shifts), (6) expense and financial-related data (e.g., store and headquarters expenses, financial indicators), (7) external environment data (e.g., trade area / population data, competitor store location and pricing information, weather, temperature, and disaster information), (8) local event / school event information, and (9) instruction and communication-related data (e.g., work instructions and policies from superiors, reports and proposals from the field, history of collaboration with headquarters).

[0024] Furthermore, among the "data related to management figures," (2) sales and product management related data (e.g., product master specifications, shelf layout data, sales performance and promotional effects) may be referred to as "data related to sales promotion (sales promotion)," and among the "data related to management figures," (4) customer and marketing related data (e.g., customer ID data (I "Purchase history information (data related to purchase history)" refers to information about the history of products purchased by a specific individual (customer). Examples of "purchase history information" include: (1) customer information (customer ID, gender, age, membership rank, etc.), (2) product information (product ID, product name, category, brand, unit price, etc.), (3) purchase information (purchase date, purchase time, purchase quantity, total amount, payment method, surveys, staff interaction records, etc.), (4) store information (store ID, store name, sales channel (store / EC), etc.), (5) campaign information (status of campaigns and events, coupon usage, point usage, discount rate, etc.), and (6) purchase frequency (number of purchases, average purchase interval, most recent purchase date, etc.).

[0025] "Predetermined decomposition method" refers to a method by which the task decomposition means 12a decomposes (divides) a predetermined task into multiple subtasks. Examples of "predetermined decomposition methods" include a method of decomposing a predetermined task into multiple subtasks in chronological order, a method of decomposing a predetermined task into multiple subtasks according to the content and type of processing, a method of decomposing a predetermined task into multiple subtasks according to the burden of processing, or a combination of these methods.

[0026] An example of a "method for breaking down a given task into multiple subtasks in chronological order" would be a method for breaking down a given task into three subtasks that are executed in chronological order: subtask 1 (process 1), subtask 2 (process 2), and subtask 3 (process 3), when the given task consists of three processes, process 1, process 2, and process 3, and these processes are executed in the order of process 1 → process 2 → process 3.

[0027] An example of a "method for decomposing a given task into multiple subtasks according to the content and type of processing" is when a given task consists of three processes: speech recognition, image creation, and data transmission. In this method, the given task is decomposed into three subtasks: subtask 1 (speech recognition), subtask 2 (image creation), and subtask 3 (data transmission), according to the content and type of processing. In this method, the processing order of subtasks 1 to 3 is not particularly limited; for example, the three subtasks may be executed simultaneously.

[0028] An example of a "method for breaking down a given task into multiple subtasks according to the processing burden" would be, for instance, if a given task consists of five processes—process 1 with a high processing burden and processes 2-5 with low processing burdens—a method of breaking down the given task into three subtasks, subtask 1 (process 1), subtask 2 (processes 2-3), and subtask 3 (processes 4-5), according to the processing burden (for example, by averaging the processing burden of the subtasks).

[0029] "Predetermined processing order" refers to the order in which the task execution means 12b executes (processes) multiple subtasks. Examples of "predetermined processing order" include executing multiple subtasks sequentially in chronological order, executing multiple tasks in a predetermined order, executing multiple subtasks in order of increasing processing burden, executing multiple subtasks in order of decreasing processing burden, executing multiple subtasks in order of earliest execution trigger, and executing multiple subtasks randomly.

[0030] The task decomposition means 12a may be configured to revise a predetermined decomposition method based on the evaluation of multiple subtasks, and to decompose a predetermined task into multiple subtasks using a method different from the predetermined decomposition method.

[0031] With this configuration, by breaking down a given task into multiple subtasks using multiple decomposition methods, it is possible to find a decomposition method for subtasks that is suitable for the given task, thereby improving the execution accuracy and processing speed of the given task.

[0032] Furthermore, the task execution means 12b may be configured to revise a predetermined processing order based on the evaluation of multiple subtasks and execute the multiple subtasks in a processing order different from the predetermined processing order.

[0033] With this configuration, by executing multiple subtasks in multiple processing sequences, it is possible to find a processing sequence for subtasks that is suitable for a given task, thereby improving the execution accuracy and processing speed of the given task.

[0034] Furthermore, the predetermined decomposition method may include a method for decomposing a predetermined task into N subtasks (where N is a positive integer greater than or equal to 2), and a method different from the predetermined decomposition method may include a method for decomposing a predetermined task into M subtasks (where M is a positive integer greater than or equal to 2) that are different from N.

[0035] With this configuration, by breaking down a given task into multiple subtasks using multiple decomposition methods, it is possible to find a decomposition method for subtasks that is suitable for the given task, thereby improving the execution accuracy and processing speed of the given task.

[0036] Furthermore, the predetermined decomposition method and predetermined processing order may be continuously reviewed until the evaluation of all subtasks reaches a predetermined level.

[0037] With this configuration, it is possible to find the optimal method and processing order for breaking down subtasks into a given task, thereby improving the execution accuracy and processing speed of that task.

[0038] Furthermore, the task decomposition means 12a may be configured to reduce the number of subtasks based on the evaluation of the subtasks, and the task execution means 12b may be configured to execute some of the subtasks in parallel based on the evaluation of the subtasks.

[0039] With this configuration, the processing speed of a given task can be increased by increasing the processing speed of subtasks.

[0040] Furthermore, a predetermined decomposition method and a predetermined processing order are repeatedly provided as training data, and the system is configured to learn the relationship between the predetermined decomposition method and the predetermined processing order and a predetermined task.

[0041] With this configuration, the processing speed of a given task can be increased by improving the processing speed of the task decomposition means and the task execution means.

[0042] <Example of system configuration> Next, an example of the configuration of the information processing system 10 according to this embodiment will be described using Figure 2. Figure 2 is a system configuration diagram showing an example of the configuration of the information processing system 10 according to this embodiment.

[0043] The information processing system 10 can be configured, for example, to include a system terminal 12 that controls the entire system, and external terminals 16 that are connected to the system terminal 12 via a network NW so that they can communicate with each other.

[0044] System terminal 12 is a terminal that constitutes the AI ​​agent and controls the entire information processing system 10, and is composed of conventionally known servers, personal computers, etc. In this example, system terminal 12 is composed of one server, but it may be composed of multiple servers, personal computers, etc. The hardware configuration of system terminal 12 and the programs that system terminal 12 executes will be described later.

[0045] External terminals 16 are terminals used by users of the information processing system 10 (individuals or specific groups of people (teams, departments, companies, etc.)), and consist of personal computers, tablets, smartphones, etc. The types of external terminals 16 are not particularly limited, but examples include smartphones, personal computers, and tablets used by individuals, or smartphones, personal computers, and tablets used by employees of a company or specific groups.

[0046] The network NW is a line that allows system terminal 12 and external terminal 16 to communicate with each other, and is typically composed of a WAN (Wide Area Network), also known as the Internet. The network NW may be wired or wireless, a LAN (Local Area Network), a dedicated line such as a VPN (Virtual Private Network), or a combination of these lines.

[0047] <System Terminal / Hardware Configuration Example> Next, we will describe an example of the hardware configuration of system terminal 12.

[0048] As shown in Figure 2, the system terminal 12 is configured to include, for example, a CPU 21, and a ROM 22, RAM 23, external storage drive 25, storage device 26, input device 27, display device 28, communication unit 29, etc., all connected to the CPU 21 via a bus.

[0049] The CPU 21 is a control means that controls the entire system terminal 12, and performs processes such as executing application programs and operating systems (OS) stored in ROM 22 and storage devices 26, and storing data and files necessary for program execution in RAM 23 and storage devices 26.

[0050] ROM22 is a storage means for storing basic I / O programs and various data, and is composed of, for example, PROM, flash memory, etc. RAM23 is a storage means for temporarily storing data, and is composed of, for example, SDRAM, DRAM, etc. External storage drive25 is a control means that can read and write data to recording media 24 such as magnetic tape, DVD, etc., and is composed of, for example, magnetic tape storage, DVD drive, etc.

[0051] The storage device 26 is a storage means for storing application programs, the OS, control programs, related programs, various information, etc., and is composed of, for example, a hard disk drive (HDD), a solid-state drive (SDD), etc. The input device 27 is for inputting commands (instructions), etc., to the system terminal 12, and is composed of, for example, a keyboard, a pointing device (mouse, etc.), a touch panel, etc.

[0052] The display device 28 displays commands input by the input device 27, the response output of the system terminal 12 to those commands, and various other displays, and is composed of, for example, a liquid crystal display, plasma display, organic EL, etc. The communication unit 29 is a control means that controls communication with external terminals 16, etc. via the network NW, and is composed of, for example, a communication card, etc.

[0053] <System Terminal / Function> Next, we will explain the functions of the system terminal 12.

[0054] The storage device 26 of the system terminal 12 stores a program (information processing program) that causes the system terminal 12 to function as a task decomposition means 12a, a task execution means 12b, and a task evaluation means 12c.

[0055] <System terminal / Function / Task breakdown method> Next, the task decomposition means 12a will be described.

[0056] The task decomposition means 12a is a means that can decompose a predetermined task into multiple subtasks using a predetermined decomposition method, and in this example, it is composed of a program stored in the storage device 26 of the system terminal 12, and the storage device 26, etc.

[0057] The task decomposition means 12a decomposes a task (a predetermined task) generated based on an activation trigger provided from within or outside the AI ​​agent 12 into multiple tasks using a predetermined decomposition method.

[0058] As mentioned above, a "task" is any work necessary to achieve a certain purpose or objective. Examples of tasks include tasks generated based on external triggers (e.g., user questions (prompts), sensor inputs), tasks generated (defined) by the AI ​​agent itself in response to ambiguous or abstract questions or issues (prompts) entered by the user, and tasks generated based on internal triggers (e.g., activation by an internal scheduler, occurrence of an anomaly).

[0059] "Predetermined decomposition method" refers to a method by which the task decomposition means 12a decomposes (divides) a predetermined task into multiple subtasks. Examples of "predetermined decomposition methods" include a method of decomposing a predetermined task into multiple subtasks in chronological order, a method of decomposing a predetermined task into multiple subtasks according to the content and type of processing, a method of decomposing a predetermined task into multiple subtasks according to the burden of processing, or a combination of these methods.

[0060] An example of a "method for breaking down a given task into multiple subtasks in chronological order" would be a method for breaking down a given task into three subtasks that are executed in chronological order: subtask 1 (process 1), subtask 2 (process 2), and subtask 3 (process 3), when the given task consists of three processes, process 1, process 2, and process 3, and these processes are executed in the order of process 1 → process 2 → process 3.

[0061] An example of a "method for decomposing a given task into multiple subtasks according to the content and type of processing" is when a given task consists of three processes: speech recognition, image creation, and data transmission. In this method, the given task is decomposed into three subtasks: subtask 1 (speech recognition), subtask 2 (image creation), and subtask 3 (data transmission), according to the content and type of processing. In this method, the processing order of subtasks 1 to 3 is not particularly limited; for example, the three subtasks may be executed simultaneously.

[0062] An example of a "method for breaking down a given task into multiple subtasks according to the processing burden" would be, for instance, if a given task consists of five processes—process 1 with a high processing burden and processes 2-5 with low processing burdens—a method of breaking down the given task into three subtasks, subtask 1 (process 1), subtask 2 (processes 2-3), and subtask 3 (processes 4-5), according to the processing burden (for example, by averaging the processing burden of the subtasks).

[0063] For example, if a task is given to the AI ​​agent 12 from outside as a predetermined task, the task decomposition means 12a decomposes the processing necessary to accomplish the task into multiple processes using a predetermined decomposition method, based on the knowledge acquired through machine learning and the knowledge acquired from the external terminal 16, and stores each of the decomposed processes in the storage device 26, associating it with the user's ID as a subtask.

[0064] For example, if an external terminal 16 provides a task (prompt) such as "Predict sales of product a," the task decomposition means 12a decomposes the processes (tasks) necessary to accomplish the task (predict sales of product a) into multiple processes (subtasks) according to the content and type of the processes (tasks), based on the knowledge acquired through machine learning and the knowledge obtained from the external terminal 16.

[0065] The processes necessary to realize a sales forecast for product a include, for example, (process 1) collecting sales performance data for product a (e.g., POS data), (process 2) collecting external factor data (weather, season, etc.), (process 3) collecting market data (market trends, sales trends of similar products, etc.), (process 4) collecting competitor data (prices of competing products, etc.), (process 5) selecting a forecasting model (moving average method, exponential smoothing method, etc.), (process 6) creating a sales forecast, and (process 7) visualizing the sales forecast (creating a forecast graph, etc.).

[0066] The task decomposition means 12a decomposes each of the multiple decomposed (process 1) to (process 7) into subtasks 1 to 7, as shown in Figure 3(a), and then stores these subtasks 1 to 7 in the storage device 26, associating them with the user's ID.

[0067] <System terminal / Function / Task execution method> Next, the task execution means 12b will be described.

[0068] The task execution means 12b is a means capable of executing multiple subtasks in a predetermined processing order, and in this example, it is composed of a program stored in the storage device 26 of the system terminal 12, and the storage device 26, etc.

[0069] The task execution means 12b refers to a plurality of subtasks stored in the storage device 26 by the task decomposition means 12a and executes a predetermined task by executing the plurality of subtasks in a predetermined processing order.

[0070] As stated above, the "predetermined processing order" refers to the order in which the task execution means 12b executes (processes) multiple subtasks. Examples of the "predetermined processing order" include executing multiple subtasks sequentially in chronological order, executing multiple tasks in a predetermined order, executing multiple subtasks in order of increasing processing burden, executing multiple subtasks in order of decreasing processing burden, executing multiple subtasks in order of earliest execution trigger, and executing multiple subtasks randomly.

[0071] For example, consider a case where a given task is "Predict sales of product a," and multiple subtasks are stored in the memory device 26, including: (Subtask 1) Collection of sales performance data for product a, (Subtask 2) Collection of external factor data (weather, season, etc.), (Subtask 3) Collection of market data (market trends, sales trends of similar products, etc.), (Subtask 4) Collection of competitor data (prices of competing products, etc.), (Subtask 5) Selection of a forecasting model (moving average method, exponential smoothing method, etc.), (Subtask 6) Creation of a sales forecast, and (Subtask 7) Visualization of the sales forecast (creation of a forecast graph, etc.).

[0072] First, the task execution means 12b refers to the multiple subtasks stored in the memory device 26 and determines whether each task is a subtask that can be executed in parallel or a subtask that should be executed in a predetermined order, and then determines the optimal execution method that can execute the multiple subtasks in the most efficient way.

[0073] In the previous case, subtasks 1-4 are not dependent on each other and are therefore judged to be subtasks that can be executed in parallel. On the other hand, subtasks 5-7 are dependent on each other and need to be processed in chronological order. Therefore, it is judged that they should be executed in the order of their task numbers (subtask 5 → subtask 6 → subtask 7) after all subtasks 1-4 have been completed. As the optimal execution method, it is chosen to execute subtasks 1-4 in parallel, and then execute the remaining subtasks 5-7 in this order.

[0074] Next, the task execution means 12b executes a predetermined task by executing multiple subtasks according to the determined optimal execution method (predetermined processing order).

[0075] In the previous case, as shown in Figure 3(b), the task execution means 12b executes subtasks 1 to 4 in parallel, and after all of subtasks 1 to 4 are completed, it executes the remaining subtasks 5 to 7 in that order, thereby performing the task of "forecasting sales for product a".

[0076] <System terminal / function / task evaluation method> Next, the task evaluation means 12c will be described.

[0077] The task evaluation means 12c is a means for evaluating a plurality of subtasks based on the execution results of the plurality of subtasks, and in this example, it is composed of a program stored in the storage device 26 of the system terminal 12, and the storage device 26, etc.

[0078] First, the task evaluation means 12c evaluates (weights) multiple subtasks based on the execution results of the processing performed by the task execution means 12b.

[0079] Here, the evaluation (weighting) of subtasks is performed using the following indicators: the execution time of the subtask, the difference between the scheduled execution time and the actual execution time of the subtask (e.g., whether there is a delay, the number of delays), the utilization rate of resources required for the execution of the subtask (e.g., server, CPU, memory, etc.), the accuracy and precision of the subtask execution, the amount of information obtained from the execution of the subtask, and the variability of the subtask processing results.

[0080] For example, consider a case where task execution means 12b executes subtasks 1 to 7 to perform processing for the task (a predetermined task) "to forecast the sales of product a," and generates the sales forecast result.

[0081] In this case, for example, if the task evaluation means 12c analyzes the sales forecast results based on knowledge acquired through machine learning and knowledge obtained from an external terminal 16, and determines that the market trend is not sufficiently reflected, it will determine that the amount of information in (subtask 3) "collection of market data" is small and inaccurate, and set the evaluation score for (subtask 3) to, for example, 3 (low evaluation) on a 10-point scale, as shown in Figure 3(c).

[0082] On the other hand, if the task evaluation means 12c analyzes the results of the sales forecast based on the knowledge acquired through machine learning and the knowledge obtained from the external terminal 16, and determines that the execution speed of collecting sales performance data is fast, it will determine that the execution speed of (subtask 1) "collection of sales performance data for product a" is fast, and set the evaluation score for (subtask 1) to, for example, 9 out of 10 (high evaluation), as shown in Figure 3(c).

[0083] The task evaluation means 12c also evaluates other subtasks besides subtask 1 and subtask 3 based on knowledge acquired through machine learning and knowledge obtained from the external terminal 16, and sets evaluation scores. Here, as shown in Figure 3(c), we assume that the evaluation scores for the other subtasks are set to, for example, 6 on a 10-point scale (the standard evaluation).

[0084] The task decomposition means 12a described above revises the predetermined decomposition method based on the evaluation (weighting) of subtasks performed by the task evaluation means 12c.

[0085] For example, in the case above, subtasks other than subtask 3 will not be reviewed because their evaluation scores are above the predetermined threshold (e.g., 6), while subtask 3 will be reviewed because its evaluation score is below the predetermined threshold (e.g., 6).

[0086] Here, reviewing subtasks involves taking measures to improve the metrics used in evaluating the subtasks (e.g., subtask execution time). Examples of subtask review include changing the processing content of a subtask, subdividing the processing of a subtask, deleting a subtask, adding a new subtask, changing the order in which subtasks are executed, and changing the priority of subtask execution.

[0087] In this example, initially, as shown in Figure 3(a), the task decomposition means 12a decomposed the task (predetermined task) "to forecast the sales of product a" into seven subtasks, subtasks 1 to 7, according to a predetermined decomposition method.

[0088] However, since the evaluation score of subtask 3 was below a predetermined threshold (for example, 6), and there was a need to improve the accuracy of the information, the task decomposition means 12a, as shown in Figure 3(c), decomposes the task (predetermined task) "Predict sales of product a" into six subtasks, subtasks 1-2 and 4-7, according to a decomposition method different from the predetermined decomposition method, and stores these subtasks 1-2 and 4-7 in the storage device 26, associating them with the user's ID.

[0089] Furthermore, the task execution means 12b described above revises the predetermined processing order based on the evaluation (weighting) of subtasks performed by the task evaluation means 12c.

[0090] In this example, as shown in Figure 3(b), the task execution means 12b initially executed subtasks 1 to 4 in parallel according to a predetermined processing order, and then executed the remaining subtasks 5 to 7 in that order.

[0091] However, in addition to the task decomposition means 12a revising the task to "predict sales of product a" (a predetermined task) into six subtasks, subtasks 1-2 and 4-7, the evaluation score of subtask 1 is above a predetermined threshold (for example, 6), and the need to execute subtasks 1-4 in parallel has decreased. Therefore, as shown in Figure 3(c), the task execution means 12b executes the six subtasks in a different processing order from the predetermined processing order, in the order of subtask 2 → subtask 1 → subtasks 4-7.

[0092] As this example demonstrates, by evaluating and weighting multiple tasks that constitute a given task, and by reviewing the decomposition method and processing order of the given task, the execution accuracy and processing speed of the given task can be improved.

[0093] Thereafter, the task decomposition means 12a continues to review the predetermined decomposition method and the task execution means 12b continues to review the predetermined processing order until the evaluation scores of all subtasks reach a predetermined threshold (for example, 6).

[0094] Then, once the task decomposition means 12a has completed the review of the predetermined decomposition method and the task execution means 12b has completed the review of the predetermined processing order, the task execution means 12b executes the multiple subtasks after the review has been finalized and outputs a solution for the predetermined task (problem).

[0095] For example, in the state shown in Figure 3(c), once the task decomposition means 12a has completed reviewing the predetermined decomposition method and the task execution means 12b has completed reviewing the predetermined processing order, the six subtasks 1-2 and 4-7 are executed in the order of subtask 2 → subtask 1 → subtasks 4-7, and the sales forecast result (solution) is transmitted to the external terminal 16 used by the user as the execution result (answer) for the predetermined task (problem).

[0096] Furthermore, the revision of the prescribed decomposition method and prescribed processing order is not limited to the deletion of subtasks, but may also include, for example, splitting subtasks, changing the processing content of subtasks, adding new subtasks, changing the order in which subtasks are executed, changing the priority of execution of subtasks, or swapping all subtasks.

[0097] Therefore, for example, if the processing time for (Task 3) shown in Figure 3(a) is long, the predetermined decomposition method may be reviewed, and the processing of (Task 3) may be divided into two parts: (Task 3-1) collecting market trends and (Task 3-2) collecting sales trends of similar products, thereby increasing the number of subtasks to be divided from 7 to 8.

[0098] Furthermore, for example, if the processing time for collecting conflict data (Task 4) shown in Figure 3(a) is long, the predetermined decomposition method may be reviewed to delete the task and reduce the number of subtasks to be divided from 7 to 6.

[0099] Furthermore, if, for example, an evaluation based on knowledge acquired through machine learning or knowledge obtained from an external terminal 16 determines that the accuracy of the sales forecast for product a is low, the predetermined decomposition method may be reviewed and the number of subtasks to be divided into may be increased to four or more in order to improve the accuracy of the sales forecast by collecting additional data other than the data collected in tasks 1 to 4.

[0100] Furthermore, by reviewing the predetermined processing order, tasks that were previously executed in parallel may be prioritized and executed in sequence, or conversely, tasks that were previously prioritized and executed in sequence may be executed in parallel.

[0101] Furthermore, weights and biases may be assigned to subtasks, and the weights and biases of subtasks may be updated according to the error between the processing performed by the task execution means 12b and the processing that was originally intended to be performed, or the weights and biases of subtasks may be updated according to the evaluation score assigned to the subtask by the task evaluation means 12c.

[0102] For example, the system may be configured to not execute a subtask if the weight assigned to it is below a predetermined threshold, and to execute a subtask if the weight assigned to it is equal to or greater than a predetermined threshold. Alternatively, subtasks may be executed in order from those with the largest assigned weights. Furthermore, if there is a large error between the processing performed by the task execution means 12b and the processing that was originally intended to be performed, or if the evaluation scores assigned to subtasks by the task evaluation means 12c are generally low, the accuracy of the processing may be improved by adding a bias to the processing.

[0103] Furthermore, if there are multiple subtasks that perform the same or similar processing, merging these subtasks and executing them together can reduce the overall execution time and optimize the process.

[0104] Furthermore, when executing a series of tasks consisting of multiple different types of processing, the multiple types of processing may be divided into multiple subtasks (or independently of processing units) depending on the content of the series of tasks, and the series of tasks may be executed by executing these multiple subtasks.

[0105] With this configuration, for example, when generating an agent that extracts information from a website and transfers it to a file, it becomes possible to build agents that can handle a wide variety of tasks, such as "an agent that extracts information from a website and transfers it to a file" or "an agent that extracts and transfers information." This makes it possible to build a versatile, autonomous agent that can process and learn a series of business processes with a single agent and be applied to different tasks.

[0106] <Information Processing Systems / Summary> As described above, the information processing system according to this embodiment (for example, the information processing system 10 shown in Figures 1 and 2) is an information processing system configured to have an AI agent (for example, the AI ​​agent 12 shown in Figure 1, the system terminal 12 shown in Figure 2), wherein the AI ​​agent comprises a task decomposition means (for example, the task decomposition means 12a shown in Figures 1 and 2) capable of decomposing a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means (for example, the task execution means 12b shown in Figures 1 and 2) capable of executing the plurality of subtasks in a predetermined processing order, and a task evaluation means (for example, the task evaluation means 12c shown in Figures 1 and 2) that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks.

[0107] Furthermore, the information processing method according to this embodiment (for example, the method executed by the information processing system 10 shown in Figures 1 and 2) is an information processing method executed using an AI agent which is a computer (for example, the AI ​​agent 12 shown in Figure 1, the system terminal 12 shown in Figure 2), wherein the AI ​​agent is configured to perform at least a task decomposition step (for example, a process executed by the task decomposition means 12a shown in Figures 1 and 2) which can decompose a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution step (for example, a process executed by the task execution means 12b shown in Figures 1 and 2) which can execute the plurality of subtasks in a predetermined processing order, and a task evaluation step (for example, a process executed by the task evaluation means 12c shown in Figures 1 and 2) which evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks.

[0108] Furthermore, the program for the information processing system according to this embodiment (for example, the information processing system 10 shown in Figures 1 and 2) is a program for an information processing system configured to have an AI agent (for example, the AI ​​agent 12 shown in Figure 1, the system terminal 12 shown in Figure 2), and is characterized in that the AI ​​agent, which is a computer, functions as a task decomposition means (for example, the task decomposition means 12a shown in Figures 1 and 2) capable of decomposing a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means (for example, the task execution means 12b shown in Figures 1 and 2) capable of executing the plurality of subtasks in a predetermined processing order, and a task evaluation means (for example, the task evaluation means 12c shown in Figures 1 and 2) that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks.

[0109] Furthermore, the AI ​​agent according to this embodiment (for example, the AI ​​agent 12 shown in Figure 1, the system terminal 12 shown in Figure 2) is an AI agent that is a computer that operates as a substitute for a human and has the ability to learn and make decisions on its own, and is configured to function as a task decomposition means (for example, the task decomposition means 12a shown in Figures 1 and 2) that can decompose a predetermined task into a plurality of subtasks using a predetermined decomposition method, a task execution means (for example, the task execution means 12b shown in Figures 1 and 2) that can execute the plurality of subtasks in a predetermined processing order, and a task evaluation means (for example, the task evaluation means 12c shown in Figures 1 and 2) that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, and is configured to review the predetermined decomposition method and the predetermined processing order based on the evaluation of the plurality of subtasks, and is configured to learn the relationship between the predetermined decomposition method and the predetermined processing order and learn the relationship between the predetermined decomposition method and the predetermined processing order and the predetermined task, by repeatedly providing the predetermined decomposition method and the predetermined processing order as learning data.

[0110] According to the information processing system, information processing method, information processing program, and AI agent of this embodiment, the execution accuracy and processing speed of a predetermined task can be improved by evaluating and weighting multiple tasks that constitute a predetermined task, and by reviewing the decomposition method and processing order of the predetermined task.

[0111] Furthermore, the task decomposition means may be configured to revise the predetermined decomposition method based on the evaluation of the plurality of subtasks, and to decompose the predetermined task into a plurality of subtasks in a manner different from the predetermined decomposition method.

[0112] With this configuration, by breaking down a given task into multiple subtasks using multiple decomposition methods, it is possible to find a decomposition method for subtasks that is suitable for the given task, thereby improving the execution accuracy and processing speed of the given task.

[0113] Furthermore, the task execution means may be configured to revise the predetermined processing order based on the evaluation of the plurality of subtasks and execute the plurality of subtasks in a processing order different from the predetermined processing order.

[0114] With this configuration, by executing multiple subtasks in multiple processing sequences, it is possible to find a processing sequence for subtasks that is suitable for a given task, thereby improving the execution accuracy and processing speed of the given task.

[0115] Furthermore, the predetermined decomposition method may include a method for decomposing the predetermined task into N subtasks (where N is a positive integer of 2 or more), and a method different from the predetermined decomposition method may include a method for decomposing the predetermined task into M subtasks (where M is a positive integer of 2 or more) that are different from the N subtasks.

[0116] With this configuration, by breaking down a given task into multiple subtasks using multiple decomposition methods, it is possible to find a decomposition method for subtasks that is suitable for the given task, thereby improving the execution accuracy and processing speed of the given task.

[0117] Furthermore, the predetermined decomposition method and the predetermined processing order may be reviewed until the evaluation of all of the multiple subtasks reaches a predetermined level.

[0118] With this configuration, it is possible to find the optimal method and processing order for breaking down subtasks into a given task, thereby improving the execution accuracy and processing speed of that task.

[0119] Furthermore, the task decomposition means may be configured to reduce the number of subtasks based on an evaluation of the subtasks, and the task execution means may be configured to execute some of the subtasks in parallel based on an evaluation of the subtasks.

[0120] With this configuration, the processing speed of a given task can be increased by increasing the processing speed of subtasks.

[0121] Furthermore, the predetermined decomposition method and the predetermined processing sequence may be repeatedly provided as training data, and the system may be configured to learn the relationship between the predetermined decomposition method and the predetermined processing sequence and the predetermined task.

[0122] With this configuration, the processing speed of a given task can be increased by improving the processing speed of the task decomposition means and the task execution means.

[0123] It should be noted that the information processing system, information processing method, information processing program, and AI agent according to the present invention are not limited to the embodiments described above, and various modifications can be made without departing from the spirit of the present invention.

[0124] Therefore, for example, the "predetermined task" according to the present invention is not limited to a prompt given from an external terminal 16, but may also be a task that arises based on an external event (e.g., a signal input from a sensor) or an internal event (e.g., activation by an internal scheduler, occurrence of an anomaly). [Industrial applicability]

[0125] The information processing system, information processing method, information processing program, and AI agent according to the present invention can be widely applied in manufacturing, service, retail, and other industries. [Explanation of Symbols]

[0126] 10 Information Processing Systems 12 System Terminals 12a Task decomposition means 12b Task execution means 12c Task evaluation method 16 External terminals 21 CPU 22 ROM 23 RAM 24 Recording media 25 External storage drives 26 Storage device 27 Input devices 28 Display device 29 Communications Department

Claims

1. An information processing system configured to include an AI agent, The aforementioned AI agent, A task decomposition means capable of decomposing a predetermined task into multiple subtasks using a predetermined decomposition method, Task execution means capable of executing the plurality of subtasks in a predetermined processing order, The system includes a task evaluation means that evaluates the plurality of subtasks based on the execution results of the plurality of subtasks, The system is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the aforementioned multiple subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the magnitude of the processing burden, The task execution means can execute the plurality of subtasks in order of increasing processing load, The task evaluation means assigns weights to each of the multiple subtasks based on the execution results and evaluation indicators of the multiple subtasks. The task decomposition means decomposes or deletes subtasks if the weighting does not meet a predetermined standard. An information processing system characterized by the following:

2. In the information processing system described in claim 1, The task decomposition means is capable of decomposing a predetermined task into a plurality of subtasks according to the magnitude of the processing burden, such that the processing burden of the decomposed subtasks is averaged. An information processing system characterized by the following:

3. In the information processing system according to claim 1 or 2, The task execution means can refer to the multiple subtasks and execute subtasks that are not dependent on each other in parallel, while subtasks that are dependent on each other can be executed in an order according to their dependencies. An information processing system characterized by the following:

4. An information processing method that is performed using an AI agent, which is a computer, The aforementioned AI agent, A task decomposition step that allows a predetermined task to be broken down into multiple subtasks using a predetermined decomposition method, A task execution step that enables the execution of the plurality of subtasks in a predetermined processing order, A task evaluation step is performed to evaluate the plurality of subtasks based on the execution results of the plurality of subtasks, and at least the following steps are performed: The system is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the aforementioned multiple subtasks. The task decomposition step allows the predetermined task to be decomposed into a plurality of subtasks according to the magnitude of the processing burden, The task execution step allows the multiple subtasks to be executed in order of increasing processing load, The task evaluation step involves weighting each of the multiple subtasks based on the execution results and evaluation indicators of the multiple subtasks. The task decomposition step includes decomposing or deleting subtasks if their weighting does not meet a predetermined standard. An information processing method characterized by the following:

5. A program for an information processing system configured to include an AI agent, The aforementioned AI agent, which is a computer, A task decomposition means capable of decomposing a predetermined task into multiple subtasks using a predetermined decomposition method, Task execution means capable of executing the plurality of subtasks in a predetermined processing order, This is configured to function as a task evaluation means that evaluates the multiple subtasks based on the execution results of the multiple subtasks, The system is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the aforementioned multiple subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the magnitude of the processing burden, The task execution means can execute the plurality of subtasks in order of increasing processing load, The task evaluation means assigns weights to each of the multiple subtasks based on the execution results and evaluation indicators of the multiple subtasks. The task decomposition means decomposes or deletes subtasks if the weighting does not meet a predetermined standard. An information processing program characterized by the following features.

6. An AI agent is a computer that acts as a substitute for a human and has the ability to learn and make decisions on its own. The aforementioned AI agent, A task decomposition means capable of decomposing a predetermined task into multiple subtasks using a predetermined decomposition method, Task execution means capable of executing the plurality of subtasks in a predetermined processing order, This is configured to function as a task evaluation means that evaluates the multiple subtasks based on the execution results of the multiple subtasks, The system is configured to revise the predetermined decomposition method and the predetermined processing order based on the evaluation of the aforementioned multiple subtasks. The task decomposition means can decompose the predetermined task into a plurality of subtasks according to the magnitude of the processing burden, The task execution means can execute the plurality of subtasks in order of increasing processing load, The task evaluation means assigns weights to each of the multiple subtasks based on the execution results and evaluation indicators of the multiple subtasks. The task decomposition means, if there are subtasks whose weighting does not meet a predetermined standard, decomposes or deletes the subtasks. The predetermined decomposition method and the predetermined processing sequence are repeatedly provided as training data, and the system is configured to learn the relationship between the predetermined decomposition method and the predetermined processing sequence and the predetermined task. An AI agent characterized by the following.

Images