Artificial intelligence management method and device, computer, storage medium and product

Abstract

The embodiment of the invention discloses an artificial intelligence management method and device, a computer, a storage medium and a product, and the method comprises the steps: receiving a model task issued by a task issuing node, and sharing the model task to N artificial intelligence models; n is a positive integer; the N artificial intelligence models are independently trained; obtaining candidate task execution results of the N artificial intelligence models for the model task, and sending the N candidate task execution results to a result evaluation node, so that the result evaluation node performs result detection on the N candidate task execution results; and obtaining a result evaluation node, and for a target task execution result consensus by the N candidate task execution results, feeding back the target task execution result to the task release node. According to the invention, the security of the artificial intelligence model can be improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to an artificial intelligence management method, device, computer, storage medium and product. Background Art

[0002] With the parameter scale of large artificial intelligence (AI) models climbing into the hundreds of billions, the intelligence demonstrated by AI has shown an "emergence" phenomenon in a form that humans cannot understand. This phenomenon is likely to lead to the development of AI exceeding human control, thus bringing irreversible risks to human society. At the same time, since AI users may try to hypnotize AI in various ways to bypass the basic behavior norms set for AI, such as the recently famous "grandmother attack" on large AI models, which can output activation codes of the system or production manuals of dangerous tools, etc. as bedtime stories by asking AI to play the role of its own grandmother, that is, by deceiving AI to output content that violates laws and regulations. Therefore, the restraint of AI has become an urgent matter. Summary of the Invention

[0003] Embodiments of this application provide an artificial intelligence management method, device, computer, storage medium and product, which can improve the security of artificial intelligence models.

[0004] On the one hand, embodiments of this application provide an artificial intelligence management method, which includes:

[0005] Receiving a model task published by a task publishing node, and sharing the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained;

[0006] Obtaining candidate task execution results of the N artificial intelligence models for the model task respectively, and sending the N candidate task execution results to a result evaluation node, so that the result evaluation node conducts result detection on the N candidate task execution results;

[0007] Obtaining the target task execution result reached a consensus by the result evaluation node for the N candidate task execution results, and feeding back the target task execution result to the task publishing node.

[0008] On the one hand, embodiments of this application provide an artificial intelligence management device, which includes:

[0009] A task sharing module, configured to receive a model task published by a task publishing node, and share the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained;

[0010] A result acquisition module for acquiring the candidate task execution results of N artificial intelligence models for a model task respectively;

[0011] A result sending module for sending the N candidate task execution results to a result evaluation node so that the result evaluation node can perform result detection on the N candidate task execution results;

[0012] A result determination module for acquiring the target task execution result reached a consensus by the result evaluation node for the N candidate task execution results;

[0013] A result feedback module for feeding back the target task execution result to the task publishing node.

[0014] Among them, the result determination module includes:

[0015] A vote counting unit for acquiring the result vote information of the result evaluation node for the N candidate task execution results respectively, and determining the result statistical values corresponding to the N candidate task execution results respectively based on the result vote information corresponding to the N candidate task execution results;

[0016] A result determination unit for determining the target task execution result based on the result statistical values corresponding to the N candidate task execution results respectively.

[0017] Among them, the vote counting unit includes:

[0018] A result vote sub-unit for, in the i-th round of result voting process, acquiring the (i - 1)-th result vote information generated by the result evaluation node in the (i - 1)-th round of result voting process, generating the i-th result vote information according to the (i - 1)-th result vote information, and sending the i-th result vote information generated by the first blockchain node to the second blockchain node; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node; the second blockchain node is the blockchain node other than the first blockchain node in the result evaluation node; i is a positive integer;

[0019] A vote generation sub-unit for, if the i-th round of result voting process does not meet the vote iteration condition, acquiring the i-th result vote information corresponding to the second blockchain node, and generating the (i + 1)-th result vote information of the first blockchain node based on the i-th result vote information of the result evaluation node;

[0020] A result statistics sub-unit for, if the i-th round of result voting process meets the vote iteration condition, acquiring the P rounds of result vote information of the result evaluation node, and statistically calculating the result statistical values corresponding to the N candidate task execution results respectively based on the P rounds of result vote information of the result evaluation node; P is a positive integer, and P is the number of rounds of the result voting process indicated by the vote iteration condition.

[0021] Among them, the vote statistics unit includes:

[0022] A result scoring sub-unit, which is used to score the execution results of N candidate tasks to obtain the first result quality scores corresponding to the execution results of the N candidate tasks respectively, and send the first result quality scores corresponding to the execution results of the N candidate tasks respectively to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model tasks issued by the task publishing node;

[0023] A score statistics sub-unit, which is used to obtain the second result quality scores of the second blockchain node for the execution results of the N candidate tasks respectively, and determine the result statistical value of the execution result of the candidate task based on the first result quality score and the second result quality score of each candidate task execution result; the first result quality score and the second result quality score of each candidate task execution result form the result voting information of the execution result of the candidate task.

[0024] Among them, the vote statistics unit includes:

[0025] A vote statistics sub-unit, which is used to obtain the first result voting information of the result evaluation node for the execution results of the N candidate tasks respectively, and determine the first result statistical value corresponding to the execution results of the N candidate tasks respectively based on the first result voting information corresponding to the execution results of the N candidate tasks respectively;

[0026] A result sorting sub-unit, which is used to sort the execution results of the N candidate tasks based on the first result statistical values corresponding to the execution results of the N candidate tasks respectively to obtain a result sequence;

[0027] A determination and invocation sub-unit, which is used to, if the difference degree between the first result voting information of the first two candidate task execution results in the result sequence is less than the data difference threshold, then based on the result sequence, execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively;

[0028] The determination and invocation sub-unit, which is used to, if the similarity between the first result voting information of the first k candidate task execution results in the result sequence is greater than or equal to the data difference threshold, then generate second result voting information for the first k candidate task execution results, determine the second result statistical values of the first k candidate task execution results respectively based on the second result voting information corresponding to the execution results of the first k candidate tasks respectively, and execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively for the second result statistical values corresponding to the execution results of the first k candidate tasks respectively.

[0029] Among them, the result determination unit is specifically used for:

[0030] Determine the execution result of the candidate task with the largest result statistical value as the result to be confirmed, and determine the result to be confirmed as the target task execution result; or,

[0031] Determine the execution result of the candidate task with the largest result statistical value as the result to be confirmed, broadcast the result to be confirmed to the second blockchain node for verification, and when the result evaluation node verifies and passes the result to be confirmed, determine the result to be confirmed as the target task execution result.

[0032] Among them, when obtaining the result voting information of the result evaluation node for the execution results of N candidate tasks respectively, the voting statistics unit further includes:

[0033] A constraint evaluation subunit, configured to obtain the artificial intelligence model behavior constraints from the blockchain network, evaluate the execution results of N candidate tasks respectively by using the artificial intelligence model constraints, obtain the result voting information of the first blockchain node, and send the result voting information of the first blockchain node to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model task issued by the task publishing node;

[0034] An information receiving subunit, configured to receive the result voting information of the execution results of N candidate tasks obtained by the second blockchain node by using the artificial intelligence model behavior constraints.

[0035] Among them, the device further includes:

[0036] A model recording module, configured to record the first artificial intelligence model and the second artificial intelligence model based on the result statistical values respectively corresponding to the execution results of N candidate tasks; the first artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the largest result statistical value among N artificial intelligence models; the second artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the smallest result statistical value among N artificial intelligence models;

[0037] A model management module, configured to manage N artificial intelligence models based on the historical record data respectively corresponding to N artificial intelligence models; the historical record data includes the record data for the first artificial intelligence model and the second artificial intelligence model.

[0038] Among them, the model management module includes:

[0039] An exception obtaining unit, configured to obtain the model exception times respectively corresponding to N artificial intelligence models from the historical record data respectively corresponding to N artificial intelligence models; the model exception times refer to the number of times when the historical result statistical value of the corresponding artificial intelligence model is the smallest;

[0040] Anomaly determination unit, configured to determine an artificial intelligence model with the number of model anomalies greater than or equal to the model anomaly threshold as an abnormal model, and send the abnormal model to a management node, so that the management node performs detection and processing on the abnormal model.

[0041] Wherein, the device further includes:

[0042] Model update module, configured to, if the first blockchain node is a management node, detect the abnormal model, determine the abnormal parameters learned in the abnormal model, delete or adjust the abnormal parameters to obtain an updated model; or, obtain a model constraint for the abnormal model, and write the model constraint into the abnormal model to obtain an updated model; the first blockchain node is a blockchain node that receives model tasks published by a task publishing node;

[0043] Model on-chain module, configured to upload the updated model to a blockchain network;

[0044] Abnormal feedback module, configured to, if the first blockchain node is not a management node, execute the process of sending the abnormal model to the management node.

[0045] Wherein, the number of result evaluation nodes is M, and M is a positive integer; the result sending module includes:

[0046] Credit acquisition unit, configured to acquire node credit data corresponding to M result evaluation nodes respectively, and acquire a credit threshold corresponding to the model task;

[0047] Node determination unit, configured to determine, among the M result evaluation nodes, a result evaluation node whose node credit data meets the credit threshold as a first result evaluation node, and send N candidate task execution results to the first result evaluation node.

[0048] Wherein, the device further includes:

[0049] Update determination module, configured to acquire the deviation degree between the result voting information corresponding to M result evaluation nodes respectively and the target task execution result, and determine the credit update value corresponding to each of the M result evaluation nodes; any result voting information refers to the voting data of the corresponding result evaluation node on the N candidate task execution results; the result voting information corresponding to M result evaluation nodes respectively is used to determine the target task execution result;

[0050] Credit update module, configured to update the node credit data corresponding to M result evaluation nodes respectively by using the credit update values corresponding to M result evaluation nodes respectively to obtain the updated credit data corresponding to M result evaluation nodes respectively;

[0051] A permission processing module is used to perform evaluation permission limit processing on result evaluation nodes where the updated credit data is less than the basic credit threshold.

[0052] Among them, the number of result evaluation nodes is M, and M is a positive integer; the device further includes:

[0053] An asset receiving module is used to receive the first digital asset transferred by the task publishing node;

[0054] A coefficient determination module is used to obtain the deviation degree between the result voting information corresponding to each of the M result evaluation nodes and the target task execution result, and determine the asset allocation coefficients corresponding to each of the M result evaluation nodes based on the deviation degrees corresponding to each of the M result evaluation nodes;

[0055] An asset transfer module is used to divide the first digital asset into a second digital asset and a third digital asset, and transfer the second digital asset to the target artificial intelligence model corresponding to the target task execution result;

[0056] An asset allocation module is used to divide the third digital asset into M node digital assets based on the asset allocation coefficients corresponding to each of the M result evaluation nodes, and allocate the M node digital assets to the M result evaluation nodes based on the asset allocation coefficients corresponding to each of the M result evaluation nodes.

[0057] On the one hand, an embodiment of the present application provides a computer device, including a processor, a memory, and an input / output interface;

[0058] The processor is respectively connected to the memory and the input / output interface. Among them, the input / output interface is used to receive and output data, the memory is used to store a computer program, and the processor is used to call the computer program so that the computer device including the processor executes the artificial intelligence management method in an aspect of the embodiment of the present application.

[0059] On the one hand, an embodiment of the present application provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded and executed by a processor so that a computer device having the processor executes the artificial intelligence management method in an aspect of the embodiment of the present application.

[0060] One aspect of the embodiments of the present application provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in various alternative manners in one aspect of the embodiments of the present application. In other words, when the computer instructions are executed by the processor, the methods provided in various alternative manners in one aspect of the embodiments of the present application are implemented.

[0061] Implementing the embodiments of the present application will have the following beneficial effects:

[0062] In the embodiments of the present application, a model task published by a task publishing node can be received, and the model task is shared to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; candidate task execution results of the N artificial intelligence models for the model task are obtained, and the N candidate task execution results are sent to a result evaluation node, so that the result evaluation node performs result detection on the N candidate task execution results; a target task execution result consensus by the result evaluation node for the N candidate task execution results is obtained, and the target task execution result is fed back to the task publishing node. That is to say, the same type of artificial intelligence model can be independently trained respectively to obtain N artificial intelligence models with the same function, and then the N artificial intelligence models are used to process the same model task, so that calculation results (i.e., candidate task execution results) of the N AI models for the model task can be obtained. Blockchain nodes on the blockchain network can perform consensus voting on a series of candidate task execution results to obtain a target task execution result, so that the final target execution result is processed by multiple AI models and multiple result evaluation nodes, which can improve the security and result accuracy of the artificial intelligence model. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0064] Figure 1 It is a network interaction architecture diagram for artificial intelligence management provided by the embodiments of the present application;

[0065] Figure 2 It is another network interaction architecture diagram for artificial intelligence management provided by the embodiments of the present application;

[0066] Figure 3 It is a schematic diagram of an artificial intelligence management scenario provided by an embodiment of the present application;

[0067] Figure 4 It is a flowchart of a method for artificial intelligence management provided by an embodiment of the present application;

[0068] Figure 5 It is a flowchart of a method for task processing and feedback provided by an embodiment of the present application;

[0069] Figure 6 It is a schematic diagram of an artificial intelligence management device provided by an embodiment of the present application;

[0070] Figure 7 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0071] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0072] Among them, if it is necessary to collect data of an object (such as a user, etc.) in the present application, a prompt interface or a pop-up window is displayed before and during the collection. The prompt interface or the pop-up window is used to prompt the user that some data is currently being collected. Only after obtaining the user's confirmation operation on the prompt interface or the pop-up window, the relevant steps for data acquisition are started, otherwise it ends. Moreover, the obtained user data will be used in reasonable, legal scenarios or uses, etc. Optionally, in some scenarios where user data needs to be used but the user's authorization has not been obtained, authorization can also be requested from the user, and the user data will be used when the authorization is passed. That is to say, the use of user data in the present application complies with the relevant regulations of laws and regulations, that is, the use of user data is reasonable and legal.

[0073] Among them, blockchain is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithms. Blockchain, in essence, is a decentralized database, a string of data blocks generated by using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity (anti-counterfeiting) of the information and generate the next block. Blockchain can include the blockchain underlying platform, the platform product service layer, and the application service layer.

[0074] Among them, the underlying blockchain platform may include processing modules such as user management, basic services, smart contracts, and operation detection. Among them, the user management module is responsible for the identity information management of all blockchain participants, including maintaining the generation of public and private keys (account management), key management, and the maintenance of the correspondence between the real identity of the user and the blockchain address (permission management). And under authorized circumstances, it supervises and audits the transaction situations of certain real identities, and provides the rule configuration for risk control (risk control and auditing); the basic service module is deployed on all blockchain node devices to verify the validity of business requests, and after reaching a consensus on valid requests, records them on the storage. For a new business request, the basic service first performs interface adaptation parsing and authentication processing (interface adaptation), then encrypts the business information through a consensus algorithm (consensus management), transmits it to the shared ledger in a complete and consistent manner after encryption (network communication), and records and stores it; the smart contract module is responsible for the registration and issuance of contracts, as well as contract triggering and contract execution. Developers can define contract logic through a certain programming language, publish it to the blockchain (contract registration), trigger the execution by calling keys or other events according to the logic of the contract terms, complete the contract logic, and at the same time provide functions for contract upgrade and cancellation; the operation detection module is mainly responsible for the deployment, configuration modification, contract setting, cloud adaptation during the product release process, and the visual output of the real-time state during product operation, such as: alarm, detecting network conditions, and detecting the health status of node devices, etc.

[0075] Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines, enabling the machines to have the functions of perception, reasoning, and decision-making.

[0076] Artificial intelligence technology is an interdisciplinary subject, involving a wide range of fields, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, autonomous driving, and intelligent transportation.

[0077] Emergence: A metaphor for the large-scale appearance of (people and things). It refers to the quantitative change of things over time. They appear in large numbers during the same period; they appear suddenly. Emergence is a transition from a lower level to a higher level. It is a mutation in the performance and structure of a macro system based on the evolution of micro entities. During this process, new qualities can emerge from old qualities. In AI, when the data scale is greater than a critical threshold (such as 60 billion parameters, etc.), the AI will exhibit unprecedented new capabilities, and this phenomenon is called AI emergence.

[0078] In the embodiments of the present application, by constructing a blockchain network between the AI model and the user, the behavior of the AI model is restricted. When the user needs the AI model to execute a certain model task, the model task can be sent to the blockchain network. At the same time, multiple independently trained AI models deployed in the blockchain network receive the model task and perform independent operations on the model task. Each AI model feeds back its own calculation result for the model task (i.e., the candidate task execution result) to the blockchain network; the blockchain nodes in the blockchain network can conduct a consensus vote on a series of candidate task execution results to determine the target task execution result of the model task. Since the target task execution result is jointly determined by multiple AI models and multiple blockchain nodes, it can be considered that the accuracy of the AI model's result is guaranteed and the security of the AI model is improved.

[0079] Specifically, the participants accessing the blockchain network include task publishers, AI models (i.e., task executors), and result evaluators, which can be considered to form an AI management system. That is to say, this AI management system can be considered to include task publishers (which can be called task publishing nodes), AI models, and result evaluators (which can be called result evaluation nodes). The architecture of this AI management system can be seen in Figure 1 , Figure 1 is a network interaction architecture diagram for artificial intelligence management provided by the embodiments of the present application. As Figure 1 shown, when the task publishing node 101 generates a model task, it can publish the model task to the blockchain network. Specifically, it can publish the model task to the model management node 102 where the AI model is deployed. Among them, the number of model management nodes 102 can be multiple, and multiple means at least two, such as Figure 1 the model management node 102a, model management node 102b, and model management node 102c shown in Figure 1Result evaluation nodes 103a, 103b, 103c, etc. shown in [figure]. At this time, each result evaluation node can receive the candidate task execution results corresponding to N AI models respectively. Each result evaluation node can perform result detection on the candidate task execution results corresponding to N AI models respectively, and consensus-determine the target task execution result for the model task, and send the target task execution result to the task publishing node 101.

[0080] Alternatively, refer to Figure 2 , Figure 2 which is another network interaction architecture diagram for artificial intelligence management provided by an embodiment of the present application. As Figure 2 shown, the task publishing node 201 can publish a model task on the blockchain network, and share the model task to N AI models 202 through the blockchain network, such as AI model 202a, AI model 202b, and AI model 202c, etc. Each AI model can process the model task, obtain the candidate task execution result for the model task, and send the candidate task execution result to the result evaluation node 203. The number of the result evaluation nodes 203 is M, such as result evaluation node 203a, result evaluation node 203b, and result evaluation node 203c, etc. Further, the result evaluation node 203 can evaluate the candidate task execution results of each AI model, determine the target task execution result, and perform an on-chain process on the target task execution result. The task publishing node 201 can obtain the target task execution result of the model task from the blockchain network.

[0081] That is to say, N AI models are deployed in the blockchain network, and M result evaluation nodes are blockchain nodes in the blockchain network. The task publishing node can be a blockchain node deployed in the blockchain network or an off-chain node that can access the blockchain network. Optionally, the model management node deployed with the AI model can partially overlap with the result evaluation node, or the AI model can be deployed on the result evaluation node, or the model management node and the result evaluation node can be independent of each other, which is not limited here.

[0082] Specifically, please refer to Figure 3 , Figure 3 which is a schematic diagram of an artificial intelligence management scenario provided by an embodiment of the present application. As Figure 3 shown, the task publishing node 301 can publish a model task in the blockchain network. N AI models 302 accessing the blockchain network can receive the model task from the blockchain network, where N is a positive integer, such as Figure 3AI models such as AI model 302a, AI model 302b, and AI model 302c shown in []. The N AI models can respectively process the model task, obtain their respective candidate task execution results for the model task, and feedback the candidate task execution results to the blockchain network. Finally, the result evaluation node 303 can conduct a consensus vote on the N candidate task execution results to obtain the target task execution result for the model task. Among them, the number of result evaluation nodes 303 can be M, such as Figure 3 result evaluation node 303a, result evaluation node 303b, and result evaluation node 303c shown in []. Through this AI management system, a model task can be calculated and processed by multiple AI models, and a consensus vote can be conducted by multiple result evaluation nodes, so that the target task execution result has been authenticated by multiple parties, which can improve the security of the artificial intelligence model and the accuracy of task processing.

[0083] It can be understood that the blockchain node mentioned in the embodiments of the present application can be a computer device. The computer device in the embodiments of the present application includes, but is not limited to, a terminal device or a server. In other words, the computer device can be a server or a terminal device, or a system composed of a server and a terminal device. Among them, the above-mentioned terminal device can be an electronic device, including but not limited to a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a vehicle-mounted device, an augmented reality / virtual reality (AR / VR) device, a head-mounted display, a smart TV, a wearable device, a smart speaker, a digital camera, a camera, and other mobile internet devices (MID) with network access capabilities, or terminal devices in scenarios such as trains, ships, and flights. Among them, the above-mentioned server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, vehicle-road coordination, content delivery network (CDN), and big data and artificial intelligence platforms.

[0084] Further, please refer to Figure 4 , Figure 4 which is a flowchart of a method for artificial intelligence management provided by the embodiments of the present application. As Figure 4 shown, the artificial intelligence management process includes the following steps:

[0085] Step S401: Receive the model task published by the task publishing node and share the model task with N artificial intelligence models.

[0086] In the embodiment of the present application, the first blockchain node can receive the model task published by the task publishing node and share the model task with N artificial intelligence models. Among them, the task publishing node can be an off-chain node that can access the blockchain network or any blockchain node in the blockchain network. Among them, the first blockchain node can be any blockchain node in the blockchain network. That is to say, the first blockchain node and the task publishing node may be the same node or different nodes. N is a positive integer; the N artificial intelligence models are independently trained. Optionally, the task publishing node can generate a task publishing transaction based on the model task and upload the task publishing transaction to the blockchain network. The N artificial intelligence models can obtain the model task based on the task publishing transaction. Or, a model management contract can be deployed in the blockchain network. The task publishing node can call the model management contract based on the model task and publish the model task to the N artificial intelligence models through the model management contract. Or, the first blockchain node can receive the model task published by the task publishing node and publish the model task to the N artificial intelligence models, etc. Among them, the model task may include, but is not limited to, input data and a task purpose. The input data is used to represent the data that needs to be input into the artificial intelligence model, and the task purpose is used to represent the processing method required by the model task and can be used to determine the artificial intelligence model. Specifically, the first blockchain node can determine N artificial intelligence models based on the task purpose in the model task; publish the model task to the N artificial intelligence models, and the publishing process can refer to the relevant description of the above publishing process in this step. Further, any artificial intelligence model can input the input data in the model task into the artificial intelligence model for prediction to obtain a candidate task execution result for the model task. That is, the N artificial intelligence models can respectively obtain candidate task execution results for the model task and feedback the candidate task execution results to the blockchain network.

[0087] Step S402: Obtain the candidate task execution results of the N artificial intelligence models for the model task respectively, and send the N candidate task execution results to the result evaluation node so that the result evaluation node can perform result detection on the N candidate task execution results.

[0088] In an embodiment of the present application, the first blockchain node may obtain the candidate task execution results of N artificial intelligence models for model tasks from the blockchain network. Further, the N candidate task execution results may be sent to the result evaluation node so that the result evaluation node performs result detection on the N candidate task execution results. Optionally, the number of result evaluation nodes is M, where M is a positive integer. The M result evaluation nodes may or may not include the first blockchain node.

[0089] Among them, the M result evaluation nodes may be all blockchain nodes in the blockchain network; or they may be the blockchain nodes consensus-selected by the blockchain nodes in the blockchain network. For example, the first blockchain node may obtain the node information of the blockchain nodes in the blockchain network, determine M result evaluation nodes to be confirmed based on the node information, and when the blockchain nodes in the blockchain network consensus-pass on the M result evaluation nodes to be confirmed, determine the M result evaluation nodes to be confirmed as the M result evaluation nodes. Among them, the node information may include any one or more of P node data of the corresponding blockchain node, where P is a positive integer, and the P node data may include, but is not limited to, node space occupancy rate, node bandwidth, node credit data, and node collateral resources, etc.

[0090] Specifically, in a result sending method ①, the first blockchain node may directly send the N candidate task execution results to the M result evaluation nodes so that the result evaluation nodes perform result detection on the N candidate task execution results. Or, in a result sending method ②, the node credit data corresponding to the M result evaluation nodes may be obtained, and the credit threshold corresponding to the model task may be obtained; the result evaluation nodes among the M result evaluation nodes whose node credit data meets the credit threshold are determined as the first result evaluation nodes, and the N candidate task execution results are sent to the first result evaluation nodes so that the first result evaluation nodes perform result detection on the N candidate task execution results.

[0091] Step S403: Obtain the target task execution result consensus-selected by the result evaluation node for the N candidate task execution results, and feedback the target task execution result to the task publishing node.

[0092] In the embodiment of the present application, the first blockchain node may obtain the target task execution result consensus by the result evaluation node for the execution results of N candidate tasks. Specifically, when the first blockchain node is not the result evaluation node, the first blockchain node may obtain the target task execution result fed back by M result evaluation nodes (the above result sending method ①) or the first result evaluation node (the above result sending method ②) from the blockchain network. The voting process of M result evaluation nodes or the first result evaluation node for the execution results of N candidate tasks may refer to the voting process when the first blockchain node is the result evaluation node as described below.

[0093] When the first blockchain node is the result evaluation node, the first blockchain node may obtain the result voting information of the result evaluation node for the execution results of N candidate tasks respectively. At this time, the result voting information of the M result evaluation nodes for the execution results of N candidate tasks may be as shown in Table 1 below:

[0094] Table 1

[0095]

[0096] Or, as shown in Table 2 below:

[0097] Table 2

[0098] Result evaluation node Result evaluation node 1 … Result evaluation node M Voting result Candidate task execution result a … Candidate task execution result b

[0099] That is to say, when evaluating the execution results of N candidate tasks, each result evaluation node may select one or more candidate task execution results for voting, as shown in Table 1 above, or may only vote for one candidate task execution result, as shown in Table 2 above.

[0100] Furthermore, the result statistical values corresponding to the execution results of N candidate tasks may be determined based on the result voting information corresponding to the execution results of N candidate tasks respectively. The target task execution result may be determined based on the result statistical values corresponding to the execution results of N candidate tasks respectively.

[0101] Specifically, in a result statistics method ①, the result evaluation node can perform a result voting process for a certain number of rounds or for a certain period of time on the execution results of N candidate tasks, and obtain the result voting information corresponding to the execution results of the N candidate tasks respectively. Specifically, in the i-th round of the result voting process, the first blockchain node can obtain the (i - 1)-th result voting information generated by the result evaluation node in the (i - 1)-th round of the result voting process, generate the i-th result voting information based on the (i - 1)-th result voting information, and send the i-th result voting information generated by the first blockchain node to the second blockchain node. The first blockchain node refers to the blockchain node that receives the model task published by the task publishing node; the second blockchain node is the blockchain node other than the first blockchain node in the result evaluation node. For example, in result sending method ①, the second blockchain node is the blockchain node other than the first blockchain node among the M result evaluation nodes. For example, in result sending method ②, the second blockchain node is the blockchain node other than the first blockchain node in the first result evaluation node; i is a positive integer. For example, taking result sending method ① as an example, the (i - 1)-th result voting information can be as shown in Table 3 below:

[0102] Table 3

[0103]

[0104] As shown in Table 3 above, "1" indicates that the result evaluation node has voted on the corresponding candidate task execution result, and "0" indicates that the result evaluation node has not voted on the corresponding candidate task execution result. That is to say, in this method, in each round of the result voting process, each result evaluation node can vote on one or more candidate task execution results. The first blockchain node can adjust the (i - 1)-th result voting information generated by the first blockchain node based on the (i - 1)-th result voting information generated by the result evaluation node, and obtain the i-th result voting information of the first blockchain node for the execution results of the N candidate tasks, such as updating the data in the column where the first blockchain node is located based on Table 3 above.

[0105] Furthermore, if the i-th round of the result voting process does not meet the voting iteration condition, obtain the i-th result voting information corresponding to the second blockchain node, and generate the (i + 1)-th result voting information of the first blockchain node based on the i-th result voting information of the result evaluation node. If the i-th round of the result voting process meets the voting iteration condition, obtain the P-round result voting information of the result evaluation node, and count the result statistical values corresponding to the execution results of the N candidate tasks respectively based on the P-round result voting information of the result evaluation node; P is a positive integer, and P is the number of rounds of the result voting process indicated by the voting iteration condition.

[0106] Among them, the voting iteration condition may be any one or more of the result voting process for the iteration round threshold and the duration of the result voting process reaching the voting duration. For example, if the voting iteration condition is the result voting process for the iteration round threshold, at this time, if i is the iteration round threshold, it means that the i-th round result voting process meets the voting iteration condition, otherwise it does not. For example, if the voting iteration condition is that the duration of the result voting process reaches the voting duration, at this time, if at the end of the i-th round result voting process, the duration is greater than or equal to the voting duration, it means that the i-th round result voting process meets the voting iteration condition, etc.

[0107] Alternatively, in a result statistics method ②, the execution results of N candidate tasks are scored to obtain the first result quality scores respectively corresponding to the execution results of the N candidate tasks, and the first result quality scores respectively corresponding to the execution results of the N candidate tasks are sent to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node is the blockchain node that receives the model task issued by the task publishing node. The second result quality scores respectively corresponding to the execution results of the N candidate tasks are obtained, and based on the first result quality score and the second result quality score of each candidate task execution result, the result statistical value of the candidate task execution result is determined; the first result quality score and the second result quality score of each candidate task execution result form the result voting information of the candidate task execution result. That is to say, in this way, each result voting node can score the execution results of the N candidate tasks respectively.

[0108] Alternatively, in a third result statistics method, the first blockchain node may obtain the first result voting information of the result evaluation nodes for the execution results of N candidate tasks respectively, and determine the first result statistical values corresponding to the execution results of the N candidate tasks respectively based on the first result voting information corresponding to the execution results of the N candidate tasks respectively. Sort the execution results of the N candidate tasks based on the first result statistical values corresponding to the execution results of the N candidate tasks respectively to obtain a result sequence. If the degree of difference between the first result voting information of the first two candidate task execution results in the result sequence is less than the data difference threshold, then based on the result sequence, perform the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively; if the similarity between the first result voting information of the first k candidate task execution results in the result sequence is greater than or equal to the data difference threshold, then generate second result voting information for the execution results of the first k candidate tasks, determine the second result statistical values corresponding to the execution results of the first k candidate tasks respectively based on the second result voting information corresponding to the execution results of the first k candidate tasks respectively, and perform the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively for the second result statistical values corresponding to the execution results of the first k candidate tasks respectively.

[0109] Among them, in any of the above result statistics methods, the result evaluation node may evaluate the execution result of the candidate task based on the behavior constraint of the artificial intelligence model to obtain result voting information (such as the result voting information, the first result voting information, or the second result voting information in each round as described above, etc.). Specifically, the first blockchain node may obtain the behavior constraint of the artificial intelligence model from the blockchain network, evaluate the execution results of the N candidate tasks respectively using the artificial intelligence model constraint to obtain the result voting information of the first blockchain node, and send the result voting information of the first blockchain node to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node. Receive the result voting information of the execution results of the N candidate tasks obtained by the second blockchain node using the behavior constraint of the artificial intelligence model. At this time, each result evaluation node can obtain the result voting information of each result evaluation node for the execution results of the N candidate tasks.

[0110] Furthermore, the execution result of the target task can be determined based on the result statistical values respectively corresponding to the execution results of N candidate tasks. Specifically, the execution result of the candidate task with the largest result statistical value can be determined as the result to be confirmed, and the result to be confirmed can be determined as the execution result of the target task. Since the result statistical values respectively corresponding to the execution results of N candidate tasks are jointly determined by the result evaluation nodes, it can be directly determined, which can also improve the accuracy and security of the artificial intelligence model. Alternatively, the execution result of the candidate task with the largest result statistical value can be determined as the result to be confirmed, and the result to be confirmed can be broadcast to the second blockchain node for verification. When the result evaluation node verifies and passes the result to be confirmed, the result to be confirmed can be determined as the execution result of the target task. That is to say, further verification can also be carried out, thereby improving the accuracy and security of the artificial intelligence model.

[0111] Furthermore, the execution result of the target task can be processed on the blockchain. The first blockchain node can feedback the execution result of the target task to the task publishing node, or the task publishing node can obtain the execution result of the target task from the blockchain network.

[0112] Further optionally, the task publishing node can distribute resources to N artificial intelligence models and M result evaluation nodes for this model task. Specifically, the first blockchain node can receive the first digital asset (such as denoted as Token) transferred by the task publishing node. The first digital asset can be divided into a second digital asset and a third digital asset. The second digital asset is transferred to the target artificial intelligence model corresponding to the execution result of the target task, and the third digital asset is allocated to M result evaluation nodes. Among them, the first digital asset can be divided into a second digital asset and a third digital asset by using the model node division ratio, and the model node division ratio can be the data consensus by the blockchain nodes in the blockchain network. Among them, when allocating the third digital asset to M result evaluation nodes, the third digital asset can be directly evenly distributed to M result evaluation nodes; or, the deviation degree between the result voting information respectively corresponding to M result evaluation nodes and the execution result of the target task can be obtained, and based on the deviation degrees respectively corresponding to M result evaluation nodes, the asset allocation coefficients respectively corresponding to M result evaluation nodes can be determined. Based on the asset allocation coefficients respectively corresponding to M result evaluation nodes, the third digital asset is divided into M node digital assets, and based on the asset allocation coefficients respectively corresponding to M result evaluation nodes, the M node digital assets are allocated to M result evaluation nodes, thereby optimizing the result evaluation nodes and improving the security of the nodes.

[0113] In an embodiment of the present application, a model task published by a task publishing node can be received, and the model task can be shared with N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; candidate task execution results of the N artificial intelligence models for the model task are obtained, and the N candidate task execution results are sent to a result evaluation node so that the result evaluation node can perform result detection on the N candidate task execution results; the target task execution result consensus by the result evaluation node for the N candidate task execution results is obtained, and the target task execution result is fed back to the task publishing node. That is to say, the same type of artificial intelligence model can be independently trained respectively to obtain N artificial intelligence models with the same function, and then the N artificial intelligence models are used to process the same model task, so that the calculation results (i.e., candidate task execution results) of the N AI models for the model task can be obtained. Blockchain nodes on the blockchain network can conduct a consensus vote on a series of candidate task execution results to obtain the target task execution result, so that the final target execution result is processed by multiple AI models and multiple result evaluation nodes, which can improve the security and result accuracy of the artificial intelligence model.

[0114] Further, please refer to Figure 5 , Figure 5 which is a flowchart of a task processing and feedback method provided by an embodiment of the present application. As Figure 5 shown, the process may include the following steps:

[0115] Step S501: Receive the model task published by the task publishing node, and share the model task with N artificial intelligence models.

[0116] In an embodiment of the present application, N is a positive integer; the N artificial intelligence models are independently trained, and the relevant description in step S401 of Figure 4 can be referred to for this process, and details will not be elaborated here.

[0117] Step S502: Obtain the candidate task execution results of the N artificial intelligence models for the model task respectively, and send the N candidate task execution results to the result evaluation node so that the result evaluation node can perform result detection on the N candidate task execution results.

[0118] In an embodiment of the present application, the relevant description in step S402 of Figure 4 can be referred to for this process, and details will not be elaborated here.

[0119] Step S503: Obtain the target task execution result consensus by the result evaluation node for the N candidate task execution results, and feed back the target task execution result to the task publishing node.

[0120] In the embodiment of the present application, this process can be referred to Figure 4 in the relevant description of step S403 in

[0121] Further optionally, N artificial intelligences can be managed based on the execution result of the target task for the model task, the candidate task execution results respectively corresponding to N artificial intelligence models, and the result voting information of the result evaluation node for the N candidate task execution results (as shown in step S504 below), and the result evaluation node can be credit managed (as shown in step S505 below).

[0122] Step S504, manage N artificial intelligence models based on the execution result of the target task.

[0123] In the embodiment of the present application, the first blockchain node can record the first artificial intelligence model and the second artificial intelligence model based on the result statistical values respectively corresponding to the N candidate task execution results; the first artificial intelligence model refers to the artificial intelligence model corresponding to the candidate task execution result with the largest result statistical value among the N artificial intelligence models; the second artificial intelligence model refers to the artificial intelligence model corresponding to the candidate task execution result with the smallest result statistical value among the N artificial intelligence models. Manage N artificial intelligence models based on the historical record data respectively corresponding to the N artificial intelligence models; the historical record data includes the record data for the first artificial intelligence model and the second artificial intelligence model.

[0124] Specifically, the number of model anomalies respectively corresponding to the N artificial intelligence models can be obtained from the historical record data respectively corresponding to the N artificial intelligence models; the number of model anomalies refers to the number of times when the historical result statistical value of the corresponding artificial intelligence model is the smallest. Further, the artificial intelligence model with the number of model anomalies greater than or equal to the model anomaly threshold is determined as an abnormal model, and the abnormal model is sent to the management node so that the management node can perform detection and processing on the abnormal model. Among them, the management node can be considered as the node corresponding to the management personnel or developers.

[0125] Optionally, if the first blockchain node is a management node, it can update the abnormal model to obtain an updated model, upload the updated model to the blockchain network, and invalidate the abnormal model. Specifically, it can detect the abnormal model to determine the abnormal parameters learned in the abnormal model, delete or adjust the abnormal parameters to obtain an updated model; or, obtain the model constraints for the abnormal model and write the model constraints into the abnormal model to obtain an updated model; the first blockchain node is the blockchain node that receives the model task published by the task publishing node. Optionally, if the abnormality degree of the abnormal model is greater than or equal to the model deletion threshold, the abnormal model can be directly deleted. If the first blockchain node is not a management node, the process of sending the abnormal model to the management node is executed. At this time, for the update process of the abnormal model by the management node, reference can be made to the above update process of the abnormal model by the first blockchain node. Further optionally, when uploading the updated model to the blockchain network, a key mark can be added to the updated model through the blockchain network, and the key mark is used to pay key attention to the updated model; in the first H task processes of the updated model, as long as the result statistical value is less than the marked abnormality threshold, the updated model can be deleted. H is a positive integer. If the updated model has not been deleted after H task processes, the key mark of the updated model can be deleted.

[0126] In this way, based on the situation of each task process of N artificial intelligence models, abnormal detection can be performed on the N artificial intelligence models, achieving an effect of mutual detection between different artificial intelligence models, so that abnormal models with abnormal situations such as AI emergence, abnormal model deployment, or rule tampering that may occur or have occurred can be detected in a timely manner, thereby improving the security of artificial intelligence models.

[0127] Step S505, perform credit management on the result evaluation node based on the target task execution result.

[0128] In the embodiments of the present application, the first blockchain node may obtain the deviation degrees between the result voting information respectively corresponding to M result evaluation nodes and the target task execution result, and determine the credit update values respectively corresponding to the M result evaluation nodes. Wherein, any result voting information refers to the voting data of the corresponding result evaluation node for N candidate task execution results; the result voting information respectively corresponding to the M result evaluation nodes is used to determine the target task execution result. The node credit data respectively corresponding to the M result evaluation nodes are updated by using the credit update values respectively corresponding to the M result evaluation nodes, and the updated credit data respectively corresponding to the M result evaluation nodes are obtained. For the result evaluation nodes with updated credit data less than the basic credit threshold, evaluation permission restriction processing is performed, such as canceling the result evaluation permission, or restricting the result evaluation permission within a certain period of time, etc. Alternatively, the deviation degrees between the result voting information respectively corresponding to the M result evaluation nodes and the target task execution result may be obtained, the second result evaluation nodes corresponding to the result voting information with deviation degrees greater than the node abnormality threshold are obtained, and the node deviation times of the second result evaluation nodes are updated; the node credit data of the result evaluation nodes with node deviation times greater than or equal to the abnormality times threshold are reduced.

[0129] Through the above process, based on the evaluation of the candidate task execution results by the M result evaluation nodes, the credit management of the M result evaluation nodes can be performed in reverse, so that the security of the nodes can be improved, and thus the security and accuracy of the artificial intelligence model can be improved to a certain extent.

[0130] Further, please refer to Figure 6 , Figure 6 which is a schematic diagram of an artificial intelligence management device provided by the embodiments of the present application. The artificial intelligence management device may be a computer program (including program codes, etc.) running in a computer device. For example, the artificial intelligence management device may be an application software; the device may be used to execute the corresponding steps in the method provided by the embodiments of the present application. As Figure 6 shown, the artificial intelligence management device 600 may be used for Figure 4 the computer device corresponding to the corresponding embodiment. Specifically, the device may include: a task sharing module 11, a result obtaining module 12, a result sending module 13, a result determining module 14, and a result feedback module 15.

[0131] The task sharing module 11 is configured to receive the model tasks published by the task publishing node and share the model tasks to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained;

[0132] The result obtaining module 12 is configured to obtain the candidate task execution results of the N artificial intelligence models respectively for the model tasks;

[0133] A result sending module 13, configured to send the execution results of N candidate tasks to a result evaluation node, so that the result evaluation node performs result detection on the execution results of the N candidate tasks;

[0134] A result determination module 14, configured to obtain the target task execution result consensus by the result evaluation node for the execution results of the N candidate tasks;

[0135] A result feedback module 15, configured to feedback the target task execution result to the task publishing node.

[0136] Wherein, the result determination module 14 includes:

[0137] A vote counting unit 141, configured to obtain the result vote information of the result evaluation node for the execution results of the N candidate tasks respectively, and determine the result statistical values corresponding to the execution results of the N candidate tasks respectively based on the result vote information corresponding to the execution results of the N candidate tasks respectively;

[0138] A result determination unit 142, configured to determine the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively.

[0139] Wherein, the vote counting unit 141 includes:

[0140] A result vote subunit 141a, configured to obtain the (i - 1)-th result vote information generated by the result evaluation node in the (i - 1)-th result vote process during the i-th result vote process, generate the i-th result vote information according to the (i - 1)-th result vote information, and send the i-th result vote information generated by the first blockchain node to the second blockchain node; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node; the second blockchain node is the blockchain node other than the first blockchain node in the result evaluation node; i is a positive integer;

[0141] A vote generation subunit 141b, configured to, if the i-th result vote process does not meet the vote iteration condition, obtain the i-th result vote information corresponding to the second blockchain node, and generate the (i + 1)-th result vote information of the first blockchain node based on the i-th result vote information of the result evaluation node;

[0142] A result statistics subunit 141c, configured to, if the i-th result vote process meets the vote iteration condition, obtain the P-round result vote information of the result evaluation node, and count the result statistical values corresponding to the execution results of the N candidate tasks respectively based on the P-round result vote information of the result evaluation node; P is a positive integer, and P is the number of rounds of the result vote process indicated by the vote iteration condition.

[0143] Wherein, the vote counting unit 141 includes:

[0144] The result evaluation sub - unit 141d is used to score the execution results of N candidate tasks, obtain the first result quality scores corresponding to the execution results of the N candidate tasks respectively, and send the first result quality scores corresponding to the execution results of the N candidate tasks to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node;

[0145] The score statistics sub - unit 141e is used to obtain the second result quality scores of the second blockchain node for the execution results of the N candidate tasks respectively, and determine the result statistical value of the execution result of each candidate task based on the first result quality score and the second result quality score of the execution result of each candidate task; the first result quality score and the second result quality score of the execution result of each candidate task form the result voting information of the execution result of this candidate task.

[0146] Among them, the voting statistics unit 141 includes:

[0147] The voting statistics sub - unit 141f is used to obtain the first result voting information of the result evaluation node for the execution results of the N candidate tasks respectively, and determine the first result statistical values corresponding to the execution results of the N candidate tasks respectively based on the first result voting information corresponding to the execution results of the N candidate tasks;

[0148] The result sorting sub - unit 141g is used to sort the execution results of the N candidate tasks based on the first result statistical values corresponding to the execution results of the N candidate tasks respectively, and obtain a result sequence;

[0149] The determination and invocation sub - unit 141h is used to, if the degree of difference between the first result voting information of the first two candidate task execution results in the result sequence is less than the data difference threshold, then based on the result sequence, execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively;

[0150] The determination and invocation sub - unit 141h is used to, if the similarity between the first result voting information of the first k candidate task execution results in the result sequence is greater than or equal to the data difference threshold, then generate second result voting information for the first k candidate task execution results, determine the second result statistical values of the first k candidate task execution results respectively based on the second result voting information corresponding to the execution results of the first k candidate tasks respectively, and execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively for the second result statistical values corresponding to the execution results of the first k candidate tasks respectively.

[0151] Among them, the result determination unit 142 is specifically used for:

[0152] Determine the execution result of the candidate task with the largest result statistical value as the result to be confirmed, and determine the result to be confirmed as the execution result of the target task; or,

[0153] Determine the execution result of the candidate task with the largest result statistical value as the result to be confirmed, broadcast the result to be confirmed to the second blockchain node for verification, and when the result evaluation node verifies and passes the result to be confirmed, determine the result to be confirmed as the execution result of the target task.

[0154] Among them, when obtaining the result voting information of the result evaluation node for the execution results of N candidate tasks respectively, the voting statistical unit 141 further includes:

[0155] The constraint evaluation subunit 141i is used to obtain the artificial intelligence model behavior constraints from the blockchain network, evaluate the execution results of N candidate tasks respectively using the artificial intelligence model constraints, obtain the result voting information of the first blockchain node, and send the result voting information of the first blockchain node to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node;

[0156] The information receiving subunit 141j is used to receive the result voting information of the execution results of N candidate tasks obtained by the second blockchain node using the artificial intelligence model behavior constraints.

[0157] Among them, the device 600 further includes:

[0158] The model recording module 16 is used to record the first artificial intelligence model and the second artificial intelligence model based on the result statistical values respectively corresponding to the execution results of N candidate tasks; the first artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the largest result statistical value among N artificial intelligence models; the second artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the smallest result statistical value among N artificial intelligence models;

[0159] The model management module 17 is used to manage N artificial intelligence models based on the historical record data respectively corresponding to N artificial intelligence models; the historical record data includes the record data for the first artificial intelligence model and the second artificial intelligence model.

[0160] Among them, the model management module 17 includes:

[0161] Anomaly acquisition unit 171 is configured to obtain the model anomaly counts corresponding to N artificial intelligence models respectively from the historical record data corresponding to the N artificial intelligence models; the model anomaly count refers to the number of times when the historical result statistical value of the corresponding artificial intelligence model is the smallest.

[0162] Anomaly determination unit 172 is configured to determine the artificial intelligence models with model anomaly counts greater than or equal to the model anomaly threshold as abnormal models, and send the abnormal models to the management node so that the management node can perform detection and processing on the abnormal models.

[0163] Wherein, the device 600 further includes:

[0164] Model update module 18 is configured to, if the first blockchain node is the management node, detect the abnormal model, determine the abnormal parameters learned in the abnormal model, delete or adjust the abnormal parameters to obtain an updated model; or, obtain the model constraints for the abnormal model and write the model constraints into the abnormal model to obtain an updated model; the first blockchain node is the blockchain node that receives the model tasks published by the task publishing node.

[0165] Model on-chain module 19 is configured to upload the updated model to the blockchain network.

[0166] Anomaly feedback module 20 is configured to, if the first blockchain node is not the management node, perform the process of sending the abnormal model to the management node.

[0167] Wherein, the number of result evaluation nodes is M, and M is a positive integer; the result sending module 13 includes:

[0168] Credit acquisition unit 131 is configured to obtain the node credit data corresponding to M result evaluation nodes respectively and obtain the credit threshold corresponding to the model task.

[0169] Node determination unit 132 is configured to determine the result evaluation nodes among the M result evaluation nodes whose node credit data meet the credit threshold as the first result evaluation nodes, and send the N candidate task execution results to the first result evaluation nodes.

[0170] Wherein, the device 600 further includes:

[0171] Update determination module 21 is configured to obtain the deviation degrees between the result voting information corresponding to M result evaluation nodes respectively and the target task execution result, and determine the credit update values corresponding to M result evaluation nodes respectively; any result voting information refers to the voting data of the corresponding result evaluation node on the N candidate task execution results; the result voting information corresponding to M result evaluation nodes respectively is used to determine the target task execution result.

[0172] A credit update module 22, configured to update the node credit data respectively corresponding to M result evaluation nodes by using the credit update values respectively corresponding to the M result evaluation nodes, so as to obtain the updated credit data respectively corresponding to the M result evaluation nodes;

[0173] An authority processing module 23, configured to perform evaluation authority restriction processing on the result evaluation nodes whose updated credit data is less than the basic credit threshold.

[0174] Wherein, the number of result evaluation nodes is M, and M is a positive integer; the apparatus 600 further includes:

[0175] An asset receiving module 24, configured to receive the first digital asset transferred by the task publishing node;

[0176] A coefficient determination module 25, configured to obtain the deviation degrees between the result voting information respectively corresponding to the M result evaluation nodes and the target task execution result, and determine the asset allocation coefficients respectively corresponding to the M result evaluation nodes based on the deviation degrees respectively corresponding to the M result evaluation nodes;

[0177] An asset transfer module 26, configured to divide the first digital asset into a second digital asset and a third digital asset, and transfer the second digital asset to the target artificial intelligence model corresponding to the target task execution result;

[0178] An asset allocation module 27, configured to divide the third digital asset into M node digital assets based on the asset allocation coefficients respectively corresponding to the M result evaluation nodes, and allocate the M node digital assets to the M result evaluation nodes based on the asset allocation coefficients respectively corresponding to the M result evaluation nodes.

[0179] An embodiment of the present application provides an artificial intelligence management device. The device can receive a model task issued by a task publishing node, share the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; obtain the candidate task execution results of the N artificial intelligence models for the model task respectively, and send the N candidate task execution results to a result evaluation node, so that the result evaluation node can perform result detection on the N candidate task execution results; obtain the target task execution result consensus by the result evaluation node for the N candidate task execution results, and feedback the target task execution result to the task publishing node. That is to say, the same artificial intelligence model can be independently trained respectively to obtain N artificial intelligence models with the same function, and then the N artificial intelligence models are used to process the same model task, so that the calculation results (i.e., candidate task execution results) of the N AI models for the model task can be obtained. The blockchain nodes on the blockchain network can conduct a consensus vote on a series of candidate task execution results to obtain the target task execution result, so that the final target execution result is processed by multiple AI models and multiple result evaluation nodes, which can improve the security and result accuracy of the artificial intelligence model.

[0180] See Figure 7 , Figure 7 is a schematic structural diagram of a computer device provided by an embodiment of the present application. As Figure 7 shown, the computer device in the embodiment of the present application may include: one or more processors 701, a memory 702, and an input / output interface 703. The processor 701, the memory 702, and the input / output interface 703 are connected through a bus 704. The memory 702 is used to store a computer program, and the computer program includes program instructions. The input / output interface 703 is used to receive data and output data, such as for data interaction between a task publishing node and a blockchain node, or for data interaction between result evaluation nodes, etc.; the processor 701 is used to execute the program instructions stored in the memory 702.

[0181] Among them, the processor 701 can perform the following operations:

[0182] Receive the model task issued by the task publishing node, and share the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained;

[0183] Obtain the candidate task execution results of the N artificial intelligence models for the model task respectively, and send the N candidate task execution results to the result evaluation node, so that the result evaluation node can perform result detection on the N candidate task execution results;

[0184] Obtain the result evaluation node, and feedback the target task execution result, which is the result reached through consensus on the execution results of N candidate tasks, to the task publishing node.

[0185] In some possible implementation manners, the processor 701 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0186] The memory 702 may include a read-only memory and a random access memory, and provide instructions and data to the processor 701 and the input / output interface 703. A part of the memory 702 may further include a non-volatile random access memory. For example, the memory 702 may also store information about the device type.

[0187] In a specific implementation, the computer device may execute, through each of its built-in functional modules, the implementation manners provided in each step of the Figure 4 For details, reference may be made to the implementation manners provided in each step of the Figure 4 and details are not described herein again.

[0188] By providing a computer device according to an embodiment of the present application, including: a processor, an input / output interface, and a memory, the processor obtains a computer program in the memory and executes the Figure 4For each step of the method shown in the figure, perform artificial intelligence management operations. In the embodiments of the present application, a model task published by a task publishing node is received, and the model task is shared with N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; candidate task execution results of the N artificial intelligence models for the model task are obtained, and the N candidate task execution results are sent to a result evaluation node so that the result evaluation node can perform result detection on the N candidate task execution results; the target task execution result consensus by the result evaluation node for the N candidate task execution results is obtained, and the target task execution result is fed back to the task publishing node. That is to say, the same type of artificial intelligence model can be independently trained to obtain N artificial intelligence models with the same function, and then the N artificial intelligence models are used to process the same model task, so that the calculation results (i.e., candidate task execution results) of the N AI models for the model task can be obtained. Blockchain nodes on the blockchain network can perform consensus voting on a series of candidate task execution results to obtain the target task execution result, so that the final target execution result is processed by multiple AI models and multiple result evaluation nodes, which can improve the security and result accuracy of the artificial intelligence model.

[0189] The embodiments of the present application also provide a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded and executed by the processor Figure 4 in each step of the artificial intelligence management method provided in the figure. For specific details, reference can be made to the Figure 4 implementation methods provided in each step in the figure, which will not be elaborated here. In addition, the description of the beneficial effects of using the same method will not be elaborated either. For the technical details not disclosed in the embodiments of the computer-readable storage medium involved in the present application, please refer to the description of the method embodiments of the present application. As an example, the computer program can be deployed to be executed on a computer device, or on multiple computer devices located at one location, or on multiple computer devices distributed at multiple locations and interconnected through a communication network.

[0190] The computer-readable storage medium may be the artificial intelligence management device provided in any of the foregoing embodiments or the internal storage unit of the computer device, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device. Further, the computer-readable storage medium may also include both the internal storage unit and the external storage device of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store the data that has been output or is to be output.

[0191] An embodiment of the present application also provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 4 the methods provided in various alternative manners in [the specific content], realizing independent training of the same artificial intelligence model to obtain N artificial intelligence models with the same function, and then using the N artificial intelligence models to process the same model task, so that N calculation results (i.e., candidate task execution results) of the N AI models for the model task can be obtained. The blockchain nodes on the blockchain network can conduct a consensus vote on a series of candidate task execution results to obtain the target task execution result, so that the final target execution result is processed by multiple AI models and multiple result evaluation nodes, which can improve the security and result accuracy of the artificial intelligence model.

[0192] The terms "first", "second", etc. in the description, claims, and drawings of the embodiments of the present application are used to distinguish different objects, rather than to describe a specific order. In addition, the term "comprising" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or equipment that includes a series of steps or units is not limited to the listed steps or modules, but optionally further includes steps or modules not listed, or optionally further includes other step units inherent to these processes, methods, devices, products, or equipment.

[0193] In the embodiments of the present application, the term "module" or "unit" refers to a computer program with a predetermined function or a part of a computer program, which works together with other relevant parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as a processing circuit or a memory), or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an overall module or unit that includes the function of that module or unit.

[0194] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in this description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0195] The methods and related devices provided in the embodiments of the present application are described with reference to the method flowcharts and / or structure diagrams provided in the embodiments of the present application. Specifically, each process and / or block of the method flowchart and / or structure diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable artificial intelligence management devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable artificial intelligence management devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or structure diagrams Figure 1 one block or multiple blocks. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable artificial intelligence management device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or structure diagrams Figure 1 one block or multiple blocks. These computer program instructions can also be loaded onto a computer or other programmable artificial intelligence management device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or structure diagrams in one block or multiple blocks.

[0196] The steps in the method of the embodiments of the present application can be adjusted, combined, and deleted according to actual needs.

[0197] The modules in the device of the embodiments of the present application can be combined, divided, and deleted according to actual needs.

[0198] The foregoing disclosure is only for the preferred embodiments of the present application, and of course cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. An artificial intelligence management method, characterized in that, The method includes: Receiving a model task published by a task publishing node, and sharing the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; Obtaining candidate task execution results of the N artificial intelligence models respectively for the model task, and sending the N candidate task execution results to a result evaluation node, so that the result evaluation node performs result detection on the N candidate task execution results; Obtaining a target task execution result reached by the result evaluation node for the N candidate task execution results, and feeding back the target task execution result to the task publishing node.

2. The method according to claim 1, characterized in that The obtaining the target task execution result reached by the result evaluation node for the N candidate task execution results includes: Obtaining result voting information of the result evaluation node for the N candidate task execution results respectively, and determining result statistical values corresponding to the N candidate task execution results respectively based on the result voting information corresponding to the N candidate task execution results respectively; Determining a target task execution result based on the result statistical values corresponding to the N candidate task execution results respectively.

3. The method according to claim 2, wherein The obtaining the result voting information of the result evaluation node for the N candidate task execution results respectively, and determining the result statistical values corresponding to the N candidate task execution results respectively based on the result voting information corresponding to the N candidate task execution results respectively includes: In the i-th round of result voting process, obtaining the (i - 1)-th result voting information generated by the result evaluation node in the (i - 1)-th round of result voting process, generating the i-th result voting information according to the (i - 1)-th result voting information, and sending the i-th result voting information generated by the first blockchain node to the second blockchain node; the first blockchain node is the blockchain node that receives the model task published by the task publishing node; the second blockchain node is the blockchain node in the result evaluation node other than the first blockchain node; i is a positive integer; If the i-th round of result voting process does not meet the voting iteration condition, obtaining the i-th result voting information corresponding to the second blockchain node, and generating the (i + 1)-th result voting information of the first blockchain node based on the i-th result voting information of the result evaluation node; If the i-th round of result voting process meets the voting iteration condition, obtaining the P-round result voting information of the result evaluation node, and counting the result statistical values corresponding to the N candidate task execution results respectively based on the P-round result voting information of the result evaluation node; P is a positive integer, and P is the number of rounds of the result voting process indicated by the voting iteration condition.

4. The method according to claim 2, wherein The obtaining the result voting information of the result evaluation node for the N candidate task execution results respectively, and determining the result statistical values corresponding to the N candidate task execution results respectively based on the result voting information corresponding to the N candidate task execution results respectively includes: Score the execution results of the N candidate tasks to obtain the first result quality scores corresponding to the execution results of the N candidate tasks respectively, and send the first result quality scores corresponding to the execution results of the N candidate tasks to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model tasks published by the task publishing node; Obtain the second result quality scores of the second blockchain node for the execution results of the N candidate tasks respectively, and determine the result statistical value of the execution result of each candidate task based on the first result quality score and the second result quality score of the execution result of each candidate task; the first result quality score and the second result quality score of the execution result of each candidate task form the result voting information of the execution result of the candidate task.

5. The method according to claim 2, wherein The obtaining the result voting information of the result evaluation nodes for the execution results of the N candidate tasks respectively, and determining the result statistical values corresponding to the execution results of the N candidate tasks respectively based on the result voting information corresponding to the execution results of the N candidate tasks includes: Obtain the first result voting information of the result evaluation nodes for the execution results of the N candidate tasks respectively, and determine the first result statistical values corresponding to the execution results of the N candidate tasks respectively based on the first result voting information corresponding to the execution results of the N candidate tasks; Sort the execution results of the N candidate tasks based on the first result statistical values corresponding to the execution results of the N candidate tasks respectively to obtain a result sequence; If the degree of difference between the first result voting information of the first two candidate task execution results in the result sequence is less than the data difference threshold, then based on the result sequence, execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively; If the similarity between the first result voting information of the first k candidate task execution results in the result sequence is greater than or equal to the data difference threshold, then generate second result voting information for the first k candidate task execution results, and determine the second result statistical values of the first k candidate task execution results respectively based on the second result voting information corresponding to the execution results of the first k candidate tasks respectively, and execute the process of determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively for the second result statistical values corresponding to the execution results of the first k candidate tasks respectively.

6. The method according to claim 2, wherein The determining the target task execution result based on the result statistical values corresponding to the execution results of the N candidate tasks respectively includes: Determine the candidate task execution result with the largest result statistical value as the result to be confirmed, and determine the result to be confirmed as the target task execution result; or, Determine the execution result of the candidate task with the largest result statistical value as the result to be confirmed, broadcast the result to be confirmed to the second blockchain node for verification, and when the result evaluation node passes the verification of the result to be confirmed, determine the result to be confirmed as the target task execution result.

7. The method according to claim 2, wherein The obtaining of the result voting information of the result evaluation node for the execution results of N candidate tasks respectively includes: Obtain the artificial intelligence model behavior constraints from the blockchain network, use the artificial intelligence model constraints to evaluate the execution results of the N candidate tasks respectively to obtain the result voting information of the first blockchain node, and send the result voting information of the first blockchain node to the second blockchain node; the second blockchain node is the blockchain node other than the first blockchain node among the result evaluation nodes; the first blockchain node refers to the blockchain node that receives the model task published by the task publishing node; Receive the result voting information of the execution results of the N candidate tasks obtained by the second blockchain node using the artificial intelligence model behavior constraints.

8. The method according to claim 2, wherein The method further includes: Record the first artificial intelligence model and the second artificial intelligence model based on the result statistical values respectively corresponding to the execution results of the N candidate tasks; the first artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the largest result statistical value among the N artificial intelligence models; the second artificial intelligence model refers to the artificial intelligence model corresponding to the execution result of the candidate task with the smallest result statistical value among the N artificial intelligence models; Manage the N artificial intelligence models based on the historical record data respectively corresponding to the N artificial intelligence models; the historical record data includes the record data for the first artificial intelligence model and the second artificial intelligence model.

9. The method according to claim 8, wherein The managing of the N artificial intelligence models based on the historical record data respectively corresponding to the N artificial intelligence models includes: Obtain the model exception times respectively corresponding to the N artificial intelligence models from the historical record data respectively corresponding to the N artificial intelligence models; the model exception times refer to the number of times when the historical result statistical value of the corresponding artificial intelligence model is the smallest; Determine the artificial intelligence model with the model exception times greater than or equal to the model exception threshold as the abnormal model, and send the abnormal model to the management node so that the management node performs detection and processing on the abnormal model.

10. The method according to claim 9, wherein The method further includes: If the first blockchain node is the management node, detect the abnormal model, determine the abnormal parameters learned in the abnormal model, delete or adjust the abnormal parameters to obtain an updated model; or obtain the model constraints for the abnormal model and write the model constraints into the abnormal model to obtain the updated model; the first blockchain node is the blockchain node that receives the model task published by the task publishing node; Upload the updated model to the blockchain network; If the first blockchain node is not the management node, execute the process of sending the abnormal model to the management node.

11. The method according to claim 1, wherein The number of the result evaluation nodes is M, where M is a positive integer; sending the N candidate task execution results to the result evaluation nodes includes: Obtaining the node credit data respectively corresponding to the M result evaluation nodes, and obtaining the credit threshold corresponding to the model task; Determining, among the M result evaluation nodes, the result evaluation nodes whose node credit data meets the credit threshold as the first result evaluation nodes, and sending the N candidate task execution results to the first result evaluation nodes.

12. The method according to claim 11, wherein The method further includes: Obtaining the deviation degrees between the result voting information respectively corresponding to the M result evaluation nodes and the target task execution result, and determining the credit update values respectively corresponding to the M result evaluation nodes based on the deviation degrees respectively corresponding to the M result evaluation nodes; any result voting information refers to the voting data of the corresponding result evaluation node for the N candidate task execution results; the result voting information respectively corresponding to the M result evaluation nodes is used to determine the target task execution result; Updating the node credit data respectively corresponding to the M result evaluation nodes by using the credit update values respectively corresponding to the M result evaluation nodes to obtain the updated credit data respectively corresponding to the M result evaluation nodes; Performing evaluation permission restriction processing on the result evaluation nodes whose updated credit data is less than the basic credit threshold.

13. The method according to claim 2, wherein The number of the result evaluation nodes is M, where M is a positive integer; the method further includes: Receiving the first digital asset transferred by the task publishing node; Obtaining the deviation degrees between the result voting information respectively corresponding to the M result evaluation nodes and the target task execution result, and determining the asset allocation coefficients respectively corresponding to the M result evaluation nodes based on the deviation degrees respectively corresponding to the M result evaluation nodes; Dividing the first digital asset into a second digital asset and a third digital asset, and transferring the second digital asset to the target artificial intelligence model corresponding to the target task execution result; Dividing the third digital asset into M node digital assets based on the asset allocation coefficients respectively corresponding to the M result evaluation nodes, and allocating the M node digital assets to the M result evaluation nodes based on the asset allocation coefficients respectively corresponding to the M result evaluation nodes.

14. An artificial intelligence management device, characterized in that, The device includes: A task sharing module, configured to receive the model task published by the task publishing node and share the model task to N artificial intelligence models; N is a positive integer; the N artificial intelligence models are independently trained; A result obtaining module, configured to obtain the candidate task execution results respectively corresponding to the N artificial intelligence models for the model task; A result sending module, configured to send the N candidate task execution results to the result evaluation nodes so that the result evaluation nodes perform result detection on the N candidate task execution results; A result determining module, configured to obtain the target task execution result reached by consensus by the result evaluation nodes for the N candidate task execution results; A result feedback module, configured to feedback the target task execution result to the task publishing node.

15. A computer device, characterized in that, Including a processor, a memory, and an input / output interface; The processor is respectively connected to the memory and the input / output interface. Among them, the input / output interface is used to receive and output data, the memory is used to store computer programs, and the processor is used to call the computer programs so that the computer device executes the method according to any one of claims 1-13.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is adapted to be loaded and executed by a processor so that a computer device having the processor executes the method according to any one of claims 1-13.

17. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the method according to any one of claims 1-13 is implemented.

Images