JP7446621B2 - Computer system and personal information provision method - Google Patents

Description

The present invention relates to a computer system for processing personal information and a method for providing personal information using a computer .

Individuals such as employees deposit their personal data such as life log data to a server that serves as an "information bank" via a network, and when the individual consents, the personal data is disclosed to a predetermined company. Next, when the personal data is disclosed, the company analyzes the personal data, predicts services, etc. according to each person's life stage, and provides information. In technology for building such a business model, for example, Patent Document 1 provides an information processing system, an information processing method, and a program that prevent personal information from being leaked to persons with illicit purposes. .

Patent No. 6342094

The purpose of the invention is to enable users to put their personal information under self-control .

The present invention is a computer system that processes personal information,
a personal information storage means storing encrypted personal information obtained by encrypting the user's personal information;
search enablement means for making the encrypted personal information searchable while encrypted;
If the desired personal information is found as a result of the search using the search enablement means, a transaction is concluded between the person requesting the personal information and the person providing the personal information, and the personal information is transferred. A transaction execution means provided to those who wish to do so;
The search enablement means includes:
an encrypted index recording means for recording on a blockchain an encrypted index that has been subjected to an encryption process of encrypting an index for keyword searching the encrypted personal information using a common key;
a search means for searching the encrypted index using an encrypted search query that has been subjected to an encryption process of encrypting a search query that is a keyword used for the search with the common key,
The transaction execution means is
A means for making personal information provided to the applicant available for viewing by the applicant;
a smart contract processing means that uses a smart contract to transact the personal information when transaction conditions according to the intention of the person providing the personal information are satisfied;
The smart contract processing means further includes recording means for recording transactions performed by the smart contract processing means on a blockchain.

Another aspect of the present invention is a personal information providing method using a computer, comprising:
storing encrypted personal information obtained by encrypting the user's personal information;
a search enablement step for making the encrypted personal information searchable while encrypted;
If the desired personal information is found as a result of the search in the search enablement step, a transaction is concluded between the person requesting the personal information and the person providing the personal information, and the personal information is transferred. A transaction execution step provided to a person who wishes to do so;
The searchability step includes:
an encrypted index recording step of recording on a blockchain an encrypted index that has been subjected to an encryption process of encrypting an index for keyword searching the encrypted personal information using a common key;
a search step of searching the encrypted index using an encrypted search query that has been subjected to an encryption process of encrypting a search query that is a keyword used for the search with the common key,
The transaction execution step includes:
a viewing enablement step for allowing the applicant to view the personal information provided to the applicant;
a smart contract processing step for performing a transaction of the personal information using a smart contract when transaction conditions according to the intention of the person providing the personal information are satisfied;
The method further includes a recording step of recording the transaction performed by the smart contract processing step on a blockchain.

According to the present invention , the user's personal information is encrypted, and the encrypted personal information is searched while being encrypted, and when the desired personal information is found as a result of the search, the desired personal information is searched. The personal information is provided to the person who requests it after a transaction is concluded based on an agreement between the person providing the personal information and the person providing the personal information . Therefore , it becomes possible for users to put their own personal information under their own control.

FIG. 1 is a system diagram showing the overall configuration of a processing system. (A) is a diagram illustrating information stored in the HDD of a user terminal that constitutes a blockchain node, and (B) is a diagram illustrating information stored in a personal information DB of a certified business operator. be. (A) is a flowchart showing a main routine program of a user terminal of a public chain, and (B) is a flowchart showing a subroutine program of personal information recording processing and a flowchart of a server of a certified business operator. It is a flowchart showing a subroutine program of smart contract processing executed on a user terminal of a public chain. (A) is a flowchart showing a main routine program of a user terminal of a private chain, and (B) is a flowchart showing a subroutine program of personal information search processing. 2 is a flowchart showing a subroutine program for smart contract processing executed on a user terminal of a private chain. (A) is a continuation of the flowchart showing the subroutine program for smart contract processing executed on the user terminal of the private chain, and (B) is a flowchart showing the subroutine program for machine learning processing executed on the user terminal of the private chain. be. 2 is a flowchart showing a subroutine program of AI smart contract generation processing executed on a user terminal of a private chain. (A) is a flowchart showing a subroutine program for simulation learning processing executed on a user terminal of a private chain, and (B) is a flowchart showing a subroutine program for AI smart contract group generation processing executed on a user terminal of a private chain. It is. 2 is a flowchart showing a subroutine program for smart contract trust contract processing executed on a user terminal of a private chain. (A) is a flowchart showing a subroutine program for reinforcement learning processing of a personalized AI smart contract trained model, and (B) is a flowchart showing a main routine program executed by a user terminal of a consortium chain. (A) is a flowchart showing a subroutine program of IoT sensor data aggregation processing executed by a user terminal of a consortium chain, and (B) is a flowchart showing a subroutine program of simulation processing executed by a user terminal of a consortium chain. . (A) is a flowchart showing a subroutine program for AI smart contract group generation processing executed by a user terminal of a consortium chain, and (B) is a flowchart showing a subroutine program for smart contract processing executed by a user terminal of a consortium chain. It is. This is an explanatory diagram that makes information recorded using blockchain unreadable. (A) is a normal state where it can be viewed, and (B) is a diagram where it is rendered indecipherable so that it cannot be viewed. It is. FIG. 2 is a diagram illustrating information stored in an HDD of a user terminal that constitutes a node of a blockchain. 3 is a flowchart showing a main routine program for a public chain user terminal and a private chain user terminal. (A) is a flowchart showing a subroutine program for personal information recording processing on a blockchain, and (B) is a flowchart showing a subroutine program for processing that records cannot be read. 3 is a flowchart showing a subroutine program for personal information acquisition processing and personal information provision processing. (A) is a diagram illustrating information stored in the HDD of a user terminal that constitutes a blockchain node, and (B) is a diagram illustrating information stored in a personal information DB of a certified business operator. be. 2 is a flowchart showing a main routine program for a public chain user terminal, a server of an authorized business operator, and a private chain user terminal. It is a flowchart showing a subroutine program of personal information recording processing and hash value recording processing on a blockchain. 3 is a flowchart showing a subroutine program for record unreadable processing. 3 is a flowchart showing a subroutine program of personal information provision processing, personal information acquisition processing, and ciphertext transmission processing. FIG. 1 is a system diagram showing the overall configuration of a processing system. 2 is a flowchart showing a main routine program of a public chain user terminal, a key registration center server, and a private chain user terminal. (A) is a flowchart showing a subroutine program for personal information recording processing and key registration processing on a blockchain, and (B) is a flowchart showing a subroutine program for record unreadable request processing and record unreadable processing. 12 is a flowchart showing a subroutine program of a half common key providing process, a data acquisition process, and a data decryption process. FIG. 2 is an explanatory diagram of a mirror world as a simulation environment. It is a diagram showing a specific example of a city digital twin in a mirror world. It is a flowchart of the main routine of a mirror world server and a user terminal. 2 is a flowchart showing a subroutine program of personal AI generation and sales processing. 3 is a flowchart showing a subroutine program of simulation preparation processing and simulation preparation response processing. 3 is a flowchart showing a subroutine program of simulation processing. (A) is a schematic diagram of the multi-service DAO construction system, and (B) is a flowchart of the main routine between the mirror world server and the user terminal. 3 is a flowchart showing a subroutine program of simulation reinforcement learning preparation processing and simulation reinforcement learning preparation response processing. It is an explanatory diagram of registering a multi-service DAO digital twin as a simulation target in the mirror world. It is a schematic system diagram of simulation reinforcement learning of multi-service DAO. 3 is a flowchart showing a subroutine program of DAO agent reinforcement learning processing. (A) is a diagram showing a reward table stored as knowledge by the DAO agent, and (B) is a flowchart showing a subroutine program of persona agent reinforcement learning processing. (A) is a flowchart showing a subroutine program for idea generation service execution processing, and (B) is a flowchart showing a subroutine program for improved idea service execution processing. (A) is a flowchart showing a subroutine program for commercialized service execution processing, and (B) is a flowchart showing a subroutine program for anti-infringement service execution processing. (A) is a flowchart showing a subroutine program of token purchase service execution processing, and (B) is a diagram illustrating price fluctuations in a fluctuating market of tokens accompanying token purchases by a group of persona agents. It is a diagram showing an element integrated DAO construction system. (A) is a flowchart of the main routine between the mirror world server and the user terminal, and (B) is a flowchart showing a subroutine program of simulation reinforcement learning preparation processing and simulation reinforcement learning preparation response processing. It is an explanatory diagram of registering an element integrated DAO digital twin as a simulation target in the mirror world. 1 is a diagram illustrating a schematic system illustrating simulation reinforcement learning of an element-integrated DAO digital twin; FIG. (A) is a diagram showing the performance and distribution rate calculation algorithm stored as knowledge by the material procurement element agent, and (B) is a diagram showing the performance and distribution rate calculation algorithm stored as knowledge by the assembly element agent. FIG. 3C is a diagram showing a performance and distribution rate calculation algorithm stored as knowledge by the advertising element agent. (A) is a diagram showing performance and distribution rate calculation algorithms stored as knowledge by the sales element agent, and (B) is a diagram showing a remuneration table stored as knowledge by the supervising agent. 3 is a flowchart showing a subroutine program of simulation reinforcement learning processing. 3 is a flowchart showing a subroutine program of a supervising agent reinforcement learning process. 12 is a flowchart showing a subroutine program of material procurement element agent reinforcement learning processing. (A) is a flowchart showing a subroutine program of information collection processing by a crawler, and (B) is a diagram showing various data stored in the material procurement DB. 3 is a flowchart showing a subroutine program of assembly element agent reinforcement learning processing. It is a flowchart which shows the subroutine program of promotion element agent reinforcement learning processing. (A) is a flowchart showing a subroutine program of information collection processing by a crawler, and (B) is a diagram showing various data stored in the advertisement DB. It is a flowchart which shows the subroutine program of sales element agent reinforcement learning processing. (A) is a flowchart showing a subroutine program of information collection processing, and (B) is a diagram showing various data stored in the sales DB. (A) is a flowchart showing a subroutine program of reinforcement learning processing by the personal AI in charge of material procurement, and (B) is a flowchart showing a subroutine program of reinforcement learning processing by the personal AI in charge of assembly. (A) is a flowchart showing a subroutine program of AI reinforcement learning processing for a person in charge of advertising, and (B) is a flowchart showing a subroutine program of AI reinforcement learning processing for a person in charge of sales. FIG. 2 is an explanatory diagram showing a program installation method.

[First embodiment]
A first embodiment of the present invention will be described based on FIGS. 1 to 13. First, referring to the overall system of FIG. 1, three types of blockchain networks, a private chain 2, a consortium chain 3, and a public chain 4, are connected to a centralized oracle 21. Public Chain 4 is a completely open system, allowing any individual or organization to transact there. Transactions can be effectively confirmed on the blockchain. Mining (competition for bookkeeping rights) is also free and anyone can participate. Consortium Chain 3 is a blockchain that can only be used by partners belonging to an association or union. People within it (each node) are designated as bookkeepers. Block generation is also predetermined, and other people (nodes) can trade but do not have bookkeeping rights. Private Chain 2 only records books using blockchain technology, and the right to record books is not open, but is monopolized by individuals or companies, and records only internal transactions. Polkadot is used to connect blockchains and exchange tokens and data between each blockchain. Polkadot is a blockchain that connects different blockchains. Blockchains developed using Substrate can be connected to Polkadot, and by connecting to Polkadot, they will be able to exchange tokens and data with other blockchains connected to Polkadot.

The centralized oracle 21 is a system that bridges data between the blockchain and the Internet 1, and is connected to the Internet 1 to collect various information scattered on the Internet and provide the information to the blockchain smart contract. I do.

Each node 19 of the private chain 2, consortium chain 3, and public chain 4 is configured with a user terminal such as a personal computer (hereinafter referred to as "PC") 16. This PC (hereinafter also referred to as "user terminal") 16 is connected to the Internet 1. Further connected to the Internet 1 are a server 20 of an SNS (Social Networking Service) 40 and a server 18 of a blockchain certified business operator 17. Note that the server 18 of the certified business operator 17 may participate in the blockchain as a node 19. Furthermore, a server of a certification authority that issues electronic certificates in PKI (Public Key Infrastructure) may be connected to the Internet 1.

The certified business operator 17 receives personal information, issues an electronic ID to the personal quasi-information, and records the hash value of the personal information on the blockchain. The deposited personal information is stored in a personal information database (hereinafter referred to as "personal information DB") 29. Note that the certified business operator 17 may participate in the blockchain as a node 19.

The PC 16 includes a CPU (Central Processing Unit) 10 as a control center, a RAM (Random Access Memory) 9 that functions as a work area for the CPU 10, a ROM (Read Only Memory) 11 that stores data and programs, and an HDD (hard disk). 12, an input operation section 7 such as a display and a keyboard, a communication section 5, a display section 6, an interface 8, a bus 13, and various other hardware. Various servers such as the server 20 and the server 18 are also configured with the same hardware as the PC 16, and repeated illustrations and explanations will be omitted here. Note that an SDD (Solid State Drive) may be used as the storage unit in addition to or in place of the HDD described above.

The node 19 of the consortium chain 3 is connected to an IoT (Internet of Things) device 14 and a wireless sensor network 15. Sensor signals from the IoT device 14 and the wireless sensor network 15 are input to the node 19, and a drive signal for the IoT device 14 is output from the node 19. The IoT device 14 is various sensors, actuators, etc. for IoT.

The wireless sensor network 15 is a wireless network in which a plurality of sensor-equipped wireless terminals are scattered in space, and these terminals cooperate to collect information about the environment and physical conditions. For example, we can create a sensor device using energy harvesting, M2M, or batteries, and use pressure sensors or gauge sensors to constantly monitor deterioration of metal fatigue, etc., and report any changes. Mainly installed in buildings such as bridges and tunnels. Generally includes a plurality of sensor nodes and a gateway sensor node. These nodes typically consist of one or more sensors, a wireless chip, a microprocessor, and a power source (such as a battery). Wireless sensor networks typically have ad hoc functionality and a routing algorithm to send data from each node to a central node. In other words, it has the ability to autonomously reconstruct another communication route if a failure occurs in communication between nodes. There is also an element of distributed processing as nodes work together as a group. In addition, it has the ability to operate for long periods of time without receiving power from an external source, and therefore has a power saving function or self-power generation function.

In this embodiment, the IoT device 14 and the wireless sensor network 15 are connected to the consortium chain 3 via the node 19; It may itself be part of the node 19 of the consortium chain 3.

Next, information stored in the HDD 12 of the PC 16 will be explained with reference to FIG. 2(A). The HDD 12 stores the user's private key SK, public key PK, common key K1, trapdoor common key K2, user's blockchain address, smart contract, token, and artificial intelligence (also called "AI (Artificial Intelligence)"). ), blockchain data, etc. are stored. Note that a user is a broad concept that includes not only natural persons but also corporations.

The private key SK and public key PK are a key pair used in PKI (Public Key Infrastructure), and data encrypted with the public key PK is decrypted with the private key SK. The private key SK is also used for electronic signatures. The common key K1 is a key used for common key encryption such as DES (Data Encryption Standard) and AES (Advanced Encryption Standard). Data encrypted with the common key K1 is decrypted using the same common key K1. In this embodiment, a different common key is used for each piece of personal information to be encrypted. In the first embodiment, an index for keyword search is provided for encrypted personal information E _K1 (personal information) in cipher text. The index is encrypted using the common key K2. In order to perform a keyword search, a search is performed using an encrypted search query (referred to as a "trapdoor") in which the keyword (search query) used for the search is encrypted using the common key K2. This common key K2 is stored in the HDD 12 as a trap door common key K2.

A user's address on the blockchain is generated through the following process.
1 Generate a public key using ECDSA from the private key.
2 Pass the public key through the hash function SHA-256 to obtain the hash value.
3 Pass the hash value further through hash function RIPEMD-160 to obtain a hash value.
4 Add 00 as a prefix to the beginning of the hash value.
5 Pass through hash function SHA-256.
6 Pass it through the hash function SHA-256 again.
7 Add a 4-byte checksum to the end.
8 Encode in Base58 format.

A smart contract is a computer protocol intended for smooth verification, condition confirmation, execution, execution, and negotiation of contracts. A token is a unique currency issued on a blockchain by a company or individual.

Next, we will explain blockchain data. The data in each block of the blockchain includes the hash value of the previous block, a nonce, and data of multiple transactions (also called transactions). Additionally, although not shown, a timestamp is also embedded in the blockchain. Such a blockchain is generated by each node 19 performing blockchain processing (see S3, S19, S30, S51, S117, S122, S153, etc. described later) and is added as a new blockchain. Blockchain processing mainly consists of three phases: transaction, propagation and recording.

The transaction phase is generally referred to as a transaction, and refers to legal acts such as buying and selling, transfer, and lending. More specifically, this transaction phase can be divided into three phases: generation → signature → propagation.

The generation phase is to generate a transaction. For example, Mr. A decides to ``lend dormant PC resources (computing resources) for 39005 seconds and obtain 25.78 tokens'' to Mr. B, and the transaction is generated. electronically sign that it was generated. This electronic signature generates a hash value by passing transaction data through a predetermined hash function, and encrypts the hash value using the private keys SK of the parties to the transaction (Mr. A and Mr. B). It is. Alternatively, a digital public key certificate may be issued by a certification authority. Although Figure 2 shows an example of lending PC resources (computing resources), the scope of lending is not limited to this, and includes, for example, electric power self-generated at home or business, the user's specialized knowledge and experience, and skills. Possible values include personal connections (including online human networks), trust, and other values.

The propagation phase is to have other nodes confirm that the transaction has been correctly generated and signed. If it is determined that the transaction was not correctly generated and signed, the transaction will be discarded.

In the recording phase, when it is confirmed that the transaction has been correctly generated and signed, the miner performs mining and records the transaction. Transactions that are confirmed to have been correctly generated and signed are moved to a location called a mining pool. Then, miners select transactions to record from the mining pool and mine them.

Mining is the process of calculating nonces. A nonce is a value used to adjust the hash value so that when a block's data is passed through a hash function, a very small hash value with many leading zeros is generated. A new block is generated when a nonce whose hash value is less than or equal to the target value can be calculated.

Furthermore, as shown in transaction I on the right side of Figure 2, the transaction data includes the hash value of E _K1 (personal information), which is the user's personal information encrypted with the key K1, its electronic ID, and the index of the personal information. It also includes the consideration for providing personal information (provided with 2.4 tokens in Figure 2) and E _K2 (provided with index + 2.4 tokens) which is encrypted with key K2. Specific examples of personal information include vital information such as the user's heart rate, blood pressure, body temperature, and brain waves, behavioral history information such as purchase history and website browsing history, user location information such as GPS, race, creed, and society. Personal status, medical history, electronic medical record data, ID (identification), information posted to SNS, etc.

The information posted to the SNS etc. is obtained by transferring past posted information that has already been posted to the SNS 25 and stored in the server 20 from the server 20 to the personal information DB 29 and the blockchain. Specifically, the user encrypts all of his/her past posted information and stores it in the personal information DB 29 of the authorized business operator, and also records the hash value on the blockchain. After that, the user does not post on the SNS 25, but encrypts and stores the posted content in the certified business's personal information DB 29, and records the hash value on the blockchain. This makes it possible for the user to retrieve personal information from a business operator such as an SNS and put it under self-management.

Note that the compensation for providing personal information (2.4 tokens in Figure 2) may be recorded in plain text on the blockchain without being encrypted. In that case, other users can search the blockchain and know the price without having to obtain the encryption key K2. Furthermore, transaction conditions such as compensation for providing personal information and compensation for lending PC resources (computing resources) (in Figure 2, 25.78 tokens are obtained by lending PC resources (computing resources) for 39005 seconds) are set in the smart contract. It is also possible to put it in the form of code and automate transactions (legal acts) using smart contracts.

The encrypted personal information itself, which is the object of the hash value recorded as transaction I, is stored in the personal information DB 29 of the authorized business operator 17. Specifically, as shown in FIG. 2(B), the certified business operator 17 issues the encrypted personal information E _K1 (personal information) in association with the electronic ID issued to the encrypted personal information E _K1 (personal information). is stored in the personal information DB 29.

The index is an index for keyword searching for encrypted personal information E _K1 (personal information). In this embodiment, a common key encryption method is adopted in which the index is encrypted using the common key K2. Therefore, in order to perform a keyword search, the keyword (search query) used for the search must be encrypted using the common key K2. Search using a search query (this is called a "trapdoor"). The common key K2 is a different key for each user, but the same key is used for encrypted indexes of the same user. Therefore, using the smart contract described below, for example, if user A performs a transaction to distribute common key K2 to user B, user B will use E _K2 (search query) to retrieve all encrypted indexes of user A. It will be searchable on the blockchain.

Note that a searchable encryption such as a homomorphic encryption or a completely homomorphic encryption, which allows searching while encrypting the ciphertext, may be used. At that time, personal information may be encrypted using homomorphic encryption or completely homomorphic encryption, and the encrypted personal information may be directly recorded on the blockchain. Alternatively, the encrypted personal information E _K1 (personal information) may be directly recorded on the blockchain.

Next, a flowchart of the main routine program of the user terminal of the public chain 19 will be described with reference to FIG. 3(A). Personal information recording processing is performed in step S (hereinafter simply referred to as "S") 1, smart contract processing is performed in S2, and blockchain processing is performed in S3.

The personal information recording process is a process in which the owner of personal information encrypts personal information, registers it with the authorized business operator 17, and records the hash value of the encrypted personal information on the blockchain. Smart contract processing is processing that automatically performs legal acts such as conclusion and enforcement of contracts according to predetermined rules. The specific content of the blockchain processing is as described above based on FIG. 2(A).

The personal information recording process will be explained with reference to FIG. 3(B). In S5, it is determined whether a personal information registration operation has been performed on a user terminal forming a node 19 of the public chain 19. If there is no such personal information, the personal information recording process returns and proceeds to the smart contract process in S2. If it is determined that a personal information registration operation has been performed, in S6, the personal information stored in the memory of the user terminal (HDD 12, etc.) is encrypted with the key K1, and then digitally signed with the private key SK. Moreover, the index is encrypted with the key K2 and transmitted to the server 18 of the authorized business operator 17.

The server 18 of the authorized business operator 17, which has received it in S7, issues an electronic ID and generates a hash value of the received encrypted personal information _EK1 (personal information). Next, the issued electronic ID is returned to the user terminal (S9). The user terminal that receives it stores the electronic ID in its memory (HDD 12, etc.). In S10, the server 18 of the certified business operator 17 performs processing for recording the electronic ID, the hash value, and the encrypted index on the blockchain.

A flowchart of the smart contract processing subroutine program shown in S2 will be explained. Referring to FIG. 4, in S13, it is determined whether a distribution contract for the common key K2 has been established, and if not, in S15, a loan contract for PC resources (computing resources) of the user terminal has been established. If not, it is determined in S16 whether or not a personal information provision contract has been established, and if not, it is determined in S17 that an ordering contract for ordering custom-made products, etc. has been established. If not, it is determined in S22 whether or not a sales contract for the product, etc. has been concluded, and if not, the process returns. These decisions are made by smart contracts. For example, if the consideration mentioned above is coded as a smart contract, the smart contract determines whether the conditions of consideration, etc. of both parties match, and if it is determined that they match, the contract is concluded and Execute automatically.

If it is determined that the distribution contract for the common key K2 has been established, the control proceeds to S14, and after the common key K2 is transmitted to the distribution destination, the control proceeds to S19. In S19, processing is performed to record the established contract as a transaction on the blockchain. If it is determined that a loan contract for PC resources (computing resources) has been established, the control proceeds to S18, and a process for lending PC resources (computing resources) is performed. If it is determined that a personal information provision contract has been established, the control proceeds to S20, in which the electronic ID of the personal information to be provided and the signature agreeing to the provision of personal information are returned to the providing party, and the personal information to be provided is The common key K1 used for encryption is encrypted with the public key of the provider and sent back to the provider.

If it is determined that an ordering contract has been established, control proceeds to S21, and after ordering processing is performed, control proceeds to S19. If it is determined that a sales contract has been established, the control proceeds to S23, a process for obtaining the purchase object is executed, and the process proceeds to S19.

Next, a flowchart of the main routine program of the user terminal configuring the node 19 of the private chain 2 will be explained based on FIG. 5(A). Personal information search processing is performed in S28, smart contract processing is performed in S29, blockchain processing is performed in S30, machine learning processing is performed in S31, AI smart contract generation processing is performed in S33, and Smart contract trust contract processing is performed. An AI smart contract is a concept that includes both "integration type" and "cooperation type". The "integrated type" is one in which AI and a smart contract are integrated, machine learning is performed based on contract (legal act) data, and the smart contract itself is made into AI. The "cooperation type" is one in which AI that performs machine learning based on contract (legal act) data and a smart contract collaborate. In the case of the collaboration type, the trained AI (hereinafter referred to as "cooperation AI") adds, changes, updates, etc. to the smart contract depending on the situation.

Personal information search processing is a process of searching an encrypted index recorded on a blockchain using a trapdoor (encrypted search query). Smart contract processing is processing that automatically performs legal acts such as conclusion and enforcement of contracts according to predetermined rules. The specific content of the blockchain processing is as described above based on FIG. 2(A). Machine learning processing is processing that generates a trained model of artificial intelligence by performing machine learning using personal information of a large number of users as learning data. More specifically, after using machine learning as training data to generate a general trained model of artificial intelligence, a huge amount of personal information in a form that cannot identify the owner of the personal information, data that can identify the owner of the personal information (for example, block This is a process of generating a personalized trained model personalized for each individual information owner (each address on the blockchain) using personal information classified by each address in the chain.

The AI smart contract generation process is a process in which personal information related to contracts (legal acts) is used as learning data for machine learning to generate a trained model of a smart contract using artificial intelligence. More specifically, after using machine learning as learning data on a huge amount of personal information related to contracts (legal acts) in which the owner of the personal information cannot be identified, a general trained model for smart contracts using artificial intelligence is generated. A personalized AI smart contract that is personalized for each individual information owner (e.g., each blockchain address) using personal information related to contracts (legal acts) classified by data that can identify the information owner (e.g., blockchain address). This is the process of generating a trained model.

Smart contract trust contract processing is a process that provides a service in which a person contracts out processing to automatically perform legal acts such as contract conclusion and execution on behalf of the principal. More specifically, a personalized AI smart contract trained model is generated for the trustor, and the personalized AI smart contract trained model is used to execute a legal act on behalf of the trustor. A reward for the AI is determined based on the result of the execution, and the trained model of the personalized AI smart contract is further trained using the reward.

Next, a flowchart of the subroutine program of the personal information search process shown in S28 will be explained based on FIG. 5(B). In S37, it is determined whether the common key K2 is stored, and if not, the process returns. If K2 is stored in S45, which will be described later, it is determined in S37 that the common key K2 is stored, and control proceeds to S38. In S38, processing is performed to search the encrypted index on the blockchain using the search query (trapdoor) encrypted with K2. As a result of the search, it is determined in S39 whether or not there is personal information to be obtained. If it is determined that there is no personal information to be obtained, the process returns, but if it is determined that there is personal information to be obtained, the electronic ID of the personal information to be obtained is stored in S40.

Next, a flowchart of the smart contract processing subroutine program shown in S29 will be explained based on FIGS. 6 and 7(A). In S42, it is determined whether or not there is personal information desired to be searched. This determination is made, for example, by sequentially negotiating with the smart contract of a personal information owner who has not yet been searched, and when the conditions are met, the personal information of that personal information owner is determined to be the personal information desired to be searched. If it is determined in S42 that there is no personal information desired to be retrieved, it is determined in S46 whether or not the personal information desired to be obtained is stored, and if it is determined that there is no personal information desired to be obtained, control proceeds to S55 in FIG. 7(A). , it is determined whether a loan contract for PC resources (computing resources) has been established, and if it is determined that it has not been established, it is determined in S56 whether or not an ordering contract has been established, and if it has not been established. If it is determined, it is determined in S57 whether or not a sales contract has been established, and if it is determined that it has not been established, the process returns.

If it is determined in S42 that there is personal information desired to be searched, control proceeds to S43, where a common key K2 is requested from the personal information owner. Specifically, it requests the common key K2 by transmitting its own address and attribute certificate to the address on the blockchain of the owner of the personal information desired to search. In step S44, it is determined whether there is a reply of K2 or not, and the process waits until there is a reply. The personal information owner or the smart contract of the personal information owner checks the transmitted attribute certificate and determines whether or not K2 can be returned, and if it is determined that it is okay to reply, returns K2. When the personal information owner returns K2, control proceeds to S45, and after storing the returned K2, control proceeds to S54. In S54, the established contract is stored in the blockchain as a transaction. In this case, a contract indicating that the common key K2 used to encrypt the index is distributed from the address of the person who sent the reply to the address of the user who received the reply is stored in the blockchain.

If it is determined in S46 that the personal information desired to be obtained is stored, the control proceeds to S47, where processing is performed to request the personal information owner to request the personal information desired to be obtained. Specifically, the user requests the personal information he or she wishes to obtain by transmitting his or her address and attribute certificate to the address on the blockchain of the owner of the personal information he wishes to obtain. On the user terminal of the public chain that receives the information, the smart contract checks the attribute certificate and determines whether the personal information provision conditions are met, and if it is determined that the personal information may be provided (in S16). YES), reply with the electronic ID of the personal information to be provided and a signature agreeing to the provision of the personal information, and encrypt the common key K1 used to encrypt the personal information to be provided with the public key of the provider and reply ( (See S20).

If there is a reply, it is determined in S48 that there has been a reply from the personal information owner, and the control proceeds to S49, where the returned encryption common key E _PK (K1) is used as the own private key SK. K1 is calculated by performing the decoding operation, that is, D _SK (E _PK (K1)). Next, in S50, a process of transmitting the electronic ID and signature returned from the personal information owner to the server 18 of the authorized business operator 17 is performed. The server 18 of the authorized business operator 17 that receives the information checks the sent signature, searches the personal information DB 29 (see FIG. 2 (B)) based on the received electronic ID, and searches the personal information DB 29 (see FIG. 2 (B)) for the received electronic ID. The encrypted personal information _EK1 (personal information) stored in association with is read out and sent back.

If the reply is received, it is determined as YES in S51 and the control proceeds to S52, where an operation is performed to decrypt the returned encrypted personal information E _K1 (personal information) using K1 calculated in S49. Execute D _K1 (E _K1 (personal information)) to obtain plaintext personal information. The personal information is stored in S53. Thereafter, the process proceeds to S54, where processing is performed to record the personal information provision contract as a transaction on the blockchain.

Next, if a loan contract for PC resources (computing resources) is established with the user terminal of the public chain (YES in S15), YES is determined in S55 and control proceeds to S58, in which the PC resources (computing resources) are The borrowing process is performed. Control then moves to S54, where processing is performed to record the loan contract as a transaction on the blockchain. If an order contract is established with the user terminal of the public chain (YES in S17), it is determined in S56 that the order contract has been established, and the control proceeds to S59, where a process is performed to store the received order for the order. Control then moves to S54, where processing is performed to record the order contract as a transaction on the blockchain.

If a sales contract is established with the user terminal of the public chain (YES in S15), it is determined in S57 that a loan contract for PC resources (computing resources) has been established, and control proceeds to S60 to provide the object for sale. Processing is performed for this purpose. Control then moves to S54, where processing is performed to record the sales contract as a transaction on the blockchain. In addition, in the smart contract processing described above, actions (for example, S42, S47, S58, Instead of executing steps S56 and S60), the action may be executed after obtaining the consent of the owner of the smart contract. This owner's consent may also be obtained when executing a contract using the smart contract described below.

Next, a flowchart of the subroutine program of the machine learning process shown in S31 will be explained based on FIG. 7(B). In S63, a process is performed in which a huge amount of personal information stored in a form in which the owner of the personal information cannot be identified is converted into learning data. There are various learning algorithms used in this machine learning, such as regression and discrimination as supervised learning, model estimation and data mining as unsupervised learning, and reinforcement learning and deep learning as intermediate methods. is available.

Next, in S64, machine learning is performed using learning data using the borrowed PC resources (computing resources). For example, in the case of regression as supervised learning, a large amount of data set consisting of input information (vector x) and correct answer information y is used as training data (learning data). In the case of supervised learning, the function ci(x) (ci:x→y) that maps the input x to the correct answer y is learned, so the learned model includes the function ci. Note that the machine learning performed by the machine learning means 34 is not limited to supervised learning, and may include unsupervised learning such as model estimation and pattern mining (data mining), and intermediate methods between supervised learning and unsupervised learning. It can be anything, such as semi-supervised learning, reinforcement learning, or deep learning.

A general learned model is generated by machine learning in S64, and is stored (S65). This general trained model is a model that uses as training data a huge amount of personal information that is stored in a form that cannot identify the owner of the personal information of many users, and can be widely applied to a large number of users. This is an average trained model.

Next, it is determined in S66 whether or not there is an order received, and if there is not, the process returns, but if there is, it is determined in S67 whether or not the order is received for artificial intelligence. If the order is not for artificial intelligence, the process returns; however, if the order is for artificial intelligence, a process is performed in which personal information is requested from the orderer's address in step S68. If there is a reply with personal information from the orderer, a YES determination is made in S69 and the process proceeds to S70. In S70, a process is performed to personalize the general trained model based on the returned personal information to generate a personalized trained model. The process of personalizing this general trained model to generate a personalized trained model is described in Japanese Patent No. 6432859. The personal information necessary for personalization is collected using blockchain, which has the advantage of avoiding privacy issues as much as possible by ensuring the anonymity of blockchain. Next, in S71, the personalized learned model is sent to the address of the orderer.

Next, a flowchart of the subroutine program of the AI smart contract generation process shown in S32 will be explained based on FIG. 8. In S80, personal information related to a contract (legal act) is extracted from a large amount of personal information stored in a form that does not allow identification of the owner of the personal information, and is converted into learning data. Next, in S81, machine learning is performed using learning data using borrowed PC resources (computing resources). The processes in S80 and S81 are the same as those described in S63 and S64 above, and will not be repeated here.

Next, in S82, a process of generating and storing a trained model of a general AI smart contract is performed. Next, in S83, simulation learning processing is executed. This simulation learning process uses a large group of AI smart contracts to perform legal acts such as contract verification, condition confirmation, enforcement, implementation, and negotiation, which are normally performed by many humans, virtually within a computer. (simulation) and performs reinforcement learning by rewarding each AI smart contract according to its performance. Since reinforcement learning is performed through computer simulation rather than reinforcement learning in the real world, it has the advantage of being able to perform a huge amount of reinforcement learning in a short period of time. Reinforcement learning is a mechanism in which an agent placed in a certain environmental state acquires a policy that maximizes the cumulative reward from the initial state to the goal, based on the reward given when an action is selected. That's true. In reinforcement learning, learning progresses through the interaction between a software agent (hereinafter referred to as an ``agent''), which is a type of AI, and the environment. An agent is a type of AI, and refers to software that behaves autonomously and permanently with a certain degree of decision-making ability while communicating with users and software. When an agent performs a certain action a on the environment, the state s of the environment changes and a reward r is given to the agent when a certain target state is reached. The agent learns a function that inputs the state s and outputs the action a in order to maximize the reward r.

Reinforcement learning progresses over time by repeating the following simple steps.
1. The agent receives the observation o (or directly the state of the environment s) from the environment and returns an action a to the environment based on the policy π.
2 The environment changes to the next state s′ based on the action a received from the agent and the current state s, and based on that transition, the next observation o′ and the goodness or badness of the previous action, called reward r, are calculated. Returns a single number (a scalar quantity) to the agent.
3 Progress of time: t←t+1
Here, ← represents an assignment operation.
For example, an alpha zero type reinforcement learning algorithm may be used as the reinforcement learning. This alpha-zero reinforcement learning algorithm differs from algorithms such as DQN (Deep Q-Network) in that it uses Monte Carlo tree search (MCTS) for searching, and searches for all values and policies. The system uses a neural network to make predictions and corrects the predictions based solely on the experience gained through self-play through tree search. Compared to the conventional AlphaGo, the value network that predicts values and the policy network that predicts policies are integrated into one neural network, and the prediction accuracy is improved through multi-task learning. Additionally, improved performance of neural networks eliminates the need for processor layout processing (extending the search tree until a reward is obtained) during tree search, allowing faster searches. Furthermore, evolutionary computation, genetic algorithms, and generative adversarial networks may be used.

Next, in S84, it is determined whether or not there is an order memory of the AI smart contract, and if there is not, the process returns. If the order is stored in S59 described above, YES is determined in S84, and control proceeds to S85, where it is determined whether the stored order is an order for a simulation-trained AI smart contract. If the order is not for a simulation-trained AI smart contract, the order is for a personalized AI smart contract trained model, and in that case, control proceeds to S86, and personal information related to the contract (legal act) is sent to the orderer's address. Processing to request is performed. When personal information is returned from the orderer, YES is determined in S87, and control proceeds to S88.

In S88, a process is performed to personalize the general AI smart contract trained model based on the returned personal information regarding the contract (legal act) to generate a personalized AI smart contract trained model. The process of personalizing this general AI smart contract trained model to generate a personalized AI smart contract trained model is described in Japanese Patent No. 6432859. The personalized trained model is sent to the orderer's address in S89.

On the other hand, in the case of receiving an order for a simulation-trained AI smart contract, YES is determined in S85, control proceeds to S90, and the simulation-trained AI smart contract is sent to the orderer's address.

Next, a flowchart of the subroutine program of the simulation learning process shown in S83 will be explained based on FIG. 9(A). In S334, it is determined whether or not there is a simulation input, and if there is no input, the process returns. This simulation is inputted by the user terminal of Private Chain 2, and includes information on policies and laws that the government is trying to adopt (for example, reduced tax rates accompanying the consumption tax increase, amended immigration control law, and the European Union (EU) in the UK). ), partial or full adoption of basic income, amendment of Article 9 of the Japanese Constitution, etc.), trading simulations in investment markets such as stock trading and futures trading, and company management simulations. , or consumption behavior simulation. Furthermore, simulations of promotions of new products (including financial products and life insurance) and new services through various media may also be used. If it is determined in S334 that there is a simulation input, control proceeds to S335, and AI smart contract group generation processing is executed.

A flowchart of a subroutine program for this AI smart contract group generation process will be explained based on FIG. 9(B). In S344, a process is performed to set a group of personas that match the input simulation using borrowed PC resources (computing resources). A persona is generally defined as a virtual person who is a typical target for a company, product, or service. In this embodiment, a persona is defined as a typical human image targeted by the simulation content as a virtual person. For example, in the case of a consumption behavior simulation under the reduced tax rate associated with the consumption tax increase mentioned above, personas corresponding to general consumers are used for each group grouped by gender, age, region, annual income, etc. Set as a persona.

The number of personas to be set should be proportional to the number of users belonging to the group. For example, the population distribution of general consumers by age group is 5% in their teens, 5% in their 20s, 10% in their 30s, 10% in their 40s, 20% in their 50s, 20% in their 60s, 20% in their 70s, and 5% in their 80s. , 5% are in their 90s, then the number of personas representing teenagers is 1, the number of personas representing people in their 20s is 1, the number of personas representing people in their 30s is 2, the number of personas representing people in their 40s is 2, There are 4 personas representing people in their 50s, 4 personas representing people in their 60s, 4 personas representing people in their 70s, 1 persona representing people in their 80s, and 1 persona representing people in their 90s. , set.

Next, in S345, a process of selecting a group of users belonging to each persona is performed using the borrowed PC resources (computing resources). Next, in S346, a process is performed to group the users belonging to each persona and collect the transaction data of the users for each group from the blockchain. For example, in the case of a consumption behavior simulation under the reduced tax rate associated with the consumption tax increase mentioned above, users are grouped by gender, age, region, annual income, etc., and the transaction data of the user group is analyzed for each group. Collect from blockchain. For selecting the user group and collecting the transaction data of the user group in S345 and S346, it is useful to use, for example, a database of survey response monitor members held by an Internet survey research company. Internet survey research companies store the contact information (email addresses, etc.) of survey response monitor members in a database in association with attributes such as gender, age group, place of residence, unmarried, married, occupation, annual household income, etc. Use monitor member data by attribute. Similarly, in S145 and S146, S586 and S587, S622 and S623, etc., which will be described later, it is useful to use a database of survey response monitor members held by an Internet survey company. Next, in S347, processing is performed to perform machine learning using the transaction data as learning data using the borrowed PC resources (computing resources) to generate a learned AI smart contract for each persona. This AI smart contract is generated in the same number as the number of corresponding persona settings. This prepares the environment for executing the simulation, and the simulation is performed within that environment.

Returning to FIG. 9A, each AI smart contract generated as described above executes the contract (legal act) according to act a (S336). This "action a" is action a as a result of reinforcement learning in S338. Next, in S337, a process is performed to record the contract established between each AI smart contract on the blockchain.

Next, in S338, a process is performed in which the borrowed PC resources (computing resources) are used to calculate the reward r based on the concluded contract details, and TD learning is performed to obtain the action a according to the optimal policy π ^* . be exposed. For example, in the case of a consumption behavior simulation under the reduced tax rate due to the above-mentioned consumption tax increase, the smaller the value of (expenses before tax increase - expense after tax increase), the higher the reward r is given. Then, it is determined in S339 whether or not the simulation has ended, and if it has not ended yet, the control returns to S336 and repeats S337→S338→S339→S336 to proceed with reinforcement learning. At the stage where the simulation is completed, it is determined as YES in S339, and the control proceeds to S340, where the AI smart contract that obtained the highest reward r is stored, and then the process returns. It should be noted that the storage is not limited to the AI smart contract that obtained the highest reward r, and for example, the top 5% of AI smart contracts may be stored. In addition, it is not necessary to record on the blockchain in S337, and in that case, in the above-mentioned cooperation type, "AI smart contract" in S335, S336, S340, and S347 is changed to "AI for collaboration". do. In other words, in the case of a simulation in a computer, it does not involve the execution of a contract (legal act) in the real world, so if there is no recording on the blockchain, there is no need to use a smart contract, and each collaboration This is because it is sufficient for AIs to perform reinforcement learning by performing action a. When the reinforcement learning is finished and the actual quoting stage occurs, the trained collaboration AI can cooperate with the smart contract to execute the contract and record it on the blockchain.

Next, based on FIG. 10, a flowchart of the subroutine program of the smart contract trust acceptance process shown in S33 will be explained. In S94, it is determined whether or not there is an order memory for the smart contract trust, and if it is determined that there is no order memory, the process returns. The content of the order stored in S59 is checked, and if the order is for a smart contract trust, YES is determined in S94, and control proceeds to S95. In S95, personal information regarding the contract (legal act) is requested from the orderer's address, and when the personal information is returned from the orderer, a YES determination is made in S96 and control proceeds to S97.

In S97, a process is performed to personalize the general AI smart contract trained model based on the returned personal information regarding the contract (legal act) to generate a personalized AI smart contract trained model. The process of personalizing this general AI smart contract trained model to generate a personalized AI smart contract trained model is described in Japanese Patent No. 6432859.

Next, in S98, processing is performed to associate and store the orderer's address and the personalized AI smart contract trained model. Using this stored personalized AI smart contract trained model, a trust contract process is performed for the orderer (S99). Next, in S100, reinforcement learning processing of the personalized AI smart contract trained model is executed.

A flowchart of the subroutine program of the reinforcement learning process of the personalized AI smart contract trained model shown in S100 will be explained based on FIG. 11(A). In reinforcement learning, knowledge of modeling the environment, i.e. state transition probability and reward probability distribution, is used to estimate the Q value when the value of performing an action at in state st is Q(st,at). If the environment model is given, a model-based method may be used, but if the environment model is unknown, TD (Temporal Difference) learning is used. First, since we need to search the environment, we use the ε-greedy method. Introduce the concept of temperature so that at the beginning of the search, you try a variety of actions, and as things settle down, you choose as many optimal actions as possible. Let temperature be T, and choose an action according to the probability expressed by the following formula.

P(a｜s)={exp(Q(s,a)/T)}/{Σexp(Q(s,b)/T} (note that b∈A is written under Σ, and the above formula (The description is omitted here.)
Here, a is the action, Q(s,a) is the value when performing action a in state s,

T is called the temperature during annealing; if it is high, actions are selected with a probability close to equal probability, and if it is low, it is biased toward the optimal one. As learning progresses, decreasing the value of T stabilizes the learning results. Such a Q-value estimation method may be applied to both the reinforcement learning described above and the reinforcement learning described below. Further, although a general Neumann type computer is used as a computer for performing machine learning such as reinforcement learning, a neural net processor (NNP) may also be used. The NNP chip is equipped with many "artificial neurons" modeled after real neurons, and each neuron works together in a network. Alternatively, a quantum computer that employs a "quantum annealing method" may be used. In particular, by using a quantum computer that employs the "quantum annealing method," the time required for optimization calculations in machine learning can be significantly reduced.

In S105, an evaluation of the result of the trust contract processing using the trained model of the personalized AI smart contract is received from the trustor. Next, in S106, a process of calculating the reward r based on the received evaluation is performed. Next, in S107, a process is performed to obtain an action a according to the optimal policy π ^* by TD learning using the borrowed PC resources (computing resources). Next, in S108, the contract according to act a is executed on behalf of the trustor. By receiving an evaluation of the results from the trustor (S105), the processes of S106 to S108 are executed.

Next, a flowchart of a main routine program of a user terminal forming a node of consortium chain 3 will be explained based on FIG. 11(B). IoT sensor data aggregation processing is executed in S114, smart contract processing is executed in S115, simulation processing is executed in S116, and blockchain processing is executed in S117. The specific content of the blockchain processing is as described above based on FIG. 2(A).

Next, a flowchart of the subroutine program of the IoT sensor data aggregation process shown in S114 will be explained based on FIG. 12(A). In S120, a process of classifying and grouping IoT sensor data by type, period, region, etc. is performed. This process includes not only IoT sensor data but also data from wireless sensor networks. Next, in S121, processing is performed to determine the value of each grouped data. According to this determined value, the consideration (amount of tokens) for providing data is coded as a smart contract for each corresponding data. Next, in S122, a process of recording each grouped data on the blockchain is performed.

Next, a flowchart of the smart contract processing subroutine program shown in S115 will be explained based on FIG. 13(B). In S150, it is determined whether a sales contract has been concluded with the person who wishes to obtain the data. This determination is automatically determined to be YES in S150 when the mutual conditions regarding the compensation (amount of tokens) for providing data encoded as a smart contract match. If the determination in S150 is NO, control proceeds to S151, where it is determined whether a PC resource loan contract has been established, and if not, the process returns.

If YES is determined in S150, control proceeds to S152, where data is transmitted to the address of the person who wishes to obtain it, and processing is performed to obtain a token as compensation. The established contract is recorded in the blockchain as a transaction in S153. On the other hand, if a loan contract for PC resources (computing resources) is established, the control proceeds to S154, where a PC resource (computing resource) borrowing process is performed, and the contract is recorded in the blockchain as a transaction at S153.

Next, a flowchart of the subroutine program of the simulation processing shown in S116 will be explained based on FIG. 12(B). This simulation process uses a large number of AI smart contracts to perform legal acts such as contract verification, condition confirmation, execution, execution, and negotiation, which are normally performed by many people, virtually within a computer ( (simulation) and verify what kind of simulation results will be obtained under certain conditions. Specific examples of the above-mentioned "under certain conditions" include policies and laws that the government is trying to adopt (e.g., reduced tax rates associated with a consumption tax increase, revised immigration control laws, the UK's withdrawal from the European Union (EU), basic income). partial or full adoption of Article 9 of the Japanese Constitution, etc.), marketing-related conditions (e.g., setting of prices and consideration for new products (including financial products and life insurance) and new services, promotion through various media) investment market-related conditions (for example, weather conditions in futures trading, monetary tightening policy in the stock market, etc.), etc.

In S134, it is determined whether or not there is a request for simulation, and if there is no request, the process returns. If it is determined in S134 that there is a request for simulation, control proceeds to S135, and AI smart contract group generation processing is executed.

A flowchart of a subroutine program for this AI smart contract group generation process will be explained based on FIG. 13(A). In S144, a process is performed to set a group of personas that match the requested simulation using borrowed PC resources (computing resources). A persona is generally defined as a virtual person who is a typical target for a company, product, or service. In this embodiment, a persona is defined as a typical human image targeted by the simulation content as a virtual person. For example, in the case of simulating the reduced tax rate due to the consumption tax increase mentioned above, personas corresponding to general consumers are set as personas for each group grouped by gender, age, region, annual income, etc. .

The number of personas to be set should be proportional to the number of users belonging to the group. For example, the population distribution of the general public by age is 5% in their teens, 5% in their 20s, 10% in their 30s, 10% in their 40s, 20% in their 50s, 20% in their 60s, 20% in their 70s, 5% in their 80s, If 5% are in their 90s, the number of personas representing teenagers is 1, the number of personas representing people in their 20s is 1, the number of personas representing people in their 30s is 2, the number of personas representing people in their 40s is 2, and the number of personas is 50. There are 4 personas representing people in their 20s, 4 personas representing people in their 60s, 4 personas representing people in their 70s, 1 persona representing people in their 80s, and 1 persona representing people in their 90s. Set.

Next, in S145, a process of selecting a group of users belonging to each persona is performed using the borrowed PC resources (computing resources). Next, in S146, a process is performed in which the user groups belonging to each persona are grouped and transaction data of the user group is collected from the blockchain for each group. For example, in the case of simulating the reduced tax rate due to the consumption tax increase mentioned above, users are grouped by gender, age, region, annual income, etc., and the transaction data of each group is collected from the blockchain. do. Next, in S147, processing is performed to perform machine learning using the transaction data as learning data using the borrowed PC resources (computing resources) to generate a learned AI smart contract for each persona. This AI smart contract is generated in the same number as the number of corresponding persona settings. This prepares the environment for executing the simulation, and the simulation is performed within that environment.

Returning to FIG. 12(B), each AI smart contract generated as described above executes the contract (legal act) according to act a (S136). This "action a" is action a as a result of reinforcement learning in S138. Next, in S137, a process is performed to record the contract established between each AI smart contract on the blockchain. Additionally, changes in the situation as the simulation progresses are recorded on the blockchain. For example, in the case of a simulation of the reduced tax rate due to the consumption tax increase mentioned above, how domestic demand and the economy have changed as the simulation progresses is recorded on the blockchain.

Next, in S138, a process is performed in which the borrowed PC resources (computing resources) are used to calculate the reward r based on the concluded contract details, and TD learning is performed to obtain the action a according to the optimal policy π ^* . be exposed. Then, it is determined in S139 whether or not the simulation has ended, and if it has not ended yet, the control returns to S136 and repeats S137→S138→S139→S136 to advance reinforcement learning. When the simulation is completed, a YES determination is made in S139, and the control proceeds to S140, where processing for deriving the simulation results is performed, and then the process returns.

A specific example of processing to derive simulation results is, for example, in the case of simulating economic fluctuations due to the adoption of a reduced tax rate due to a consumption tax increase, it is possible to calculate how each item of the economic trend index has changed as a result of the simulation. Derive. In the case of a simulation associated with a financial tightening policy in the stock market, how the stock market fluctuated as a result of the simulation is derived. In addition, we will vary the specific aspects of the reduced tax rate associated with the consumption tax increase (for example, what items are eligible for the reduced tax rate, the reduced tax rate for each item eligible for the reduced tax rate, etc.) to determine the optimal reduced tax rate form. A simulation optimization method may be used to determine the . In this case, as the optimal form of the reduced tax rate, the expected value E is (tax revenue increase rate (%) + diffusion index (DI) as an economic trend index / 50), and the simulation result that maximizes the expected value E is calculated. Try to get it. Let θ be the control parameter (aspect of reduced tax rate) in the simulation, let Y(θ) be the simulation result, and find θ at maxE[Y(θ)]. Specific methods for optimizing such simulations include, for example, Particle Swarm Optimization (PSO) as a metaheuristic algorithm, cases where it is difficult to express the objective function analytically, and information on the differentiation of the objective function. DFO (derivative free optimization) is used to find the optimal solution in situations where it is not possible to use . Note that it is not necessary to record on the blockchain in S137, and in that case, in the above-mentioned cooperation type, the "AI smart contract" in S135, S136, and S147 is changed to "AI for collaboration." In other words, in the case of a simulation in a computer, it does not involve the execution of a contract (legal act) in the real world, so if there is no recording on the blockchain, there is no need to use a smart contract, and each collaboration This is because it is sufficient for AIs to perform reinforcement learning by performing action a.
[Modified example]

(1) The certified business operator 17 stores the user's encrypted personal information E _K1 (personal information), but instead records the encrypted personal information E _K1 (personal information) directly on the blockchain. You may also do so.

(2) Transaction data other than personal information, such as transaction C and transaction F in FIG. 2, may also be encrypted with the key K1 or the like and recorded on the blockchain in the same way as personal information.

(3) In the above explanation, the operation processing of the node 19 of each blockchain network 2, 3, and 4 was shown, but the operation processing shown for the node 19 of the private chain 2 is also applied to the node 19 of the other blockchain networks 3 and 4. The operation process shown for node 19 of consortium chain 3 may be performed by nodes 19 of other blockchain networks 2 and 4, and the operation process shown for node 19 of public chain 4 may be performed by other blocks. It may also be performed at the nodes 19 of the chain networks 2 and 3. This modification may be similarly applied to the embodiments described below.

(4) Borrowed PC resources (computing resources) may be used to perform mining (competition for bookkeeping rights) in the blockchain. In this case, PC resources (computing resources) may be lent at an hourly rate as in the first embodiment, but what is the profit (tokens, etc.) that miners who succeed in mining (competition for bookkeeping rights) receive? The discount may be distributed (divided) to the lender of PC resources (computing resources). The dividend ratio (dividend amount) is controlled to be proportional to the loan amount of PC resources (computing resources) (number of PCs loaned x loan time, etc.).

(5) It is also possible to carry out some project by borrowing (using) the above-mentioned PC resources (computing resources), privately generated electric power, and the like. Specific examples of projects include research and development (e.g., artificial intelligence development, machine learning, human genome analysis, new product development, new drug development, etc.), exploration and excavation of rare metals, oil, natural gas, marine resources, etc., space development, etc. However, it is possible. In this case, the lender (provider) may obtain consideration (tokens, etc.) according to the loan amount (provided amount) of the loan (provided) target as in the first embodiment, but the success of the project A percentage of the profits obtained by the project executor (individual, corporation, or group) may be distributed (divided) to the resource lender (resource provider). The dividend ratio (dividend amount) is controlled to be proportional to the loan amount (provided amount) of resources.

Furthermore, the resource loaner (resource provider) may acquire the right to receive the dividend (hereinafter referred to as "dividend enjoyment right") instead of receiving the dividend itself. This dividend enjoyment right may be controlled so that the lender (provider) acquires it in the form of a token issued by the project executor, for example. Then, the lender (provider) may control the acquired dividend enjoyment rights (tokens) so that they can be transferred to others at a price (tokens) according to the market price at that time. With this configuration, the dividend enjoyment rights (tokens) can be managed as if they were traded in a secondary market in the stock market.

(6) In S71, the generated personalized trained model is sent to the orderer's address and delivered, but in addition to or in place of that, orders can be placed using the generated personalized trained model. It may also be possible to provide personalized services to users.

(7) In the above explanation, smart contracts automate contract verification, condition confirmation, execution, implementation, and negotiation, but it is necessary to obtain the user's consent before concluding and executing a contract (legal act such as a transaction). It may be controlled as desired. In addition, instead of requesting the user's consent for all contracts (legal acts such as transactions), the system determines whether or not it is a predetermined important contract (legal acts such as transactions), and control may be performed so that the user's consent is requested if it is determined that the act is a legal act). Furthermore, it is determined whether the contract is one that requires urgent conclusion and execution (legal acts such as transactions), and if it is determined that the contract is urgent (legal acts such as transactions), the user It is also possible to perform control to conclude and execute a contract (legal act such as a transaction) without obtaining consent, and control to report it to the user himself/herself later. This modification may be similarly applied to the embodiments described below.

(8) The above-mentioned programs that operate on the user terminals 16, etc. and various servers that constitute the nodes 19 of each blockchain may be downloaded and installed from a predetermined website, etc.; The program may be recorded on a medium (non-transitory recording medium) and distributed, and the person who purchases the CD-ROM 99 or the like may install the program on the user terminal 16 and various servers (see Figure 60). .

(9) In the above description, a centralized oracle 21 is used, but a distributed oracle that is managed in a distributed manner throughout the network may also be used. The information collected by multiple oracles distributed throughout the network is collected, the average information is extracted, and the average information is regarded as correct information and is incorporated into the blockchain and used for smart contracts. This is based on the idea proposed by James Sulowiecki in his book ``Everyone's Right'': ``By gathering information in a group, the group comes to a better conclusion than any individual in the group could think of.'' It is based on the theory that it is possible to Then, by giving rewards such as tokens to oracles that collect information close to average information, they are given an incentive to operate decentralized oracles.

In summary, there is an extraction means that collects information collected by multiple oracles distributed on a network and extracts average information, a recruitment means that employs the average information extracted by the extraction means, and an oracle. the plurality of oracles include a first oracle and a second oracle, and the reward granting means provides information closer to the average information than the first oracle. More rewards are given to the collected second oracles. Note that the reward given to the first oracle may be 0 or may be negative.

(10) Regarding the determination of the establishment of the K2 distribution contract in S13 and/or the establishment of the personal information provision contract in S16, information that specifies the intention of the personal information owner (hereinafter referred to as "intention identification information") is required. ) may be provided. Specifically, when the owner of personal information receives recommendations for products or services that match him or her at a physical store or electronic shopping mall, etc., the owner of the personal information may provide specific information on his/her own mobile device (smartphone, IC card, etc.). By reading the information and entering a PIN that indicates that the contract can be concluded, the personal information owner's smart contract is notified of the intention-specific information consisting of the specific information and the PIN, and the smart contract is executed according to the intention-specific information. will judge.

By doing so, the user can utilize the personal information retrieved from the SNS etc. operator and placed under self-management according to his or her own will.
[Second embodiment]

Next, a second embodiment will be described. This second embodiment addresses the need for the right to delete personal information (the so-called right to be forgotten) by applying encryption technology to personal information recorded using blockchain. It is something to respond to. Blockchain is characterized by the fact that once recorded information cannot be tampered with or is extremely difficult to tamper with, and for this reason it is impossible or extremely difficult to delete information once recorded (hereinafter referred to as "impossibility of deletion"). ). On the other hand, the General Data Protection Regulation (GDPR) in Europe requires that the right to delete personal information (the so-called right to be forgotten), which allows the owner of personal information to delete recorded personal information, is guaranteed. This GDPR requirement for the right to delete personal information and the impossibility of deletion in blockchain are in direct conflict, creating a contradictory dilemma. In other words, the second embodiment solves the dilemma in which the requirement to guarantee the right to delete information to be deleted and the impossibility of deletion are contradictory. The outline will be explained based on FIG. 14.

FIG. 14(A) shows a normal state in which the deletion right is not exercised, and FIG. 14(B) shows a state in which the deletion right is exercised and the information is rendered indecipherable. Referring to FIG. 14A, an information holder (also referred to as an information owner) 40 double-encrypts information using two common keys KA and KB. Expressed as a formula, it becomes E _KA (E _KB (information)). Next, the encrypted information E _KA (E _KB (information)) is recorded on a blockchain or the like. Note that the common key KA is stored in a confidential state in the user terminal of the information holder 40, etc.

In this state, when the information requester 41 requests information from the information owner 40, the information owner 40 decrypts the recorded encrypted information E _KA (E _KB (information)) using the key KA. Expressed as a formula, D _KA (E _KA (E _KB (information)) = E _KB (information). Then, the information owner 40 sends this E _KB (information) and the common key KB to the information requester 41. Send.

The information requester 41 who has received them decrypts the received _EKB (information) using the received half common key KB. Expressed as a formula, D _KB (E _KB (information))=information. Thereby, the information requester 41 can obtain plaintext information.

Next, a state in which information is made unreadable by exercising the deletion right will be explained based on FIG. 14(B). The information owner 40 updates one of the half keys KA and KB used to encrypt the information to be made indecipherable to a random number R (≠KA). Next, when the information requester 41 who has already stored the shared common key KB requests information from the information owner 40, the information owner 40 uses the stored encrypted information E _KA (E _KB (information )) with key R (random number). Expressed as a formula, it becomes D _R (E _KA (E _KB (information)). Then, the information owner 40 transmits this D _R (E _KA (E _KB (information)) to the information requester 41 .

The information requester 41 who received it decrypts D _R (E _KA (E _KB (information)) using the already stored half common key KB. Expressed in the formula, D _KB (D _R (E _KA ( E _KB (information))))≠information.In this way, in the unreadable state, even the information requester 41 who has already memorized the half common key KB cannot obtain the plaintext information. , it is possible to resolve the dilemma where the request to guarantee the right to delete information that you want to delete and the impossibility of deletion are contradictory.In addition, copying is prohibited so that information such as personal information recorded on blockchain etc. cannot be copied and pasted. If processing is carried out, the guarantee of the right to delete information will be more complete.Note that there is no need to limit to double encryption using two keys KA and KB; You may also use multiple encryption using a random number R. In this case, by replacing at least one of the n keys with a random number R, the state becomes unreadable.

The outline of the second embodiment described above will be explained in more detail. Regarding common points with the first embodiment, repeated explanation will be omitted, and differences will be mainly explained. FIG. 15 corresponds to FIG. 2 in the first embodiment. Referring to transaction I in the blockchain, in this second embodiment, encrypted personal information E _KA (E _KB (personal information)) is directly recorded in the block. Therefore, the certified business operator 17 is not necessary in the second embodiment. Here, KA and KB are half common keys.

Next, with reference to FIG. 16, a flowchart of the main routine of the user terminal configuring the node 19 of the private chain 2 and the user terminal 16 configuring the node of the public chain 4 will be described. This main routine omits the flowchart of the operation process shown in the first embodiment, and only shows a flowchart of the operation process added to or modified from the operation process shown in the first embodiment. In the user terminal 16 constituting a node of the public chain 4, personal information recording processing is performed on the blockchain in S160, processing to make the record undecipherable is performed in S161, and personal information provision processing is performed in S162. At the user terminal forming the node 19 of the private chain 2, personal information acquisition processing is performed in S170.

The process of recording personal information on the blockchain is the process of recording personal information on the blockchain. Processing to make records unreadable is a process for making information unreadable by exercising the deletion right. The personal information provision process is a process in which the user terminal 16 of the public chain 4 provides personal information to the user terminal of the private chain 2. The personal information acquisition process is a process in which the user terminal of the private chain 2 acquires personal information from the user terminal 16 of the public chain.

A flowchart of a subroutine program for personal information recording processing on a blockchain will be described based on FIG. 17(A). In S174, a process of generating two random numbers is performed. For example, in the case of DES, two 56-bit random numbers are generated and these 56-bit random numbers are used as the half common keys KA and KB. In the case of ADS, two 128-bit random numbers are generated and these 128-bit random numbers are used as half common keys KA and KB.

Next, in S177, processing is performed to record E _KA (E _KB (personal information)), E _K2 (index + compensation for providing personal information), and a ciphertext identifier on the blockchain. This ciphertext identifier is an identifier for specifying E _KA (E _KB (personal information)), which is encrypted personal information, and corresponds to the electronic ID in the first embodiment.

Next, in S183, the KA and KB and the ciphertext identifier are stored in association with each other in the HDD 12 of the user terminal 16.

Next, a flowchart of a subroutine program for recording unreadable processing will be explained based on FIG. 17(B). In S190, it is determined whether or not there is a ciphertext that is desired to be made unreadable among the ciphertexts (for example, _EKA ( _EKB (personal information)) recorded on the blockchain. If there is no ciphertext, the process returns; however, if there is a ciphertext to be made unreadable, control proceeds to S191, where the process searches the HDD 12 for the half common key KA stored in association with the ciphertext identifier of the ciphertext. will be held.

Next, in S192, a random number R is generated. For example, in the case of DES, a 56-bit random number is generated. In the case of ADS, a 128-bit random number is generated. Next, in S193, it is determined whether the generated random number R=KA. If the generated random number R is the same as the half common key KA stored in the HDD 12, control returns to S192 and a random number is generated again. If NO is determined in S193, the control proceeds to S194, and a process of updating the common key KA stored in the HDD 12 to R is performed.

Next, a flowchart of a subroutine program for personal information provision processing and personal information acquisition processing will be described based on FIG. 18. In the user terminal of the private chain 2, it is determined in S198 whether or not the common key KB of the desired personal information is already stored. If the common key KB of the other half of the personal information you wish to obtain has already been distributed from the user terminal 16 of public chain 4 to the user terminal of private chain 2, it is determined in S198 that there is storage, and the control proceeds to S203, but it is still not stored. If not, control proceeds to S199.

In S199, a process of transmitting the ciphertext identifier of the personal information desired to be obtained to the user terminal 16 of the public chain 4 and requesting the shared common key KB is performed. The user terminal 16 of the public chain 4 that receives the information in S200 uses the smart contract to determine whether or not to conduct a transaction that provides the personal information specified by the ciphertext identifier (see S16), and determines whether or not to conduct a transaction that provides the personal information specified by the ciphertext identifier (see S16). If this is done, a signature agreeing to provide personal information and a half common key KB corresponding to the ciphertext identifier are returned in step S201.

The user terminal of the private chain 2 that received it in S202 transmits the signature and the ciphertext identifier of the desired personal information to the user terminal 16 of the public chain 4 in S203. The user terminal 16 of the public chain 4 that receives the information in S206 uses the smart contract to determine whether or not to conduct a transaction that provides the personal information specified by the ciphertext identifier (see S16), and determines whether or not to conduct a transaction that provides the personal information specified by the ciphertext identifier (see S16). If this is the case, in step S207, a process is performed in which D _KA (encrypted personal information) or _DR (encrypted personal information) is calculated and sent back. Specifically, if the common key KA stored in the HDD 12 of the user terminal 16 has already been updated to a random number R, D _R (encrypted personal information) is calculated and returned, but the random number is still If it has not been updated to R, processing is performed to calculate D _KA (encrypted personal information) and send it back.

The user terminal of the private chain 2, which received the reply from the user terminal 16 of the public chain 4 in S208, determines in S209 that D _KB (D _KA (encrypted personal information)) = plain text or D _KB (D _R (encrypted personal information)) Personal information)) ≠ Plaintext processing is performed. Specifically, when D _KA (encrypted personal information) is received, D _KB (D _KA (encrypted personal information)) = D _KB (D _KA (E _KA (E _KB (personal information))) )=plaintext to obtain the personal information in plaintext. On the other hand, if (D _R (encrypted personal information)) is received, D _KB (D _R (encrypted personal information)) = D _KB (D _R (E _KA (E _KB (personal information)))) ≠Plaintext is calculated, and personal information in plaintext cannot be obtained. This makes it possible to resolve the dilemma in which the request to guarantee the right to delete information to be deleted and the impossibility of deletion are contradictory.
[Modified example]

(1) In the above explanation, encrypted personal information E _KA (E _KB (personal information)) is directly recorded on the blockchain, and this large amount of encrypted personal information is stored at each node (in public chain 4, all If the data is stored in multiple computers, each node (user terminal) is required to have a huge amount of storage capacity, which is an inconvenience.As a means to solve this problem, secret sharing technology is used to store divided data on multiple computers. Divide the data into fragments and store each fragmented data in multiple nodes.Also, make the data stored in each node redundant (duplicate) and store it.Ensure sufficient redundancy. By doing so, even if a part of the fragmented data is lost, there will be no problem in restoring it, and it will also ensure that the blockchain is difficult to tamper with.Furthermore, the amount of memory responsible for storing the fragmented data can be reduced by each node. It may be controlled so that it is decided by the will, and control is given to each node in the form of a token or the like in accordance with the amount of memory it carries.

(2) In the above explanation, the encrypted personal information E _KA (E _KB (personal information)) was directly recorded on the blockchain, but in the modified examples shown in Figures 19 to 23, the encrypted personal information E _KA (E _KB (personal information)) is stored in the personal information DB 29 of the certified business operator 17, and the hash value of the encrypted personal information is recorded on the blockchain. Transaction I shown in Figure 19 (A) The hash value of E _KA (E _KB (personal information)) + E _K2 (provided by index + 2.4 tokens) + ciphertext identifier and electronic signature are recorded. Also, as shown in Figure 19 (B) The personal information DB 29 of the authorized business operator 17 stores encrypted personal information E _KA (E _KB (personal information)) in association with the ciphertext identifier.

Based on FIG. 20, the main connections between the user terminal configuring the node 19 of the private chain 2, the user terminal 16 configuring the node 19 of the public chain 4, and the server 18 of the certified business operator 17 in this modification example Describe the flowchart of the routine. The server 18 of the authorized business operator 17 participates in the blockchain as a node 19. S215 executes personal information recording processing on the blockchain, S216 executes record unreadable processing, S217 executes personal information provision processing, S220 executes hash value recording processing, and S221 encrypted text transmission processing. is executed, and personal information acquisition processing is executed in S224.

In the process of recording personal information on the blockchain, the user terminal 16 constituting the node 19 of the public chain 4 sends encrypted personal information E _KA (E _KB (personal information)) to the server 18 of the certified business operator 17. This is the process of sending. The hash value recording process is a process in which the server 18 of the certified business operator 17 that receives encrypted personal information E _KA (E _KB (personal information)), etc. stores it, generates its hash value, and records it on the blockchain. It is. Record unreadable processing is a process for making encrypted personal information E _KA (E _KB (personal information)), etc., which is desired to be made unreadable, unreadable. The personal information provision process is a process executed by the user terminal 16 forming the node 19 of the public chain 4 to provide personal information to the user terminal forming the node 19 of the private chain 2. The personal information acquisition process is a process in which the user terminals forming the nodes 19 of the private chain 2 acquire personal information. In the ciphertext transmission process, the server 18 of the certified business operator 17 transmits ciphertext such as encrypted personal information E _KA (E _KB (personal information)) to the user terminal forming the node 19 of the private chain 2. It is processing.

The details of each process will be explained below based on the flowchart of each subroutine program, and differences from the second embodiment will be mainly explained.

A flowchart of a subroutine program for personal information recording processing and hash value recording processing on the blockchain will be explained based on FIG. 21. In S231, E _KA (E _KB (personal information)) and E _K2 (index + compensation for providing personal information) are transmitted to the server 18 of the certified business operator 17 . The server 18 that has received it in S240 performs processing to generate a hash value of E _KA (E _KB (personal information)) and a ciphertext identifier in S241. Next, in S242, a process is performed to record the hash value of E _KA (E _KB (personal information)), E _K2 (index + compensation for providing personal information), and a ciphertext identifier on the blockchain.

Next, in S243, a process of transmitting the ciphertext identifier to the user terminal 16 of the public chain 4 is performed. The user terminal 16 of the public chain 4 that received the key in S232 performs a process of storing the half keys KA and KB and the received ciphertext identifier in association with each other in the HDD 12 in S233. In S244, the server 18 of the authorized business operator 17 performs a process of associating E _KA (E _KB (personal information)) with a ciphertext identifier and storing them in the personal information DB 29 .

The recording undeciphering process shown in FIG. 22 is the same as that already explained in FIG. 17(B) of the second embodiment, and therefore, a repeated explanation will be omitted.

Next, a flowchart of a subroutine program for personal information providing processing, personal information obtaining processing, and ciphertext transmission processing will be described based on FIG. 23. In this modified example, when the user terminal of the private chain 2 receives the signature agreeing to provide personal information and the common key KB corresponding to the ciphertext identifier from the user terminal 16 of the public chain 4 (S264), the user terminal of the private chain 2 performs S265. , transmits the received signature and the ciphertext identifier of the personal information desired to be obtained to the server 18 of the authorized business operator 17. The server 18 of the authorized business operator 17 that received the message in S266 checks the signature, searches the personal information DB 29 for the ciphertext corresponding to the ciphertext identifier, and sends it to the user terminal 16 of the public chain 4 in S267. Processing takes place.

The user terminal 16 of the public chain 4 that received it in S268 calculates D _KA (encrypted personal information) or D _R (encrypted personal information) and sends it back to the user terminal of the private chain 2 in S269. will be held. Specifically, if the common key KA stored in the HDD 12 of the user terminal 16 has already been updated to R, _DR (encrypted personal information) is calculated and returned, but if it is not yet updated to R If it has not been updated, processing is performed to calculate D _KA (encrypted personal information) and send it back.

The user terminal of the private chain 2, which received the reply from the user terminal 16 of the public chain 4 in S270, determines in S271 that D _KB (D _KA (encrypted personal information)) = plain text or D _KB (D _R (encrypted Personal information)) ≠ Plaintext processing is performed. Specifically, when D _KA (encrypted personal information) is received, D _KB (D _KA (encrypted personal information)) = D _KB (D _KA (E _KA (E _KB (personal information))) )=plaintext to obtain the personal information in plaintext. On the other hand, if (D _R (encrypted personal information)) is received, D _KB (D _R (encrypted personal information)) = D _KB (D _R (E _KA (E _KB (personal information)))) ≠Plaintext is calculated, and personal information in plaintext cannot be obtained. This makes it possible to resolve the dilemma in which the request to guarantee the right to delete information to be deleted and the impossibility of deletion are contradictory.

Note that the server 18 of the certified business operator 17 may be connected to the user terminal of the private chain 2 and the user terminal of the pub chain 4 via the Internet 1 without participating in the blockchain as a node 19. .

(3) In the above explanation, the personal information owner held the shared key KA (stored in the HDD 12 of the user terminal 16), but instead, the key registration center 30, which is an example of a designated organization (third party organization) The common key KA may be registered in the key DB 32 of . Note that the common key KA is stored in the key DB 32 in a secret state. This modification will be explained based on FIGS. 24 to 27.

Referring to FIG. 24, a server 31 of a key registration center 30 is connected to the Internet 1. A key DB 32 connected to the server 31 stores a ciphertext identifier and a half common key KA in association with each other for each address of a user who is each node 19 of the public chain 4. If there is a request from the user to make the record undecipherable, the shared common key KA stored in association with the ciphertext identifier corresponding to the requested record is updated to a random number R. In FIG. 24, the common key half stored in association with the ciphertext identifier 307cd4 at address 0x6079dd has been updated to the random number 1R2, and the common key half stored in association with the ciphertext identifier 4arb56 at address 0x6080dd has been updated to random number 1R2. It has been updated to a random number 2Rn, and the common key half stored in association with the ciphertext identifier e2c87r at address 0x6978dd has been updated to a random number mR1.

Next, a flowchart of the main routine of the user terminal 16 of the public chain 4, the server 31 of the key registration center 30, and the user terminal of the private chain 2 will be explained based on FIG. Points in common with the second embodiment will not be repeated, and differences will be mainly explained.

In the user terminal 16 of the public chain 4, a process for recording personal information on the blockchain is executed in S468, a process for requesting to make the record unreadable is executed in S469, and a process for providing a decryption key is executed in S470. In the server 31 of the key registration center 30, a key registration process is executed in S463, a record undecipherable process is executed in S464, and a data decryption process is executed in S465. At the user terminal of private chain 2, data acquisition processing is executed in S460.

Next, a flowchart of a subroutine program for personal information recording processing and key registration processing on the blockchain will be described based on FIG. 26(A). In the user terminal 16 of the public chain 4, E _KA (E _KB (personal information)), E _K2 (index + compensation for providing personal information), and a ciphertext identifier are recorded on the blockchain in S479, and one half is registered in S480. A process of transmitting the common key KA and the ciphertext identifier to the key registration center 30 is performed.

The server 31 of the key registration center 30 that received it in S474 performs a process of storing the received half common key KA and the ciphertext identifier in the key DB 32 in association with each other in S475.

Next, a flowchart of a subroutine program for the record unreadable request process and the record unreadable process will be described based on FIG. 26(B). In the user terminal 16 of the public chain 4, it is determined in S494 whether or not there is a ciphertext to be made unreadable, and if there is no ciphertext, the return is made; A process of transmitting the information to the server 31 of the key registration center 30 is performed.

The server 31 of the key registration center 30 that received it in S485 performs a process of searching the key DB 32 for the half common key KA stored in association with the received ciphertext identifier. Next, in S487, a random number R is generated, and in S488, it is determined whether the random number R=KA. If R=KA, the random number R is generated again in S487, and when R≠KA, the process of updating the common key KA to R is performed in S489.

Next, with reference to FIG. 27, a flowchart of a subroutine program of a common key provision process, a data acquisition process, and a data decryption process will be described. In the user terminal of the private chain 2, a process of transmitting the ciphertext identifier of the desired data to the server 31 of the key registration center 30 is performed in S503. In S505, the server 31 of the key registration center 30 that has received it in S504 performs a process of searching the block chain for the ciphertext (encrypted personal information, etc.) corresponding to the ciphertext identifier. Next, in S506, processing is performed to search for the half common key KA or R corresponding to the ciphertext identifier.

Next, in S507, processing is performed to return D _KA (encrypted personal information) or _DR (encrypted personal information) to the user terminal of private chain 2. Specifically, if the half common key KA stored in the key DB 32 of the key registration center 30 has already been updated to R, D _R (encrypted personal information) is calculated and returned, but the If it has not been updated to R, processing is performed to calculate D _KA (encrypted personal information) and send it back.

The user terminal of private chain 2 that received it in S509 calculates D _KA (D _KB (encrypted personal information)) = plain text or D _KA (D _R (encrypted personal information)) ≠ plain text in S510. Processing takes place. Specifically, when D _KA (encrypted personal information) is received, D _KB (D _KA (encrypted personal information)) = D _KB (D _KA (E _KA (E _KB (personal information))) )=plaintext to obtain the personal information in plaintext. On the other hand, if (D _R (encrypted personal information)) is received, D _KB (D _R (encrypted personal information)) = D _KB (D _R (E _KA (E _KB (personal information)))) ≠Plaintext is calculated, and personal information in plaintext cannot be obtained. This makes it possible to resolve the dilemma in which the request to guarantee the right to delete information to be deleted and the impossibility of deletion are contradictory. Moreover, since the process of updating the common key KA to R is performed at the key registration center 30, it is easy to ensure the reliability of updating KA to R and guaranteeing deletion rights. For example, there is an advantage that updating KA to R can be easily performed under audit by a predetermined organization.

(4) Another way to solve the dilemma of the conflict between the request to guarantee the right to delete information and the impossibility of deletion is to use encrypted personal information stored in the personal information DB 29 of the certified business operator 17. E _KA (E _KB (personal information)) may be deleted at the request of the owner of the personal information. In that case, a contradictory situation will occur in which the hash value of personal information is recorded on the blockchain but the corresponding personal information is not stored in the personal information DB 29. However, it is necessary to tolerate this contradiction. If possible, deleting personal information is also an effective method.

(5) Information that guarantees the right to delete is not limited to personal information, but also includes information posted on SNS and blogs (including data on posted photos and videos), notarized documents such as wills and voluntary guardianship contracts, and private documents. It can be any kind of information, such as the articles of incorporation of a company, etc., or other information that requires a fixed date. Further, in the second embodiment, the information that guarantees deletion rights is recorded using a blockchain, but the blockchain is only an example, and other information may be used for recording.

(6) The aforementioned programs that operate on the user terminals 16, etc. and various servers that constitute the nodes 19 of each blockchain may be downloaded and installed from a predetermined website, etc.; The program may be recorded on a medium (non-transitory recording medium) and distributed, and the person who purchases the CD-ROM 99 or the like may install the program on the user terminal 16 and various servers (see Figure 60). .

(7) In the above explanation, the data is encrypted twice by encrypting it once with the common key KA, then encrypted again with the common key KB, and then decrypted once with the common key KB. is decrypted twice using the common key KA, thereby converting it into plain text. However, the invention is not limited to this, and encryption may be performed multiple times using the shared key KA or KB, and decryption may be performed multiple times using the shared key KA or KB. Further, the number of common keys KA and KB is not limited to two, and three or more common keys may be used.

Furthermore, one key K is generated by calculating the exclusive OR of the common keys KA and KB (KA(+)KB=K), and personal information is encrypted with that key K. (E _K (personal information)), one half of the common key KA is registered in the key DB 32 of the key registration center 30, and one half of the common key KB is distributed to the person requesting the personal information. Upon receiving a request from a personal information requester, the personal information owner sends the encrypted personal information to the server 31 of the key registration center 30, and the personal information requester sends the distributed common key KB to the key registration center 30. Send to server 31. The server 31 of the key registration center 30 calculates the exclusive OR of the received half common key KB and the half common key KA registered in the key DB 32 to generate one key K. (KA(+)KB=K), the received encrypted personal information (E _K (personal information)) is decrypted using the key K to make it plaintext, and then (D _K (E _K (personal information)) = plaintext. ), the plaintext personal information may be sent to the personal information requester. The above (+) is a symbolic representation of exclusive OR.

Note that the exclusive OR is just one example, and any algorithm may be used as long as it generates one key K from the common keys KA and KB.

In addition, in the case of the above method in which the key K is generated using an additive group such as exclusive OR, the advantage is that security can be maintained by periodically updating the common keys KA and KB. There is. For example, when one half of the common key KA is updated to KC, the other half of the common key KB=K(+)KC, which can be determined by calculation. By updating the shared common keys KA and KB in this way, it is not only possible to prevent the leakage of the shared common keys, but also to ensure that the personal information requester who once distributed the shared common key KB can access the blockchain again. It becomes possible to prevent encrypted personal information from being viewed by making it impossible to decrypt it. By updating the shared common key in this way every time the shared common key KB is distributed, even if the person who received the distributed common key KB misappropriates the shared common key KB to another person, the It becomes possible to make encrypted personal information on the blockchain impossible to decrypt by a person who has been misappropriated. In other words, the distributed common key KB can be made into a one-time key that can be used only once.

Further, the encryption key is not limited to the common key, and a public key cryptosystem such as RSA or elliptic cryptography may be used.

Further, in realizing the above-mentioned key update, a cryptographic algorithm that satisfies the following conditions may be adopted.
Denote the plaintext as M, the ciphertext as C, and the encryption keys as KA, KB, KC, and KD,
E _KA (E _KB (M)) = E _KC (E _KD (M)) = C
An algorithm that satisfies the formula.

If such an algorithm is a common key encryption algorithm, when the common keys KA and KB are updated to KC and KD, the ciphertext C recorded in the blockchain is decrypted using the common keys KC and KD. By doing this, plaintext M can be obtained. On the other hand, in the case of a public key cryptographic algorithm, when the private keys KA and KB are updated to KC and KD, the cryptography recorded in the blockchain with the pair of public keys PKC and PKD corresponding to the private keys KC and KD. By decoding sentence C, plaintext M can be obtained.
(8) In the above explanation, the information holder transmits the encrypted personal information (D _KA (encrypted personal information) or D _R (encrypted personal information)) and the common key KB to the information requester (S201). , S207), the information requester himself used the shared key KB to decrypt the encrypted personal information to plain text (S209), but the decryption using the shared key KB was performed by a third party (a designated service organization) may also be responsible. In this case, the information holder sends the encrypted personal information (D _KA (encrypted personal information) or D _R (encrypted personal information)) and the common key KB to a third party organization (prescribed service organization). , decrypted by a third-party organization (prescribed service organization) and sent to the information requester.
[Characteristics of disclosure content]

Next, the features of the disclosed contents of the embodiments described above are listed below.
(Feature 1)
[Technical field]

Feature 1 relates to a processing system and program for an information recording method that is difficult to tamper with or erase, such as a blockchain.
[Background technology]

Blockchain has been generally known as an information recording method that is difficult to tamper with or erase. For example, Japanese Patent Application Laid-Open No. 2018-128723 records various information related to cargo transportation using this blockchain.
[Summary of feature 1]
[Problem that Feature 1 attempts to solve]

However, information records made using such blockchains are not only difficult to falsify, but also difficult to erase (hereinafter referred to as "irresolvability"). As a result, once personal information is recorded using blockchain, even if the owner of the personal information wants to erase it, it cannot be erased, and the right to erase personal information (the so-called right to be forgotten) is impaired. There is a drawback.

In other words, a dilemma arises in which guaranteeing the authenticity of recorded information and guaranteeing the right to delete that information are contradictory.

Feature 1 was devised in view of the above circumstances, and its purpose is to resolve the dilemma in which guaranteeing the authenticity of recorded information and guaranteeing the right to delete that information are contradictory.
[Means to solve the problem]

The subject matter of feature 1 is shown as the following items, for example.
(Item 1)
An encryption means (for example, S174, S177, or S228, S231, or S478, S479) that performs an encryption process to encode information to be recorded (for example, personal information);
A recording means for recording information after the encryption process (for example, S177 and blockchain, or S231, S240, S242, S244, blockchain and personal information DB 29, or S479 and blockchain);
A decryption means (for example, S201, S202, S207 to S209, or S263 to S271) performs a decryption process on the information recorded by the recording means using a first key and a second key to convert it into plaintext information. , or S500 to S510),
comprising a decoding disabling means (for example, S191 to S194, or S250 to S254, or S494, S495, S485 to S489) that puts the information recorded by the recording means into a decoding disabling state in which the information recorded by the recording means cannot be decoded;
The decryption means includes a second key secrecy holding means (for example, S194, S233, or S475) that keeps the second key (for example, half common key KA) confidential;
The decryption disabling means updates the second key held by the second key secrecy holding means to another key (for example, random number R) to disable decryption (for example, from S190 to S194). , or S250 to S254, or S494, S495, S485 to S489), a processing system.

(Item 2)
The decryption means is a first key distribution means (for example, S200, S201, or S2562, S263, or S500, S501) that distributes the first key (for example, half common key KB) to those who wish to view the information. The processing system according to item 1, further comprising:
(Item 3)
The processing system according to item 1 or 2, further comprising a search unit (for example, S37 to S40, S42 to S45) that searches the information recorded by the recording unit without converting it into plain text.

(Item 4)
The information recorded by the recording means includes personal information,
The decoding means disables the decoding of the personal information of the personal information owner in response to a request from the personal information owner (for example, S190 to S194, or S250 to S254, or S494, S495, or S485). ~S489), the processing system according to any one of items 1 to 3.

(Item 5)
a step of performing an encryption process (for example, S174, S177, or S228, S231, or S478, S479) to encrypt information to be recorded (for example, personal information);
Recorded by a recording means (for example, S177 and blockchain, or S231, S240, S242, S244, blockchain and personal information DB29, or S479 and blockchain) that records the information after the encryption process. A decryption step in which information is decrypted using a first key and a second key to convert it into plaintext information (for example, S201, S202, S207 to S209, or S263 to S271, or S500 to S510) and,
A step (for example, S191 to S194, or S250 to S254, or S494, S495, or S485 to S489) of rendering the information recorded by the recording means in a decodable state in which it cannot be decoded;
make the computer run
The decryption step includes a step (for example, S194, S233, or S475) of keeping the second key (for example, half common key KA) secret;
The step of making the decryption impossible includes updating the second key held in the holding step to another key (for example, a random number R) to make the decryption impossible (for example, S190 to S194). , or S250 to S254, or S494, S495, S485 to S489), a program.

(Effect of feature 1)
According to feature 1, it is possible to eliminate as much as possible the dilemma in which guaranteeing the authenticity of recorded information and guaranteeing the right to delete that information are contradictory.
(Feature 2)
[Technical field]

Feature 2 relates to smart contracts used, for example, in blockchains.
[Background technology]

A smart contract is a computer protocol intended for smooth contract verification, condition confirmation, execution, execution, and negotiation, and has traditionally been used in blockchains and other systems. This smart contract has been known for automating contracts, transactions, etc. (for example, Japanese Patent No. 6403177).
[Summary of feature 2]
[Issues that Feature 2 attempts to solve]

In the field of smart contracts, there is a demand for sophisticated smart contracts that can execute legal acts such as various contracts or transactions on behalf of users themselves, such as sales contracts and rental contracts.

The purpose of feature 2, which was devised in view of the actual situation, is to provide an advanced smart contract that allows users to perform legal acts on their own behalf.
[Means to solve the problem]
The subject matter of feature 2 is shown as the following items, for example.
(Item 1)
Machine learning means (for example, S80 to S82) that inputs information regarding legal acts performed by multiple natural persons or legal entities as data for machine learning and generates a general model;
Means for personalizing the general model into a model suitable for the user, the personalizing means for personalizing based on information regarding legal acts performed by the user (for example, S86 to S88, or S94 to S98) and
A computer comprising: smart contract generation means (for example, S86 to S88, or S94 to S98) for generating a smart contract for executing a legal act on behalf of the user using the personalized model. system.

(Item 2)
A means for personalizing a general model generated by inputting information about legal acts performed by multiple natural persons or legal entities as data for machine learning into a model suitable for the user, and personalization means (for example, S80 to S82, S86 to S88, or S94 to S98) for personalizing based on information related to legal acts;
A computer comprising: smart contract generation means (for example, S86 to S88, or S94 to S98) for generating a smart contract for executing a legal act on behalf of the user using the personalized model. system.

(Item 3)
A means for personalizing a general model generated by inputting information about legal acts performed by multiple natural persons or legal entities as data for machine learning into a model suitable for the user, and personalization means (for example, S80 to S82, S86 to S88, or S94 to S98) for personalizing based on information related to legal acts;
A computer system comprising: a service providing means (for example, S99) that provides a service for executing a legal act on behalf of the user using the personalized model as a smart contract.

(Item 4)
Reinforcement learning means (for example, S99) for learning a policy for maximizing the accumulation of the reward by giving the model a reward for the legal act performed in connection with the provision of the service by the service providing means (for example, S99). For example, the computer system according to item 3, further comprising steps S105 to S108).

(Item 5)
Predetermined themes (for example, policies and laws that the government is planning to adopt (for example, reduced tax rates due to consumption tax increase, amended immigration control law, UK's withdrawal from the European Union, partial or full basic income) 9 of the Japanese Constitution) is adopted, simulations of trading in the investment market such as stock trading and futures trading, company management simulation, consumption behavior simulation, etc.) can be performed in a computer. A computer system that advances reinforcement learning by using
selection means (for example, S344, S345) for selecting a group of users belonging to a plurality of personas that match the theme of the simulation;
a collection means (for example, S346) that groups the user group selected by the selection means into the plurality of personas and collects information regarding legal acts performed by the user group for each group;
A generating means (for example, S347) that performs machine learning using the collected information regarding legal acts as learning data to generate a learned smart contract model group for each persona;
a simulation means (for example, S336 to S339) that executes a simulation in a computer of performing a legal act between the generated learned smart contract models;
The simulation means is a reinforcement learning means (for example, S336, S338).

(Item 6)
A computer system that performs simulation in a computer to advance reinforcement learning,
By running a simulation in a computer in which a group of learned smart contract models generated by machine learning perform a legal act, and giving a reward for the executed legal act to the learned smart contract model, the trained smart contract model A computer system comprising reinforcement learning means (for example, S336, S338) that performs a simulation reinforcement learning process in which a contract model learns a policy that maximizes the accumulation of the reward.

(Item 7)
Item 6 further includes a selection means (for example, S340) for selecting a trained smart contract model to be actually used from the group of trained smart contract models based on the performance of the reinforcement learning result by the reinforcement learning means. Computer system as described.

(note)
The above-mentioned "machine learning data" for general model generation and "machine learning data" used for personalization only need to include "information related to legal acts," and other than "information related to legal acts." information (for example, access history to websites, GPS location information, etc.) may also be included. The above-mentioned "smart contract generation means" includes, for example, a case where an artificial intelligence such as a personal assistant, which is machine-learned based on information related to legal acts, is made to play the role of a smart contract.

(Effect of feature 2)
According to feature 2, it is possible to provide an advanced smart contract that allows users to execute various legal acts on their own behalf.

(Feature 3)
[Technical field]
Characteristic 3 is, for example, the policy or law that the government is trying to adopt (e.g. reduced tax rate due to consumption tax increase, amended immigration control law, the UK's withdrawal from the EU (European Union), partial or full adoption of basic income). , amendments to Article 9 of the Japanese Constitution, etc.), marketing-related conditions (e.g., setting of prices and consideration for new products (including financial products and life insurance) and new services, promotion effects through various media, etc.), investment market-related conditions (e.g., weather conditions for futures trading, monetary tightening policy for stock markets, etc.), perform a simulation in a computer under those conditions, and predict in advance what the simulation results will be. Concerning computer systems.
[Background technology]

As a computer system of this type, a new statement of changes in net assets account has been established to clarify the future debts that the people will have to bear and the resources that will be available in the future, and to support decision-making at the policy level. Accounting methods have been proposed that can clarify changes in assets due to policy decisions and simulate future burdens on citizens (for example, Japanese Patent Laid-Open No. 2006-155233).
[Summary of feature 3]
[Issues that Feature 3 attempts to solve]

In the field of simulation, there is a need for a computer system that can run a simulation on a computer that faithfully imitates the activities of natural persons and corporations in the real world, and can derive simulation results with as little deviation from the real world as possible.

The purpose of feature 3, which was devised in view of the actual situation, is to enable a simulation that faithfully imitates the activities of natural persons and corporations in the real world.
[Means to solve the problem]

The subject matter of feature 3 is shown as the following items, for example.
(Item 1)
Predetermined conditions (for example, policies or laws that the government is planning to adopt (e.g. reduced tax rate due to consumption tax increase, amended immigration control law, UK's withdrawal from the European Union, partial or full basic income) recruitment, amendments to Article 9 of the Japanese Constitution, etc.), marketing-related conditions (e.g., setting of prices and consideration for new products (including financial products and life insurance) and new services, promotion effects through various media, etc.), investment A computer system that performs simulations within a computer under market-related conditions (e.g., weather conditions in futures trading, monetary tightening policies in stock markets, etc.),
selection means (for example, S144, S145) for selecting a group of users belonging to a plurality of personas that match the conditions of the simulation;
a collection means (for example, S146) that groups the user group selected by the selection means into the plurality of personas and collects information regarding legal acts performed by the user group for each group;
a generation unit (for example, S147) that performs machine learning using the collected information regarding legal acts as learning data to generate a learned smart contract model group for each persona;
a simulation means (for example, S136 to S139) that executes a simulation in a computer of performing a legal act between the generated learned smart contract models;
derivation means (for example, S140) for deriving the results of the simulation by the simulation means,
The simulation means is a reinforcement learning means (for example, S136, S138).

(Item 2)
A computer system that performs simulation within a computer,
A simulation in which a group of learned smart contract models generated by machine learning (for example, S144 to S146) perform a legal act is executed in a computer, and a reward for the executed legal act is transferred to the learned smart contract model that has been executed. Reinforcement learning means (for example, S136, S138) for proceeding with reinforcement learning in which the learned smart contract model learns a policy that maximizes the accumulation of the reward by giving
A computer system comprising: deriving means (for example, S140) for deriving a result of a simulation in which a group of learned smart contract models perform a legal act with each other through reinforcement learning performed by the reinforcement learning means.
(Effect of feature 3)

According to feature 3, it is possible to perform a simulation that imitates the activities of natural persons and corporations in the real world as faithfully as possible.
[Third embodiment]

Next, a third embodiment will be described. In this third embodiment, by performing a simulation in a mirror world (cyber space) consisting of a digital twin of the real world (real world) as a simulation environment, for example, an optimal solution that predicts the future can be derived, It relates to a system that derives the optimal solution for incentive design in a DAO (Decentralized Autonomous Organization) and performs AI machine learning (for example, reinforcement learning).

A digital twin is a digital representation of a real-world entity or system. A mirror world is a mirror image world made up of digital twins in which all information from the physical world (real world), such as real countries, cities, societies, local governments, companies, and other organizations and people, is digitized. . Specifically, we will develop a digital twin of a person, such as machine learning (e.g., agent It consists of an assistant AI (hereinafter referred to as "personal AI") that has undergone reinforcement learning (reinforcement learning). This reinforcement learning is multi-agent reinforcement learning in which multiple personal AIs cooperate to perform reinforcement learning. The personal AI of the people who make up an organization such as a company in the real world constitutes the digital twin of the organization, and the personal AI of the people who make up the local government in the real world constitutes the digital twin of the local government. , the personal AI of the people who make up a city in the real world constitutes a digital twin of that city, and the personal AI of the people who make up a country in the real world constitutes a digital twin of that country.

In this third embodiment, real countries, cities, societies, local governments, companies, and other organizations and people in the real world take the initiative to participate and cooperate in building a mirror world as a simulation environment. Prepare a system. Specifically, by performing various simulations within the mirror world, the personal AI of the digital twin participating in the simulation is subjected to machine learning (e.g. reinforcement learning), and the trained personal AI that has learned to a higher level is made real. Give back (feedback) to the world. With the enjoyment of this benefit as an incentive, real world nations, cities, societies, local governments, companies, and other organizations and people are encouraged to take the initiative to participate and cooperate in building the mirror world.

Referring to FIG. 28, mirror world data is stored in a data center 45 in which a plurality of mirror world servers (including storage servers) 46 are installed. The hardware configuration of the mirror world server 46 is similar to the hardware configuration of the user terminal 16 shown in FIG. 1, so repeated illustration and description thereof will be omitted here. Digital twins (real-world national digital twins (e.g. Japan The entire mirror world 51 composed of a country digital twin 53), a city digital twin 54, society, local governments, organizations such as companies, people, and a global digital twin 52) is stored in the data center 45 as digital data.

The data center 45 performs a simulation using the mirror world 51 as a simulation environment, and derives an optimal solution that foresees the future by, for example, simulation optimization. The simulations include, for example, the policies and laws that the government is trying to adopt (e.g., the reduced tax rate accompanying the consumption tax increase, the revised Immigration Control Act, the UK's withdrawal from the European Union (EU), partial or partial basic income, etc.). Possible examples include simulations of transactions in investment markets such as stock trading and futures trading, simulations of company management, and simulations of consumption behavior, etc., in the case where the system is adopted (full adoption, amendment of Article 9 of the Japanese Constitution, etc.). Furthermore, simulations of promotions of new products (including financial products and life insurance) and new services through various media may also be used. The optimal solution derived through simulation optimization is fed back to the real world, and the benefits of the optimal solution are provided to the real world. In addition, trained personal AI that has undergone machine learning (for example, reinforcement learning) through simulation is returned to the real world, allowing more advanced trained personal AI to perform tasks.

By linking this personal AI and a smart contract, the above-mentioned linked type AI smart contract is configured. Note that this data center 45 is connected to the Internet 1 shown in FIGS. 1 and 24. In FIG. 28, various blockchains 2, 3, 4, SNS 19, key registration center 30, etc. are not shown.

FIG. 29 shows a specific example of the city digital twin 54 in the mirror world 51. In the city 50 of the real world 47, there are ABC 56, a person named Taro 55, and Taro's family 56. The corresponding city digital twins 54 include ABC digital twin 59, Taro digital twin (Taro's personal AI) 57, Taro's family digital twin 58, and the like. City digital twin data consisting of these data is stored in the mirror world server 46. If there are changes to various objects such as ABC56, the person Taro 55, and Taro's family 56 in Real World 49 (for example, personnel changes in the company, employment or retirement, marriage or childbirth of people, etc.), The corresponding digital twins will be updated with the changed contents. Such city digital twin data is stored in the data center 45 for each city and becomes data for the digital twin 53 of Japan 49, and city digital twin data for each country is stored in the data center 45 for each city. It becomes the digital twin data of each country, and all the digital twin data becomes the data of 52 digital twins of 48 Earths.

As a specific example, the mirror world server 46 contains Taro's digital twin (Taro's personal AI) 57, name: Taro, AI identification number: 82km9, personal AI data, and Taro's personal data (for example, life log, profile, preferences). data, electronic medical record data, vital data, etc.) are stored. As Taro's family digital twin 58, names: Taro, Sakura, Shiro, family composition: husband, wife, eldest son, AI identification numbers: 82km9, 11zk9, gf43y are stored. As ABC Digital Twin 59 Co., Ltd., names: Taro, Hanako... Saburo, positions: representative director, managing director, general manager... ordinary employee, AI identification number: 82km9, ba935, 2es14,...9w1c2 are memorized. There is.

A flowchart of the main routine program between the user terminal 16 and the mirror world server 46 will be explained based on FIGS. 30 to 33. Referring to FIG. 30A, the CPU 10 of the user terminal 16 executes a member registration request process S555 for requesting participation registration in the mirror world 51 as a simulation environment, a simulation preparation response process S556, and a simulation response process S557. The CPU 10 of the mirror world server 46 executes member registration processing 550, simulation preparation processing S551, and simulation processing S552.

The member registration process and the member registration request process will be explained based on FIG. 30B. Both of these processes are for registering members who wish to participate as digital twins in a simulation using the mirror world 51 as a simulation environment. In the member registration request process, the CPU 10 of the user terminal 16 determines whether or not to apply for registration in S560, and if it is determined not to apply for registration, the member registration process ends and returns. If it is determined that a registration application is to be made, in S561, predetermined items necessary for the registration application are sent to the mirror world server 46, and if there is a person who does not have a personal AI, that fact and the person's blockchain address are sent. is also transmitted to the mirror world server 46. Specifically, the prescribed items required for registration application include, for a digital twin of a person, the AI identification number and personal AI data of the person's personal AI, and for a digital twin of a family, the name and family composition of the family member. In the case of a company's digital twin, this includes the employee's name, position, and each AI identification number.

The CPU 10 of the mirror world server 46, which received it in S565, determines in S566 whether or not the personal AI is already owned. If the information sent in S561 includes the information that "you do not have a personal AI", the control proceeds to S567, where personal AI generation and sales processing is performed; If the information "not being used" is not included, control proceeds to S568, and predetermined items including the AI identification number sent in S562 are registered in the mirror world 51.

The personal AI generation and sales process shown in S567 will be explained based on FIG. 31. In S573, the CPU 10 of the mirror world server 46 transfers the transaction data and SNS posting data recorded in the blockchain address received in S565 (blockchain address of the user who does not have personal AI) from the blockchain. collect. Next, in S574, machine learning is performed using transaction data and posting data such as SNS as learning data to generate a learned personal AI. Next, in S575, the trained personal AI is sold to the corresponding user.

The simulation preparation process shown in S551 and the simulation preparation response process shown in S555 will be explained based on FIG. 32. In S577, the CPU 10 of the mirror world server 46 determines whether a request for simulation has been received. If it is determined that the simulation has not been received, this simulation preparation process ends and returns. If it is determined that a simulation request has been received, the control proceeds to S578, where a process is performed to find a personal AI group and a digital twin that match the requested simulation. For example, in the case of a consumption behavior simulation under the reduced tax rate associated with the consumption tax increase mentioned above, a personal AI group corresponding to general consumers is used, and the data is analyzed according to demographic statistics such as gender, age, region, annual income, etc. Personal AI group, digital twins of manufacturers of consumer goods, digital twins of retailers, etc., which are subject to the reduced tax rate, will be determined based on the percentages of personal AI groups. Next, processing is performed to request consent for the simulation from the identified personal AI group and digital twin. Specifically, the contents of the simulation are sent to the user terminals 16 of each of the user groups corresponding to the identified personal AI group and digital twin, and the user is asked whether or not they agree.

The CPU 10 of the user terminal 16 of each user group corresponding to the identified personal AI group and digital twin receives the content of the transmitted simulation in S580, and agrees to participate as an execution member of the simulation in S581. Determine whether or not to do so. This determination may be made by the personal AI, or may be made by the user himself. If it is determined that they do not agree, the simulation preparation response process ends and returns, but if it is determined that they agree, a message that they agree is sent back to the mirror world server 46 in S582.

The CPU 10 of the mirror world server 46, which received the request in S583, determines whether or not the amount of consent necessary to execute the requested simulation has been obtained. If it is determined that consent has been obtained, the AI group and digital twin for which consent has been obtained are copied and registered in the mirror world 51 as simulation targets. The registered state is shown in FIG. 29 mentioned above.

On the other hand, if it is determined that consent has not been obtained from the necessary personal AI group and digital twin, control proceeds to S585, and the persona group (manufacturer or retailer) that matches the missing personal AI group and digital twin In S586, a group of users belonging to each persona (including a group of users working at manufacturers and retailers) is selected, and in S587, users belonging to each persona are set. The groups are grouped and the transaction data of the user group (including transaction data as manufacturers and retailers) is collected from the blockchain for each group, and in S588, machine learning is performed using the transaction data as learning data to create a system for each persona. After generating and replenishing the learned personal AI group and digital twin, the process advances to S584. These steps S585 to S588 are the same processes as steps S344 to S347 in FIG. 9(B), and detailed description thereof will not be repeated here.

Next, specific control of the simulation process shown in S552 and the simulation response process shown in S557 will be explained based on FIG. 33. S593 to S595 are the same processes as S336, S338, and S339 in FIG. 9 and S136, S138, and S139 in FIG. 12 described above, and detailed description thereof will be omitted here. In S596, the simulation result is notified to the simulation requester. Specifically, the simulation results are transmitted to the user terminal 16 of the simulation requester. Next, in S597, each personal AI used in the simulation (including personal AI engaged in digital twin of a company organization, etc.) is transmitted to the user terminal 16 of each owner.

The CPU 10 of the user terminal 16, which received it in S598, determines in S599 whether or not to erase the received personal AI. The received personal AI is a learned AI that has been subjected to reinforcement learning (machine learning) by participating in a simulation, and its performance has increased accordingly, making it possible to execute advanced task processing. However, depending on the content of the simulation, the user may be receiving reinforcement learning (machine learning) that is not desired by the user, so in such a case, a YES determination is made in S599, and the received personal AI is deleted in S601. do. On the other hand, if the simulation is the content desired by the user and it is determined that the received personal AI has undergone desired reinforcement learning (machine learning), control proceeds to S600 and the received trained personal AI is overwritten and saved. . As a result, the user has the advantage of being able to obtain a personal AI with improved performance through desired reinforcement learning (machine learning). By using the enjoyment of this advantage as an incentive, organizations such as countries, cities, societies, local governments, companies, and people in the real world can be encouraged to take the initiative to participate and cooperate in building the mirror world. By overwriting and saving the personal AI in S600, the data is updated in the mirror world server 46 to the digital twin of the new personal AI after being overwritten and the digital twin of the organization made up of the new personal AI (Fig. 29). Note that instead of overwriting and saving, both the existing personal AI and the learned personal AI may be stored together and used separately as necessary.

Next, a system for deriving an optimal solution for incentive design in a DAO by performing a simulation using the mirror world as a simulation environment will be described based on FIGS. 34 to 59. FIG. 34(A) is a schematic diagram of a multi-service DAO construction system. For example, Bitcoin is a type of DAO, but nodes (miners) only perform one type of service: mining (competing for bookkeeping rights), and those who succeed in mining are given an incentive to receive Bitcoin. This allows blocks to be added and the Bitcoin system to continue autonomously. On the other hand, a DAO in which multiple types of services exist is called a multi-service DAO. For example, in the case of a DAO of a company organization, there are multiple services such as material procurement, assembly, advertising, sales, etc., and what is the optimal ratio and how much reward should be distributed to the nodes that performed these multiple services? The difficult question is whether the incentive design will be appropriate. A system for deriving an optimal solution for incentive design in such a multi-service DAO will be described.

Referring to FIG. 34(A), the mirror world server 46 storing multi-service DAO data stores a DAO agent 61, a persona agent group 62 that performs service 1, a persona agent group 63 that performs service 2, . . . It stores the data of the persona agent group 64 that performs n. Furthermore, the mirror world server 46 also stores the types of rewards r1, r2, . . . rn given to each persona agent group as a result of reinforcement learning.

The terminal 16 of the multi-service DAO construction company downloads and installs the DAO agent 61, the necessary persona agent group, and the types of rewards r1, r2, . . . rn from the mirror world server 46. The terminal 16 is each node 19 that constitutes the public chain 4. Simulation reinforcement learning is performed on the digital twin 66 of the multi-service DAO 65 operated in the public chain 4 made up of each node 19 within the mirror world 51 to derive an optimal solution for incentive design in the multi-service DAO. The optimal solution for the incentive design is applied to the actual multi-service DAO 65 in the real world 47 to create a multi-service DAO 65 with an optimal incentive design. This simulation reinforcement learning within the mirror world 51 is executed in the mirror world server 46. The control will be explained below.

Referring to FIG. 34(B), the CPU 10 of the terminal 16 performs simulation reinforcement learning preparation response processing in S606, and performs simulation reinforcement learning response processing in S607.

The CPU 10 of the mirror world server 46 performs simulation reinforcement learning preparation processing in S611, simulation reinforcement learning processing in S612, and DAO agent reinforcement learning processing in S613.

Specific control of the simulation reinforcement learning preparation process shown in S611 and the simulation reinforcement learning preparation response process shown in S606 will be explained based on FIG. 35. In the simulation reinforcement learning preparation response process, the CPU 10 of the terminal 16 determines whether to request simulation reinforcement learning in S615. If no request is made, this simulation learning preparation response processing is completed and the process returns. If a request is to be made, control proceeds to S616, where the multi-service DAO data is transmitted to the mirror world server 46 and a request is made. This multi-service DAO data includes the type of service. For example, in the case of the innovation-inducing DAO described later in FIGS. 36 to 42, there are five types of services: idea generation, improvement proposal, commercialization, infringement discovery, and token purchase, and these services are transmitted.

The CPU 10 of the mirror world server 46, which received this in S620, sets a persona group that matches each of the multi-services in S621. For example, in the case of the above-mentioned innovation-inducing DAO, the persona group for idea generation and improvement suggestions is people who often come up with inventions, the persona group for commercialization is people who are interested in commercialization, and the persona group for discovering infringements is people related to patent law and copyrights. Possible personas include people who are knowledgeable about the law, and people who are interested in investing as a persona group for purchasing tokens.

Next, in S622, a user group belonging to each persona is selected. For example, in the case of the above-mentioned innovation-inducing DAO, users who are listed as inventors of patent applications belong to the persona group of idea generation and improvement proposals, and company managers belong to the persona group of commercialization. The user group that belongs to the infringement discovery persona group is a user group of patent attorneys and lawyers, and the user group that belongs to the token purchase persona group is a user group that has purchased virtual currencies such as Bitcoin. .

Next, in S623, users belonging to each persona are grouped and transaction data of the user group is collected from the blockchain for each group, and in S624, machine learning is performed using the transaction data as learning data to learn for each persona. Generate a set of completed persona agents. These two controls are the same processes as S346 and S347 in FIG. 9(B) described above, and repeated detailed explanations will be omitted here. Next, in S625, a group of persona agents is deployed in the multi-service DAO 65 to generate a multi-service DAO digital twin 66, and registered in the mirror world 51 as a simulation target. The state is shown in FIG.

Referring to FIG. 36, a digital twin 66 of a multi-service DAO 65 consisting of a public chain is constructed in the mirror world 51. The multi-service DAO digital twin 66 is equipped with a group of persona agents in step S625, and one persona agent is installed in each node. The identification numbers of these persona agents are classified and stored in the mirror world server 46 for each service (idea proposal, improvement proposal, commercialization, infringement discovery, token purchase). For example, kc29m, 1w13a, . . . 9nad8 are stored in the mirror world server 46 as identification numbers of persona agents for idea generation services. This multi-service DAO 65 is the innovation-inducing DAO described above, and hereinafter, the multi-service DAO will be explained using the innovation-inducing DAO as an example.

Furthermore, the mirror world server 46 also stores DAO agents that provide rewards (incentives) for the actions of each persona agent. This DAO agent derives the optimal solution for incentive design by performing reinforcement learning (machine learning) on the reward distribution ratio and reward amount. A schematic system of reinforcement learning (machine learning) is shown in FIG. 37.

Referring to FIG. 37, when a group of persona agents posts an origin idea as actions a11, a12, . Rewards r11, r12, . . . r1n are given to 67. This idea generation service is a broad concept that includes ideas such as dreams, ideas, business plans, technical ideas, and copyrighted works. The environmental state S1 is also given to the persona agent group 67 that performed the idea generation service. The state S1 of this environment is, for example, the contents of the origin idea submission, the fluctuating market price of the token A1 given as a reward to each persona agent 68 who performed the idea generation service, and the like.

When the persona agent group 68 performs actions a21, a22, . Rewards r21, r22, . . . r2n are given to the agent group 68. The environmental state S1 is also given to the persona agent group 68 that performed the improvement plan service. The state S2 of this environment is, for example, the content of the improvement proposal posting, the number of "likes" given to the improvement proposal posting, and the like. The entities that give this "Like!" are limited to, for example, those who purchased the token A1 given as a reward for posting the origin idea (persona agent group 71). In this way, the reason for limiting (restricting) who gives "Likes" to interested parties (stakeholders) is to prevent fraudulent acts. If the parties that give likes are unlimited, for example, the person who posted the improvement proposal (persona agent group 68) colludes with many people (persona agents) and receives a large number of likes. This is to prevent fraudulent acts such as asking someone to attach a . The person who gave "Like" to the persona agent group 69 who performed commercialization services and the persona agent group 70 who performed infringement handling services was also the person who purchased token A1 for the same reason (persona agent group 71). It is limited to only.

When the persona agent group 69 who performed commercialization services for the above-mentioned origin idea performs actions a31, a32, ... a3n, the environmental state S3 is input to the DAO agent 61, and the persona who performed those commercialization services Rewards r31, r32, . . . r3n are given to the agent group 69. Actions a31, a32, ... a3n of the persona agent group 69 include, for example, posting a business plan, posting the progress of commercialization, posting the status of actual commercialization, and revenue from the commercialized business. Possible options include posting the amount. The environmental state S3 is also given to the persona agent group 69 that has performed commercialization services. The state S3 of this environment is, for example, the number of "Likes" given to postings of business plans, postings of the progress of commercialization, postings of profits from commercialized businesses, etc.

When the persona agent group 70 that performed infringement handling services for the above-mentioned origin idea performs actions a41, a42, ... a4n, the environmental state S4 is input to the DAO agent 61, and the persona that performed those infringement handling services Rewards r41, r42, . . . r4n are given to the agent group 70. The actions a31, a32, . . . a3n of the persona agent group 70 may include, for example, posting a report of discovery of infringement, posting a report on dealing with infringement, posting a license negotiation report, etc. Furthermore, it may also include the service of posting reports on patent applications and the acquisition of rights, which are prerequisites for these services. The environmental state S4 is also given to the persona agent group 70 that performed the infringement response service. The state S4 of this environment is, for example, the number of "Likes" given to the posting of a report of discovery of infringement, the posting of a countermeasure against infringement, the posting of a license negotiation report, etc.

When the persona agents perform the service of purchasing the token A1 granted for the idea generation service of the origin as actions a51, a52, ... a5n, the environmental state S5 is input to the DAO agent 61 and the token purchase is performed. Rewards r51, r52, . . . r5n are given to the persona agent group 71 who performed the service. The environmental state S5 is also given to the persona agent group 71 that performed the token purchasing service. The state S5 of this environment is, for example, the number of purchased tokens (or the purchased amount). The persona agent group 71 consumes virtual currency (for example, Ethereum's ETH, etc.) to purchase the token A1. The purchased tokens can be converted (exchanged) into virtual currency according to the price on the floating market, and the virtual currency can be converted (exchanged) into legal currency such as yen or dollars according to the price on the floating market.

The rewards r1 to r5 to be given to each persona agent group are determined by the DAO agent 61 based on the reward table (see FIG. 39(A)). For the persona agent who performed the idea generation service, r1=A1+B1・b+G1・g, for the persona agent who performed the improvement service, r2=A2・e+B2・b+G2・g, the business For the persona agent who performed the violation service, r3=A3・e+B3・b, and for the persona agent who performed the infringement response service, r4=A4・e+B4・b+G4・g, and the token purchase service was performed. For the persona agent, r5=B5・b+G5・g is determined.

Here, A2 to A4, B1 to B5, G1, G2, G4, and G5 are coefficients, and the DAO agent 61 converges them to optimal ones by reinforcement learning. A1 is the token, g is the license revenue, e is the number of "likes", and b is the commercialization revenue.

Note that only the license income g or commercialization income b generated after each persona agent group 68 to 71 performs the service is considered as the remuneration r2 to r5. This is to prevent fraudulent activity in which improvement services or token purchase services are later performed on an origin idea for which license income g or commercialization income b has already been generated.

Furthermore, a group of persona agents that have performed improvement services, commercialization services, or infringement response services may also perform token purchase services. Furthermore, the persona agent group 67 that performed the idea generation service may also perform improvement services, commercialization services, or infringement response services.

Details of the DAO agent reinforcement learning process shown in S613 will be explained based on FIG. 38. In this process, the DAO agent 61 performs reinforcement learning on its own to optimize the rewards r1 to r5. In S630, the DAO agent 61 determines whether or not each action a of the persona agent group has been received. If it has not been received, the process proceeds to S632, but if it is determined that it has been received, the control proceeds to S631, where each received action a is stored.

In S632, it is determined whether or not "Like" has been given. If "Like" has not been given, the process proceeds to S634, but if it is determined that "Like" has been given, in S633, a "Like" has been given for each persona agent. Remember e. In S634, it is determined whether or not there is commercialization profit. If there is not, the process proceeds to S636, but if it is determined that there is, commercialization profit b is stored in S635. In S636, it is determined whether or not there is a license profit g. If there is not, the process proceeds to S638, but if it is determined that there is, the license profit g is stored in S637.

In S638, it is determined whether or not the remuneration calculation time has come. If it is not, the process proceeds to S640. If it is determined that the time has come, in S639, each remuneration is calculated with reference to the remuneration table (FIG. 39(A)). Calculate r1 to r5 and give it to the corresponding persona agent. In S640, it is determined whether the learning update time has come, and if it has not, the DAO agent reinforcement learning process ends and returns. If it is determined that it is time to update the learning, in S641, the total given price TT of the token A1 given as a reward and the current total price TB in the fluctuating market of the given token are calculated, and in S642, the TB/TT is calculated. Calculate the DAO agent's reward R from the value. For example, compare the value of TB/TT at the time of the previous learning update and the value of TB/TT at the time of the current learning update, and if the value of TB/TT at the time of the current learning update is larger, the reward R will be larger. and if the reward R is small, the reward R is small. As a result, the reward R that the DAO agent 61 can obtain increases as the total price TB of tokens in a fluctuating market increases, and decreases as the total price TB of tokens in a fluctuating market falls.

Next, in S643, based on the reward R, TD learning is performed to obtain the actions A1 to A4, B1 to B5, G1, G2, G4, and G5 according to the optimal policy π ^* , and in S644, the reward table is Update A1-A4, B1-B5, G1, G2, G4, G5 to the requested actions A1-A4, B1-B5, G1, G2, G4, G5. As a result, the DAO agent 61 learns the optimal actions A1 to A4, B1 to B5, G1, G2, G4, and G5 to raise the total price TB in the fluctuating market of the token. Note that this learning objective is just one example; other learning objectives include increasing the number of original ideas submitted, increasing the total number of original ideas and improvement ideas submitted, and increasing the number of commercializations. It may also be to increase the total commercialization revenue.

Next, in S645, it is determined whether the reinforcement learning has been completed, and if it has not been completed, the process returns. If it is determined that the process has been completed, the learned multi-service DAO is transmitted to the simulation reinforcement learning requester in S646.

A requester of simulation reinforcement learning can operate a trained multi-service DAO (innovation-inducing DAO) 65 with an optimized incentive design in the real world 47. As a result, in this multi-service DAO (innovation-inducing DAO) 65, "each persona agent group 67 to 71" shown in FIG. 37 becomes an actual user group, and the learned DAO agent 61, the optimally designed reward (incentive) is distributed. At the stage of actual operation in this real world 47, services such as the contents of each post and contents of token sales and transactions are recorded on the blockchain with time stamps. As a result, blockchain can play the role of a notary public for the content of origin ideas and improvement proposals, and apply exceptions to loss of novelty (Article 30 of the Patent Act) and countermeasures against fraudulent applications (Article 49, Paragraph 1, Line 7 of the Patent Act). , Article 74, Article 123, Paragraph 1, Item 2) will also become easier to implement.

In addition, even at the stage of operating a multi-service DAO (innovation-inducing DAO) 65 in the real world 47, the DAO agent 61 continues machine learning (reinforcement learning) to provide even more optimal incentives that match the actual operating situation. It may be designed to be the same. In addition, each learned persona agent group 67 to 71 (learned persona agent group according to FIGS. 40(A)(B) and FIG. 41(A)(B)) is also included in the multi-service DAO (innovation-inducing DAO) 65. The persona agent groups 67 to 71 may function as advisors to the user group performing each service by transmitting the information to the requester of the simulation reinforcement learning. Furthermore, at the stage of operating the multi-service DAO (innovation-inducing DAO) 65 in the real world 47, even if the multi-service DAO 65 is a mixed type in which each service is executed by both the user group and each persona agent group 67 to 71. Alternatively, a persona agent-operated multi-service DAO (innovation-inducing DAO) 65 may be used in which each service is executed only by each persona agent group 67 to 71. Note that this innovation-inducing DAO is not limited to those generated through simulation reinforcement learning in the mirror world described above, but may be generated artificially based on other methods, such as artificial design. Moreover, it is not limited to the DAO, but may also be an organization with a specific administrator or entity (for example, a regular stock company, etc.).

Next, the main routine of the process in which the persona agent performs reinforcement learning will be explained based on FIG. 39(B). In S648, an idea generation service execution process is performed, in S649, an improvement service execution process is performed, in S650, a commercialization service execution process is performed, in S651, an infringement response service execution process is performed, and in S652, Token purchase service execution processing is performed.

Details of the idea generation service execution process shown in S648 will be explained based on FIG. 40(A). In S655, it is determined whether or not to generate an idea, and if not, the process returns. If it is determined to do so, in S656, a process of creating an idea is performed. The creation of this idea uses an AI called DABUS, for example. For example, persona agent 67 and DABUS collaborate to generate ideas. In S657, the content of posting the idea is generated, and in S658, the idea posting act a1i is executed.

In S659, it is determined whether the reward r1i has been received from the DAO agent 61, and if it has not been received, the process returns. If it is determined that it has been received, in S660, an action a according to the optimal policy π ^* is determined by TD learning based on the reward r1i. If the received reward r1i is satisfactory, this action a will continue the act of idea generation repeatedly, but if the received reward r1i is not satisfactory, other actions (e.g. improvement services, commercialization) Services, infringement response services, token purchase services, or no services at all).

Details of the improved service execution process shown in S649 will be explained based on FIG. 40(B). In S664, it is determined whether or not to post the improvement proposal, and if it is not posted, the process returns. If it is determined to post, a process of creating an improvement plan is performed in S665. For example, an AI called DABUS is used to create improvement plans. For example, the persona agent 68 and DABUS collaborate to create an improvement plan. In S666, the content of the improvement proposal posting is generated, and in S667, the improvement proposal posting action a2i is executed.

In S668, it is determined whether the reward r2i has been received from the DAO agent 61, and if it has not been received, the process returns. If it is determined that it has been received, in S669, an action a according to the optimal policy π ^* is determined by TD learning based on the reward r2i. If the received reward r2i is satisfactory, this action a will continue to repeat the act of posting improvement proposals, but if the reward r2i is not satisfactory, other actions (for example, idea generation services, The choice is commercialization services, infringement response services, token purchase services, or no services at all.

Details of the commercialization service execution process shown in S650 will be explained based on FIG. 41(A). In S674, it is determined whether or not to commercialize the process, and if the process is not commercialized, the process returns. If it is determined to be commercialized, a business plan is generated in S675, a business plan posting act a3i is executed in S676, a commercialization service is performed in S677, and an execution status posting act a3i is executed in S678. Execute. This performance status posting act a3i also includes posting of profits obtained from commercialization as described above.

In S679, it is determined whether the reward r3i has been received from the DAO agent 61, and if it has not been received, the process returns. If it is determined that the received reward r3i has been received, in S680, an action a according to the optimal policy π ^* is determined by TD learning based on the received reward r3i. If the received remuneration r3i is satisfactory, this act a will repeatedly continue the act of commercialization services, but if the remuneration r3i is not satisfactory, other acts (for example, idea generation services, Customers can choose between improvement services, infringement response services, token purchase services, or no services at all.

Details of the infringement countermeasure service execution process shown in S651 will be explained based on FIG. 41(B). In S684, it is determined whether or not to execute the infringement handling service. If it is not to be carried out, the process returns; however, if it is determined to be carried out, an investigation into the infringement is carried out in S685, and it is determined in S686 whether or not an infringement has been discovered. Note that, as described above, before conducting an investigation into infringement, it is also possible to apply for a patent or obtain rights therefor. If no infringing act is found, return is made, but if it is determined that infringing act is found, a warning letter is generated to the suspected infringer in S687, and a warning letter posting act a4i is executed in S688. Then, in S689, the infringement countermeasure action a4i, such as negotiation with the suspected infringer, is carried out, and in S690, the execution status posting action a4i is carried out.

Next, in S691, it is determined whether the reward r4i has been received from the DAO agent 61, and if it has not been received, the process returns. If it is determined that the received reward r4i has been received, in S692, an action a according to the optimal policy π ^* is determined by TD learning based on the received reward r4i. If the received remuneration r4i is satisfactory, this act a will repeatedly continue the act of commercialization services, but if the remuneration r4i is not satisfactory, other acts (for example, idea generation services, Customers can choose between improvement services, commercialization services, token purchase services, or no services at all.

Next, details of the token purchase service execution process shown in S652 will be explained based on FIG. 42(A). In S969, it is determined whether or not to purchase the token, and if the token is not purchased, the process returns. If it is determined that the token is to be purchased, the token purchase action a5i is executed in S697. Next, in S698, it is determined whether the reward r5i has been received from the DAO agent 61, and if it has not been received, the process returns. If it is determined that the received reward r5i has been received, an action a according to the optimal policy π ^* is determined by TD learning based on the received reward r5i. If the received remuneration r5i is satisfactory, this act a will repeatedly continue the act of commercialization services, but if the remuneration r5i is not satisfactory, other acts (for example, idea generation services, The customer can choose between improvement services, commercialization services, infringement response services, or no services at all.

Based on FIG. 42(B), the price fluctuation of the token 72 in the fluctuating market due to the token purchase by the persona agent group 71 will be explained. Persona agent 67 who performed the idea generation service was given 50 tokens (market capitalization 50,000 yen) 72 as reward A1, and part of the tokens 72 (10 tokens) was paid in virtual currency equivalent to 10,000 yen. Purchased by Persona Agent 71a. Next, Persona Agent 71b purchased those 10 tokens by paying 15,000 yen worth of virtual currency. As a result, the value of 10 tokens will soar to 15,000 yen. Persona Agent 71c paid 20,000 yen worth of virtual currency to purchase it. As a result, the value of 10 tokens will soar to 20,000 yen. Persona Agent 71d purchased it by paying virtual currency worth 25,000 yen. As a result, the value of 10 tokens will soar to 25,000 yen. Persona Agent 71e purchased it by paying virtual currency worth 30,000 yen. As a result, the value of 10 tokens will soar to 30,000 yen.

As a result, the 40 tokens held by Persona Agent 67 (market capitalization: 40,000 yen) will soar to a market capitalization of 120,000 yen. The value of this token rises in direct proportion to the expected value; the more popular the origin idea is, the more expensive it becomes; the more popular (more likes) improvements are posted, the more expensive it becomes; The more businesses are posted (with more likes), the higher it becomes, and the more popular (more likes) anti-infringement measures are posted, the higher it becomes.

At the stage of actually operating the multi-service DAO 65 in the real world 47, as described above, "each persona agent group 67 to 71" in FIG. 37 becomes a user group in the real world. In that case, not only the tokens 72 given to the group of users who performed the idea generation service, but also the tokens of the users themselves who perform various services (hereinafter referred to as "my tokens") may be bought and sold. When other users who view the posted service content purchase the poster's own My Tokens with expectations of the poster, the price of My Tokens in the fluctuating market soars. In this case, a part of the poster's income may be distributed to the My Token purchaser at a rate according to the purchase amount. This My Token may be issued within the Multi-Service DAO 65, but by linking it with the My Token issued to the user by a specialist company that issues and distributes My Tokens, the My Token issued by the specialist company can be used within the Multi-Service DAO 65. It may also be possible for users to buy and sell. VALU Co., Ltd. is currently a specialist in issuing and distributing MyTokens.

Next, a system in which a plurality of functional elements cooperate to construct one well-controlled DAO will be described based on FIGS. 43 to 59. Such a DAO is hereinafter referred to as an "element integrated DAO."

Referring to FIG. 43, this element-integrated DAO makes it possible to easily construct a company organization, etc. that already exists in the real world 47 using DAO, and an element DAO is generated and prepared in advance for each functional element. There is. In other words, a modular element DAO is prepared for each functional element, and the configuration is such that a desired element integrated DAO can be easily constructed by selecting and combining the necessary element DAOs. The element DAO provider 73 is provided with a server 74 and an element DAO protocol DB 75. The element DAO protocol DB 75 includes, for example, an element DAO prepared for each functional element necessary for company relations, an element DAO prepared for each functional element necessary for NPO (Nonprofit Organization) relations, and a functional element necessary for local governments. Element DAO etc. prepared for each are stored.

The element integrated DAO construction vendor receives an order for element integrated DAO construction from the client and installs the element DAO corresponding to the required functional element into the PC terminal 76 via the server 74. In the example in Figure 43, A1 element DAO (including A1 element agent), A2 element DAO (including A2 element agent), A5 element DAO (including A5 element agent), and A9 element DAO (including A9 element agent) are installed. . Each element agent A1 to A9 is an AI for performing reinforcement learning (machine learning) so that each corresponding element DAO can exhibit the best performance.

A supervising agent is also installed on the PC terminal 76. This supervising agent controls the element agents of each element DAO so that the entire element integrated DAO is optimized, and the supervising agent itself performs reinforcement learning (machine learning) to achieve overall optimization. do. Since each element agent is intended to maximize the performance of the element DAO for which it is in charge, there is a risk that using only the element agent will lead to partial optimization and that the overall optimization of the element integrated DAO may not be achieved. Therefore, a central agent is required to control the element integration DAO as a whole so that it is optimized. This can be said to be the same as, for example, searching for a Pareto optimal solution in an incomplete information game.

The element integrated DAO installed on the PC 67 is installed on multiple terminals 16 in order to perform simulation reinforcement learning using the mirror world 51 as a simulation environment. Twin 2T is generated in the mirror world 51.

The main routine of simulation reinforcement learning of this element integrated DAO will be explained based on FIG. 44(A). The CPU 10 of the user terminal 16 of the requester who requests simulation reinforcement learning of the element integrated DAO performs simulation reinforcement learning preparation response processing in S674, and performs simulation reinforcement learning response processing in S675. The CPU 10 of the mirror world server 46 performs simulation reinforcement learning preparation processing in S679, and performs simulation reinforcement learning processing in S680.

Details of the simulation reinforcement learning preparation response process shown in S674 and the simulation reinforcement learning preparation process shown in S679 will be explained based on FIG. 44(B). In the simulation reinforcement learning preparation response process, the CPU 10 of the user terminal 16 determines whether or not to request simulation reinforcement learning in S679, and returns if not requested. If it is determined that the request is to be made, the request is made by transmitting the DAO data and the personal AI group in S680. DAO data is the functional element of the organization that you want to undergo simulation reinforcement learning. For example, in the case of a furniture assembly and sales company, there are material procurement elements, assembly elements, advertising elements, and sales elements. The personal AI group is the personal AI of people who actually engage in element integration DAO in the real world 47. If there is a worker who does not have personal AI, or if the worker has not yet been determined, simulation reinforcement learning can be used as explained based on S561, S565-S568, S562, and S573-S575 above. Generate and prepare a personal AI that matches the target element integrated DAO.

In the simulation reinforcement learning preparation process, the CPU 10 of the mirror world server 46 determines in S683 whether or not there is a request for simulation reinforcement learning, and if there is no request, returns. If it is determined that there is a request for simulation reinforcement learning, in S684, the personal AI group is copied and deployed in the element integrated DAO to generate an element integrated DAO digital twin, and registered in the mirror world 51 as a simulation target. .

The state is shown in FIG. The digital twin 78 of the element integrated DAO 77 of the real world 47 is registered in the mirror world 51. The element integrated DAO digital twin 78 shown in FIG. 45 is, for example, an element integrated DAO digital twin 78 of a furniture assembly and sales company, and has functional elements such as material procurement, assembly, advertising, and sales, and is engaged in each functional element. The identification numbers of the personal AI groups of those who do so are stored in the mirror world server 46.

For this element-integrated DAO digital twin 78, an optimal incentive design is derived through simulation reinforcement learning. A schematic system of reinforcement learning (machine learning) is shown in FIG.

Referring to FIG. 46, in the element integrated DAO digital twin 78, a material procurement element agent 80, a personal AI group 84 in charge of material procurement, and an assembly element agent correspond to each functional element of material procurement, assembly, advertising, and sales. 81 and personal AI group 85 in charge of assembly, advertising element agent 82 and personal AI group 86 in charge of advertising, material procurement element agent 80 and personal AI group 84 in charge of material procurement, sales element agent 83 and personal AI group 87 in charge of sales. is being formed. A supervising agent 79 supervises each element agent 80 to 83.

The personal AI group 84 in charge of material procurement performs actions a11, a12, . The state S1 of the digital twin group 88 of the material supplier for the action a1 is input to the material procurement element agent 80 and the personal AI group 84 in charge of material procurement. This state S1 is, for example, the number of materials requested and the response price to the action a1 of price negotiation. Note that each act a11, a12, ... a1n of the personal AI group 84 in charge of material procurement is also input to the material procurement element agent 80, and the group of acts a1 is also input to the supervising agent 79 and the material procurement element agent 80. be done.

The material procurement element agent 80 calculates the performance p1 by the personal AI group 84 in charge of material procurement based on the action a1 and the state S1, and transmits the performance p1 to the supervising agent 79. The supervising agent 79 determines the remuneration r1 based on its performance p1, and transmits the remuneration r1 to the material procurement element agent 80. The material procurement element agent 80 determines the reward distribution rate based on each action a11, a12, ... a1n of the personal AI group 84 in charge of material procurement, and distributes the reward r1 to the personal AI group 84 in charge of material procurement according to the reward distribution rate. Distribute to AI84.

The personal AI group 85 in charge of assembly performs actions a21, a22, . The state S2 of the digital twin group 89 of the assembly equipment for the action a2 is input to the assembly element agent 81 and the personal AI group 85 in charge of assembly. This state S1 is, for example, the power consumption of the assembly equipment digital twin group 89 and the total working hours of the personal AI group 85 in charge of assembly who worked on the assembly equipment digital twin group 89. Incidentally, each act a21, a22, .

The assembly element agent 81 calculates the performance p2 by the personal AI group 85 in charge of assembly based on the action a2 and the state S2, and transmits the performance p2 to the supervising agent 79. The supervising agent 79 determines a reward r2 based on its performance p2, and sends the reward r2 to the assembly element agent 81. The assembly element agent 81 determines the reward distribution rate based on each act a21, a22, ... a2i of the personal AI group 85 in charge of assembly, and distributes the reward r2 to the personal AI 85 in charge of assembly according to the reward distribution rate. .

The personal AI group 86 in charge of advertising performs acts a51, a52, . The state S5 of the consumer's personal AI group 90 for the action a5 is input to the promotion element agent 82 and the personal AI group 86 in charge of promotion. This state S5 includes, for example, whether or not the consumer has purchased the product and the purchase price in response to the product recommendation act a5 to the consumer. Incidentally, each act a51, a52, .

The promotion element agent 82 calculates the performance p5 by the personal AI group 86 in charge of promotion based on the action a5 and the state S5, and transmits the performance p5 to the supervising agent 79. The supervising agent 79 determines a reward r5 based on its performance p5 and sends the reward r5 to the advertising element agent 82. The promotion element agent 82 determines the reward distribution rate based on each act a51, a52, ... a5j of the personal AI group 86 in charge of promotion, and distributes the reward r5 to the personal AI 86 in charge of promotion according to the reward distribution rate. .

The personal AI group 87 in charge of sales performs acts a91, a92, . The state S9 of the digital twin group 91 of the store and the consumer for the action a9 is input to the sales element agent 83 and the personal AI group 87 of the sales person. This state S9 is, for example, the total sales amount at the store. The sales element agent 83 also receives the status S5 of the digital twin 88 of the material supplier for the above action a5. Incidentally, each act a91, a92, .

The sales element agent 83 calculates the performance p9 by the personal AI group 87 in charge of sales based on the action a9 and the states S5 and S9, and transmits the performance p9 to the supervising agent 79. The supervising agent 79 determines a reward r9 based on its performance p9, and sends the reward r9 to the sales element agent 83. The sales element agent 83 determines the reward distribution rate based on each act a91, a92, ... a9m of the personal AI group 87 in charge of sales, and distributes the reward r9 to the personal AI 87 in charge of sales according to the reward distribution rate. .

A method of calculating each of the performances p1 to p9 and each reward distribution rate will be explained based on FIGS. 47(A), (B), and (C) and FIG. 48(A). The material procurement element agent 80 stores the algorithm for calculating the performance p1 and the distribution rate as knowledge. The algorithm for calculating the performance p1 and the distribution ratio will be explained based on FIG. 47(A). The material procurement element agent 80 sets the material purchase amount u and the current inventory quantity z for this time (from the previous YES point in S689 to the current YES point) as the state S1, and calculates the performance p1={2 (average purchase amount Calculate performance p1 using the formula: /u) + (z/average inventory)}/3. The average purchase price is the average purchase price of materials from the start of simulation reinforcement learning to the present. The average number of materials in stock is the average number of materials purchased from the start of simulation reinforcement learning to the present. As a result of this calculation formula, the performance p1 increases as the current material purchase price u decreases, and the performance p1 decreases as the current inventory quantity z increases.

In addition, when p1≧1, the reward distribution rate is calculated in proportion to the degree of approval for the collective action a1, and conversely, when p1<1, the reward distribution rate is calculated in inverse proportion to the degree of approval for the collective action a1. calculate. This includes not only those that are proportional or inversely proportional to the first power of the "degree of approval" but also those that are proportional or inversely proportional to the nth power of the "degree of approval", and the material procurement element agent 80 determines the optimal proportional function or inverse proportional function. is determined by performing reinforcement learning (machine learning). In addition, regarding the "degree of approval", the personal AI that proposed the collective action a1 itself has the highest approval level, and the material procurement element agent 80 Determine (calculate) the degree of AI approval.

The algorithm for calculating the performance p2 and the distribution ratio, which is stored as knowledge by the assembly element agent 81, will be explained based on FIG. 47(B). The assembly element agent 81 sets the current power consumption e of the assembly equipment (from the previous YES time to the current YES time in S710) and the current assembly worker total working time t as a state S2, and sets the performance p2={ Performance p2 is calculated using the following formula: (average power consumption/e) + (average total working hours/t)}/2. The current assembly worker total working time t is the total working time of the personal AI group 85 in charge of assembly who worked on the assembly equipment digital twin group 89 this time. The average power consumption is the average power consumption of the assembly equipment from the start of simulation reinforcement learning to the present. The average total working hours is the average of the total working hours of assembly workers from the start of simulation reinforcement learning to the present. As a result of this calculation formula, the performance p2 increases as the current power consumption e of the assembly equipment decreases, and the performance p2 decreases as the current total working time t of the assembly workers increases.

In addition, when p2≧1, the reward distribution rate is calculated in proportion to the degree of approval for the collective action a2, and conversely, when p2<1, the reward distribution rate is calculated in inverse proportion to the degree of approval for the collective action a2. calculate. This includes not only those that are proportional or inversely proportional to the first power of the "degree of approval," but also those that are proportional or inversely proportional to the nth power of the "degree of approval," and the assembly element agent 81 determines the optimal proportional function or inverse proportional function. Obtained by performing reinforcement learning (machine learning). In addition, the "degree of approval" is that the personal AI that proposed the collective action a2 itself has the highest approval level, and the assembly element agent 81 Determine (calculate) the degree of approval.

The algorithm for calculating the performance p5 and the distribution rate, which is stored as knowledge by the advertising element agent 82, will be explained based on FIG. 47(C). The advertising element agent 82 sets the total purchase amount k of the recommended consumer personal AI this time (from the previous YES time in S728 to the current YES time point) as a state S5, and sets the performance p5 = k/recommended consumer personal AI. Calculate performance p5 using the formula for AI's average total purchase price K. The average total purchase amount K is the average total purchase amount of the recommended consumer personal AI group 90 from the start of simulation reinforcement learning to the present. As a result of this calculation formula, the performance p5 increases as the current total purchase price k of the recommended consumer personal AI increases.

In addition, when p5≧1, the reward distribution rate is calculated in proportion to the degree of approval for the collective action a5, and conversely, when p5<1, the reward distribution rate is calculated in inverse proportion to the degree of approval for the collective action a5. calculate. This includes not only those that are proportional or inversely proportional to the first power of the "degree of approval," but also those that are proportional or inversely proportional to the nth power of the "degree of approval," and the advertising element agent 82 determines the optimal proportional function or inverse proportional function. Obtained by performing reinforcement learning (machine learning). In addition, the "degree of approval" is that the personal AI that proposed the collective action a5 itself has the highest approval level, and the advertising element agent 82 is Determine (calculate) the degree of approval.

The algorithm for calculating the performance p5 and the distribution rate, which is stored as knowledge by the sales element agent 83, will be explained based on FIG. 48(A). The sales element agent 83 sets the total sales amount h and the average total sales amount H at the store this time (from the previous YES time in S749 to the current YES time point) as a state S9, and sets this state S9 and the above-mentioned state S5. Performance p9 is calculated using the following formula: performance p9=(h-k)/(HK). The average total sales amount H is the average of the total sales amount at the store from the start of simulation reinforcement learning to the present. In addition, k is the current total purchase price of the recommended consumer personal AI, and K is the average of the total purchase price of the recommended consumer personal AI group 90 from the start of simulation reinforcement learning to the present (Fig. 47(C) and its explanation). As a result of this calculation formula, the performance p9 increases as the value obtained by subtracting the current total purchase amount k of the consumer personal AI that made the recommendation from the current total sales amount h at the store increases. The current total purchase price k of the recommended consumer personal AI is the credit of the personal AI group 86 in charge of advertising, and the credit of only the personal AI group 87 in charge of sales is the recommendation based on the current total sales price h at the store. This is because it is the value obtained by subtracting the current total purchase price k of the consumer personal AI.

In addition, when p9≧1, the reward distribution rate is calculated in proportion to the degree of approval for the collective action a9, and conversely, when p9<1, the reward distribution rate is calculated in inverse proportion to the degree of approval for the collective action a9. calculate. This includes not only those that are proportional or inversely proportional to the first power of the "degree of approval" but also those that are proportional or inversely proportional to the nth power of the "degree of approval", and the advertising element agent 83 determines the optimal proportional function or inverse proportional function. Obtained by performing reinforcement learning (machine learning). In addition, the "degree of approval" is that the personal AI that proposed the collective action a9 itself has the highest approval level, and the sales element agent 83 Determine (calculate) the degree of approval.

Next, the remuneration table 92 stored as knowledge by the supervising agent 79 will be explained based on FIG. 48(B). This remuneration table 92 stores a calculation formula for remuneration distributed by the supervising agent 79 to each element agent 80 to 83. The reward to be distributed is calculated by coefficient x (current period's profit) x (performance sent from the target element agent) ÷ (total performance sent from all element agents). Here, the "current period" refers to the period from the previous YES time in S675 to the current YES time point.

Specifically, the reward r1 to be distributed to the material procurement element agent 80 is r1=A1・Lt・p1/(p1+p2+p5+p9). The reward r2 distributed to the assembly element agent 81 is r2=A2・Lt・p2/(p1+p2+p5+p9). The reward r5 to be distributed to the advertising element agent 82 is r5=A5・Lt・p5/(p1+p2+p5+p9). The remuneration r9 to be distributed to the sales element agent 83 is r9=A9・Lt・p9/(p1+p2+p5+p9). Here, Lt is the current period's profit, and A1, A2, A5, and A9 are coefficients as actions determined by the supervising agent 79.

Next, the specific content of the simulation reinforcement learning process shown in S680 will be explained based on FIG. 49. S687 executes the supervising agent reinforcement learning process, S688 executes the material procurement element agent reinforcement learning process, S689 executes the assembly element agent reinforcement learning process, S690 executes the promotion element agent reinforcement learning process, and S691 executes the reinforcement learning process for the advertising element agent. Sales agent reinforcement learning processing is executed, material procurement personnel personal AI reinforcement learning processing is executed in S692, assembly personnel personal AI reinforcement learning processing is executed in S693, advertising personnel personal AI reinforcement learning processing is executed in S694, and S695 Executes the sales person's personal AI reinforcement learning process.

The details of the supervising agent reinforcement learning process shown in S687 will be explained based on FIG. 50. In S699, the supervising agent 79 determines whether or not each performance p sent from each element agent 80 to 83 has been received. If it has not been received, the control proceeds to S671, but if it is determined that it has been received, each received performance p is stored in S670.

Next, in S671, it is determined whether or not each of the actions a1 to a9 has been received. If not, control proceeds to S673. If it is determined that the received actions have been received, each of the received actions a1 to a9 is stored in S672. Next, in S673, it is determined whether or not there is an input of status S9 sent from the store and consumer's personal AI group 91, and if there is no input, control proceeds to S675. If it is determined that there has been an input, sales = ΣS9 is calculated in S674.

Next, in S675, it is determined whether the remuneration calculation time has come, and if it has not, the control proceeds to S677, but if it is determined that it has come, the remuneration table 92 is referred to in S676. Rewards r1, r2, r3, r5, r9 are calculated and sent to the corresponding element agents 80 to 83.

Next, in S677, it is determined whether it is time to update reinforcement learning (machine learning), and if it is not, the process returns. If it is determined that it is time for renewal, in S687, the current term's profit Lt = sales - expenses is calculated. Next, in S679, each reward r1, r2, r5, r9 is calculated and distributed to the corresponding element agents 80-83.

Next, in S680, the remuneration R of the supervising agent 79 is calculated from the profit Lt. This reward R is proportional to profit Lt. Next, in S681, based on the reward R, actions (coefficients) A1, A2, A5, A9 according to the optimal policy π ^* are determined by TD learning. Next, in S682, A1, A2, A5, and A9 of the reward table 92 are updated to the actions (coefficients) A1, A2, A5, and A9 obtained in S681. As a result, the supervising agent 79 learns the actions (coefficients) A1, A2, A5, and A9 that maximize the profit Lt.

Details of the material procurement element agent reinforcement learning process shown in S688 will be explained based on FIG. 51. In S684, the material procurement element agent 80 performs information collection processing using a crawler. A crawler is a program that periodically acquires documents, images, etc. on the web and automatically creates a database. Also called "bot", "spider", "robot", etc.

Details of the information collection process by this crawler will be explained based on FIG. 52(A). In S702, the material procurement element agent 80 receives the information collected by the crawler while patrolling the Internet. Next, in S703, the received information is stored in the material procurement DB 93.

The information stored in the material procurement DB 93 is shown in FIG. 52(B). As shown in the figure, the material procurement DB 93 stores various information necessary for material procurement operations, such as economic information, social information, weather information, inventory information, market information, ... supplier information. .

Returning to FIG. 51, the material procurement element agent 80 determines whether or not it has received the actions a11, a12, . . . a1n from the personal AI group 84 in S685. If it has not been received, the control proceeds to S687, but if it is determined that it has been received, the received acts a11, a12, . . . a1n are stored in S686. In S687, it is determined whether the status S1 from the material supplier's digital twin group 88 has been received, and if not, control proceeds to S689. If it is determined that it has been received, the received S1 is stored in S688.

In S689, it is determined whether it is time to calculate performance p1, and if it is not time to calculate performance p1, control proceeds to S692. If it is determined that it is time to calculate, in S690, performance p1={2(average purchase price/u)+(z/average quantity in stock)}/3 is calculated. The performance p1 is transmitted to the supervising agent 79 (S691).

In S692, it is determined whether the reward r1 sent from the general agent 79 has been received, and if it has not been received, the process returns. If it is determined that the information has been received, the reward distribution rate is calculated based on the reward distribution rate algorithm shown in FIG. 47(A) (S693). In S694, each reward r11, r12...r1n is calculated by multiplying the reward by each distribution rate, and in S695, each reward r11, r12...r1n is given to each material procurement personal AI group 84. In S696, an action (proportional function or inverse proportional function) according to the optimal policy π ^* is determined by TD learning based on the received reward r1. In S697, the proportional function or inverse proportional function is updated to the one determined in S696. As a result, the material procurement element agent 80 will learn the proportional function or inverse proportional function that maximizes the performance p1.

Next, details of the assembly element agent reinforcement learning process shown in S689 will be explained based on FIG. 53. In S706, the assembly element agent 81 determines whether or not it has received the actions a21, a22, . . . a2n from the personal AI group 85. If it has not been received, the control proceeds to S708, but if it is determined that it has been received, the received actions a21, a22, . . . a2n are stored in S707. In S708, it is determined whether or not the status S2 from the digital twin group 89 of the assembly equipment has been received, and if the status S2 has not been received, the control proceeds to S710. If it is determined that the message has been received, the received state S2 is stored in S709.

In S710, it is determined whether it is time to calculate performance p2, and if it is not time to calculate performance p2, control proceeds to S713. If it is determined that it is time to calculate, in S711, performance p2={(average power consumption/e)+(average total working hours/t)}/2 is calculated. The performance p2 is transmitted to the supervising agent 79 (S712).

In S713, it is determined whether the reward r2 sent from the general agent 79 has been received, and if it has not been received, the process returns. If it is determined that the information has been received, the reward distribution rate is calculated based on the reward distribution rate algorithm shown in FIG. 47(B) (S714). In S715, each reward r11, r12, . . . r1n is calculated by multiplying the reward by each distribution rate, and in S695, each reward r21, r22, . In S717, an action (proportional function or inverse proportional function) according to the optimal policy π ^* is determined by TD learning based on the received reward r2. In S718, the proportional function or inverse proportional function is updated to the one determined in S717. As a result, the assembly element agent 81 learns the proportional function or inverse proportional function that maximizes the performance p2.

Details of the promotion element agent reinforcement learning process shown in S690 will be explained based on FIG. 54. In S723, the advertising element agent 83 performs information collection processing using a crawler. The details of this process will be explained based on FIG. 55(A). The advertisement element agent 83 receives the information collected by the crawler while traveling on the Internet in S740, and stores the received information in the advertisement DB 94 in S741.

The collected data stored in the advertisement DB 94 is shown in FIG. 55(B). The advertisement DB 94 stores various behavioral data of consumers such as Taro, Jiro, Hanako, etc. For example, in the case of Taro, it is determined that there is a high possibility that he will purchase furniture based on the information that he has ordered a detached house, and the furniture is advertised to Taro. In Jiro's case, based on the information that the couple is purchasing furniture, they judge that there is a high possibility that they will purchase furniture for their new home because they are getting married soon, so they advertise the furniture to Jiro.

Returning to FIG. 54, the advertising element agent 83 determines whether or not it has received an action from each personal AI group 86 in S742. If it has not been received, control proceeds to S726, but if it is determined that it has been received, each received action is stored in S725. In S726, it is determined whether the status S5 sent from the consumer's personal AI group 90 has been received, and if it has not been received yet, control proceeds to S728. If it is determined that it has been received, the received state S5 is stored in S727.

In S728, it is determined whether or not it is time to calculate the performance p5, and if it is not the time to calculate the performance p5, control proceeds to S731. If it is determined that it is time to calculate the performance p5, then in S729, the performance p5=k/average total purchase price K of the recommended consumer personal AI is calculated. Next, in S730, the performance P5 is sent to the supervising agent 79.

In S731, it is determined whether the reward r5 sent from the general agent 79 has been received, and if it has not been received yet, the process returns. If it is determined that the information has been received, the reward distribution rate is calculated based on the reward distribution rate algorithm shown in FIG. 47(C) (S732). In S733, each reward r51, r52, . In S735, an action (proportional function or inverse proportional function) according to the optimal policy π ^* is determined by TD learning based on the received reward r5. In S736, the proportional function or inverse proportional function is updated to the one determined in S735. As a result, the promotional element agent 82 will learn the proportional or inverse proportional function that maximizes the performance p5.

Next, details of the sales element agent reinforcement learning process shown in S691 will be explained based on FIG. 56. In S744, the sales element agent 84 performs information collection processing using a crawler. The details of this process will be explained based on FIG. 57(A). In S760, the sales element agent 84 receives the information collected by the crawler while patrolling the Internet, and in S761 stores the received information in the sales DB 95. Furthermore, in S762, POS data at the store is stored in the sales DB 95.

The collected data stored in the sales DB 95 is shown in FIG. 57(B). The sales DB 95 stores various data such as weather data and POS data. "By date" in weather information is a concept that includes days of the week. Based on weather information (weather and temperature data by date and time) and POS data (sales product data by date and time), for example, it is possible to rearrange displayed products in consideration of the day of the week, time, and weather conditions. can.

Returning to FIG. 56, the sales element agent 84 determines whether or not an action from each personal AI group 87 has been received in S745. If it has not been received, control proceeds to S747, but if it is determined that it has been received, each received action is stored in S746. In S747, it is determined whether the status S9 sent from the store and consumer's personal AI group 91 has been received, and if it has not been received yet, the control proceeds to S749. If it is determined that it has been received, the received state S9 is stored in S748.

In S749, it is determined whether or not it is time to calculate performance p9, and if it is not time to calculate performance p9, control proceeds to S752. If it is determined that it is time to calculate the performance p9, the performance p9=(h−k)/(HK) is calculated in S750. Next, in S751, the performance p9 is sent to the supervising agent 79.

In S752, it is determined whether the reward r9 sent from the general agent 79 has been received, and if it has not been received yet, the process returns. If it is determined that the information has been received, the reward distribution rate is calculated based on the reward distribution rate algorithm shown in FIG. 48(A) (S753). In S754, each reward r91, r92, . In S756, an action (proportional function or inverse proportional function) according to the optimal policy π ^* is determined by TD learning based on the received reward r9. In S757, the proportional function or inverse proportional function is updated to the one determined in S756. As a result, the sales element agent 83 learns the proportional function or inverse proportional function that maximizes the performance p9.

Next, details of the material procurement personal AI reinforcement learning process shown in S692 will be explained based on FIG. 58(A). In S765, the personal AI group 84 in charge of material procurement determines whether or not to negotiate with the digital twin group 88 of material suppliers, and if not, control proceeds to S770. If it is determined to negotiate, in S766, the data stored in the financing DB 93 is viewed, action a1 is determined while holding an internal meeting with reference to the stored data (S767), and negotiation is performed with the digital twin group 88 of material suppliers. (S768). In S769, it is determined whether or not the negotiation has ended, and if it has not ended yet, the process returns to S766 and repeats the loop of S767→S768→S769→S766. When it is determined in S769 that the negotiation has ended, control proceeds to S770.

In S770, it is determined whether the rewards r11, r12, . . . r1n have been received from the material procurement element agent 80, and if not, the process returns. If it is determined that the reward has been received, in S771, based on the received reward, actions (a11, a12...a1n) according to the optimal policy π ^* are determined by TD learning. This action a1i includes moving to another company's DAO digital twin (for example, the DAO digital twin 59 of ABC Co., Ltd. in FIG. 45) if the received remuneration r1i is not satisfactory. As a result of this reinforcement learning, each personal AI in charge of material procurement will learn actions that increase the aforementioned performance p1.

Next, details of the assembling personal AI reinforcement learning process shown in S693 will be explained based on FIG. 58(B). In S775, the assembly personal AI group 85 determines whether or not to hold an internal meeting, and if not, the control proceeds to S779. If it is determined that a meeting is to be held, in S776, each personal AI in charge of assembly determines the action a2 while holding an internal meeting. Next, in S777, the assembly equipment digital twin group 89 is trial run according to the action a2, and the validity of the action a2 is verified. In S778, it is determined whether the meeting has ended, and if it has not ended, the process returns to S776 and repeats the loop of S777→S778→S776. If the result of the trial run in S777 is that the action a2 is appropriate, it is determined in S778 that the meeting is over, and the control proceeds to S779.

In S779, it is determined whether or not rewards r21, r22, . . . r2n have been received from the assembly element agent 81, and if not, the process returns. If it is determined that a reward has been received, in S780, actions (a21, a22, . . . a2i) according to the optimal policy π ^* are determined by TD learning based on the received reward. This action a2i includes moving to another company's DAO digital twin (for example, the DAO digital twin 59 of ABC Co., Ltd. in FIG. 45) if the received remuneration r2i is not satisfactory. As a result of this reinforcement learning, each of the personal AIs in charge of assembly will learn actions that increase the aforementioned performance p2.

Next, details of the advertising personal AI reinforcement learning process shown in S694 will be explained based on FIG. 59(A). The advertising personal AI group 86 determines whether or not to hold an internal meeting in S784, and if not, control proceeds to S789. If it is determined that a meeting is to be held, in S785, each personal AI in charge of advertising determines the action a5 while holding an internal meeting. Next, in S787, action a2 to the consumer is executed. In S788, it is determined whether the meeting has ended or not. If the meeting has not ended yet, the process returns to S785 and repeats the loop of S786→S787→S788. When it is determined in S788 that the meeting has ended, control proceeds to S789.

In S789, it is determined whether the rewards r51, r52, . . . r5j have been received from the advertising element agent 82, and if not, the process returns. If it is determined that a reward has been received, in S790, actions (a51, a52 . . . a5j) according to the optimal policy π ^* are determined by TD learning based on the received reward. This action a5i includes moving (changing jobs) to another company DAO digital twin (for example, ABC Corporation's DAO digital twin 59 in FIG. 45) if the received remuneration r5i is not satisfactory. As a result of this reinforcement learning, each of the advertising personal AIs will learn actions that increase the aforementioned performance p5.

Next, details of the salesperson personal AI reinforcement learning process shown in S695 will be explained based on FIG. 59(B). In S791, the sales personal AI group 87 determines whether an internal meeting is to be held, and if not, control proceeds to S795. If it is determined that a meeting is to be held, in S792, each sales person's personal AI determines the action a9 while holding an internal meeting. Next, in S793, it is determined whether the meeting has ended or not. If the meeting has not ended yet, the process returns to S792 and repeats the loop of S792→S793→S792. When it is determined in S793 that the meeting has ended, control proceeds to S794. In S794, the action determined in the above meeting is performed with respect to the consumer and the store.

Next, in S795, it is determined whether or not the rewards r91, r92, . . . r9m have been received from the sales element agent 83, and if not, the process returns. If it is determined that a reward has been received, in S796, an action (a91, a92 . . . a9m) according to the optimal policy π* is determined by TD learning based on the received reward. This action a9i includes moving to another company's DAO digital twin (for example, the DAO digital twin 59 of ABC Co., Ltd. in FIG. 45) if the received remuneration r9i is not satisfactory. As a result of this reinforcement learning, each of the sales personal AIs will learn actions that increase the aforementioned performance p9.

The element integrated DAO that has undergone simulation reinforcement learning is operated as an actual organization in the real world 47. At that stage, actual humans (users) in the real world will be in charge of "each personal AI group 84 to 87" in FIG. 46. At that time, each personal AI group 84 to 87 that has undergone simulation reinforcement learning becomes a consultant to the actual human being (user) and provides the knowledge, experience, and know-how obtained through simulation reinforcement learning to the actual human being (user). can.

The construction of an element integrated DAO explained above shows that the entire organization such as a company, NPO, or local government is created by combining element DAOs according to functions. ) may be constructed using element DAO.

The aforementioned programs that run on the user terminal 16, etc. and various servers may be downloaded and installed from a predetermined website, etc.; A person who purchases the CD-ROM 99 or the like may install the program on the user terminal 16 and various servers (see FIG. 60).
[Modified example]

(1) For example, the names of Taro, Jiro, Sakura, Saburo, etc. in the digital twin data shown in Figure 29 are pseudonyms (anonymous) from the perspective of protecting personal information, and although it is possible to identify them as the same person, they are not specific. It may be possible to make it impossible to identify individuals. In that case, the AI identification number or blockchain address may be used as a pseudonym (anonymous). Similarly, digital twins such as ABC Co., Ltd. use pseudonyms (anonymous) for the company name (organization name), and although it is possible to identify the same company (same organization), it is difficult to identify a specific company (organization). You can also do this. Furthermore, for a human being, a plurality of digital twins may be prepared by a plurality of personal AIs for one person. Furthermore, one digital twin of one person may be composed of a set of multiple personal AIs (for example, a set of professional personal AIs in various fields, etc.).

(2) In Figures 34 to 59, we have explained a system that derives the optimal solution for incentive design in a DAO by performing simulations using a multi-service DAO where multiple types of services exist as an example, but it is not limited to multi-service DAOs. , it may be a system that derives an optimal solution for incentive design through simulation for a DAO in which only one type of service exists.

(3) In FIG. 35, a trained persona agent group is generated for each persona, but each personal AI of the user group belonging to each persona is You may choose one from among them and use it as a group of persona agents. In this case, it is necessary to inquire of the user group belonging to each persona whether or not personal AI may be used in the simulation, and to obtain consent to use the personal AI. Each personal AI of the user group for which consent has been obtained is copied and used in the simulation, and the learned personal AI group after the simulation is completed is sent to each corresponding user. When each user who receives the learned personal AI determines that the learned personal AI is useful (necessary), the learned personal AI is overwritten and saved on the existing personal AI. Note that both the existing personal AI and the learned personal AI may be stored together and used separately as necessary.

(4) As multi-agent reinforcement learning, a supervising agent (master agent) responsible for overall optimization allocates rewards to each agent, and the supervising agent itself also performs reinforcement learning to converge the reward distribution behavior to the optimal one. The master agent method was demonstrated. However, multi-agent reinforcement learning is not limited to this, for example, D-learning that can converge to an optimal solution under a Markov decision process, or Bucket Brigade as a reinforcement learning algorithm in a classifier system. or Profit Sharing may be used.

(5) Simulations using digital twins are not limited to digital twins of people or organizations made up of people (for example, corporations, NPOs, etc.). For example, for an object such as an AI-equipped machine or electrical product (e.g., an AI-equipped vacuum cleaner), a digital twin of the environment in which the object operates (e.g., the room in the user's home where an independently moving AI-equipped vacuum cleaner operates) is generated in cyberspace, the AI installed on the object is simulated in advance in the digital twin environment, subjected to reinforcement learning (machine learning), and the customized (personalized) object equipped with the learned AI is applied to the object. It may also be provided to users who do so.

Since it is possible to eliminate as much as possible the dilemma in which guaranteeing the authenticity of recorded information and guaranteeing the right to delete that information are contradictory, it can be used for information recording systems that cannot be erased, such as blockchain.

1 Internet 2 Private chain 3 Consortium chain 4 Public chain 12 HDD
16 User terminal 19 Node 30 Key registration center 32 Key DB
46 Mirror World Server 51 Mirror World 52 Earth Digital Twin 53 Japan Digital Twin 54 Town Digital Twin 57 Taro Digital Twin 58 Taro Family Digital Twin 59 ABC Digital Twin Co., Ltd. 61 DAO Agent 72 Token 78 DAO Digital Twin 79 Supervising Agent.

Claims (2)

A computer system that processes personal information,
a personal information storage means storing encrypted personal information obtained by encrypting the user's personal information;
search enablement means for making the encrypted personal information searchable while encrypted;
If the desired personal information is found as a result of the search using the search enablement means, a transaction is concluded between the person requesting the personal information and the person providing the personal information, and the personal information is transferred. A transaction execution means provided to those who wish to do so;
The search enablement means includes:
an encrypted index recording means for recording on a blockchain an encrypted index that has been subjected to an encryption process of encrypting an index for keyword searching the encrypted personal information using a common key;
a search means for searching the encrypted index using an encrypted search query that has been subjected to an encryption process of encrypting a search query that is a keyword used for the search with the common key,
The transaction execution means is
A means for making personal information provided to the applicant available for viewing by the applicant;
a smart contract processing means that uses a smart contract to transact the personal information when transaction conditions according to the intention of the person providing the personal information are satisfied;
A personal information processing system further comprising recording means for recording transactions performed by the smart contract processing means on a blockchain. A method of providing personal information using a computer,
storing encrypted personal information obtained by encrypting the user's personal information;
a search enablement step for making the encrypted personal information searchable while encrypted;
If the desired personal information is found as a result of the search in the search enablement step, a transaction is concluded between the person requesting the personal information and the person providing the personal information, and the personal information is transferred. A transaction execution step provided to a person who wishes to do so;
The searchability step includes:
an encrypted index recording step of recording on a blockchain an encrypted index that has been subjected to an encryption process of encrypting an index for keyword searching the encrypted personal information using a common key;
a search step of searching the encrypted index using an encrypted search query that has been subjected to an encryption process of encrypting a search query that is a keyword used for the search with the common key,
The transaction execution step includes:
a viewing enablement step for allowing the applicant to view the personal information provided to the applicant;
a smart contract processing step for performing a transaction of the personal information using a smart contract when transaction conditions according to the intention of the person providing the personal information are satisfied;
A method for providing personal information, further comprising a recording step of recording transactions performed by the smart contract processing step on a blockchain.

Images