Method and system for sharing intelligent AI learning assistant

Abstract

The invention discloses a method and a system for sharing an intelligent AI learning assistant. The system comprises a view layer module, a business layer module, a model layer module and a storage layer module, the view layer module comprises a front-end interface, and the front-end interface is connected with all the modules and used for displaying information of all the modules and outputting operation instructions so that a user can obtain information, operate and manage the system; the business layer module is also connected with the model layer module and the storage layer module, has the functions of user registration, user sharing, user learning, Prompt sharing and the like, and is convenient for a user to carry out business editing management, knowledge sharing and self-supervised learning; the model layer module is used for constructing various learning assistant models; and the storage layer module comprises a short-term memory database, a long-term memory database and a block chain account book and is used for storing, managing and tracking data of various learning assistants. The learning assistant system has the intelligent learning characteristics of sharing, learning and tracking.

Description

Technical Field

[0001] This invention relates to the field of AI learning assistant technology, specifically to a shared intelligent AI learning assistant method and system. Background Technology

[0002] Current learning assistants provide learning services to users through pre-recorded videos and human assistance. A schematic diagram of the relevant architecture is provided below. Figure 10 As shown, current learning assistants suffer from drawbacks such as limited domain knowledge, concentrated information, and low user engagement. These drawbacks are further elaborated below:

[0003] <1> Current learning assistant applications teach users through video lectures and human assistance, but they do not yet provide comprehensive support to address the specific needs of students with certain characteristics.

[0004] <2> Current learning assistant applications have domain-specific closedness; for example, New Oriental's English learning assistant is merely internal learning material for the company.

[0005] <3> Current learning assistants lack knowledge-sharing features;

[0006] <4> Current learning assistants lack relevant incentive mechanisms;

[0007] <5> The knowledge of current learning assistants has not yet been certified and endorsed by most expert strategies.

[0008] As mentioned above, while the saying "Among three people walking together, there must be one who can teach me" holds true, current learning assistants not only fail to mobilize experts from various fields for knowledge sharing, but also fail to call upon experts in similar fields for knowledge sharing. Furthermore, current technical architectures of learning assistants do not categorize and analyze students' questions, nor do they provide recommendations based on students' relevant questions. Therefore, a method for developing a shareable, learnable, and trackable AI learning assistant is needed. Summary of the Invention

[0009] To address the aforementioned technical shortcomings, the purpose of this invention is to provide a shared, intelligent AI learning assistant method and system.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0011] This invention provides a shared intelligent AI learning assistant method and system, characterized by including, but not limited to, a view layer module, a business layer module, a model layer module, and a storage layer module; the view layer module includes a front-end interface, which is connected to the business layer module, the model layer module, and the storage layer module respectively; the business layer module is also connected to the model layer module and the storage layer module; the model layer module and the storage layer module are connected; the front-end interface is used to display information from the above modules and output operation control commands for the above modules, allowing users to obtain information and operate and manage the system;

[0012] The business layer modules include, but are not limited to, user verification units, user incentive units, user sharing units, knowledge sharing units, knowledge recommendation units, and knowledge verification units. These business layer modules have multiple functions, including but not limited to user registration, user sharing, user learning, prompt sharing, and user incentive functions, which facilitate users in business editing and management, knowledge sharing, and self-supervised learning.

[0013] The model layer module includes, but is not limited to, AI deep learning models, which include problem diffusion models, incentive mechanism models, user sharing models, MOE large language models, recommendation learning models, and endorsement strategy models. The model layer module is used to construct various learning assistant models, which include, but are not limited to, AI deep learning models, as well as federated learning models involved in privacy computing.

[0014] The storage layer module includes, but is not limited to, a short-term memory database, a long-term memory database, and a blockchain ledger. The storage layer module is used to store, manage, and track various learning assistant data.

[0015] This also includes, but is not limited to, methods for implementing user registration, user sharing, Prompt sharing, and user learning functions.

[0016] Furthermore, the method for implementing the user registration function includes, but is not limited to, the following steps:

[0017] a1. Create a blockchain network;

[0018] b1. The user authorizes the learning assistant;

[0019] c1. The learning assistant creates a DID digital identity on the blockchain network;

[0020] d1. Regulatory agencies need to verify identity information;

[0021] e1. The user uploads the certificate information to the blockchain network and generates the proof digest information of the DID;

[0022] f1. The certificate verification party will verify the validity of the certificate;

[0023] g1. User identity generation.

[0024] Furthermore, the method for implementing the user sharing function includes, but is not limited to, the following steps:

[0025] a2. Verify user identity;

[0026] b2. Users share knowledge;

[0027] c2. Use a knowledge classification model to classify knowledge;

[0028] d2. Endorsement by domain experts;

[0029] e2. Store the knowledge and expert endorsement information on the blockchain for evidence.

[0030] f2. Update the domain knowledge distance relationship;

[0031] g2. Store knowledge in distributed storage memory for users to retrieve.

[0032] Furthermore, the method for implementing the Prompt sharing function includes, but is not limited to, the following steps:

[0033] a4. Verify user identity;

[0034] b4. Verify whether the user has the skills and certification information to write excellent Prompts;

[0035] c4. Select the knowledge domain category and question type;

[0036] d4. Verify the Prompt using POS consensus;

[0037] e4. Check if the verification is successful;

[0038] f4. Perform on-chain evidence storage;

[0039] g4. Open for use.

[0040] Furthermore, the method for implementing the user learning function includes, but is not limited to, the following steps:

[0041] a3. The user raises a question;

[0042] b3. The Agent breaks down the problem;

[0043] c3. Search the decomposed atomic problems in the distributed vector library;

[0044] d3. Feed the user-provided prompt and retrieval context into the MOE large language model;

[0045] e3. The language model analyzes based on the "thought chain" and answers questions through the "action chain".

[0046] f3. Convert the text of the answer to the question into the voice of an authoritative expert;

[0047] g3. Combine the voice of an authoritative expert with a video or image recorded by the authoritative expert to create a high-definition video;

[0048] h3. Send the instructional video to the streaming media server;

[0049] i3. Users learn through videos.

[0050] The beneficial effects of this invention are as follows:

[0051] (1) From the perspective of knowledge sharing: it can not only mobilize the active sharing of knowledge among all participants, but also ensure the accuracy of knowledge by participating in the "expert endorsement and verification" strategy.

[0052] (2) From the perspective of student learning: it can not only provide explanations and sharing of knowledge points by various teachers, but also promote students' understanding of knowledge through knowledge adaptation recommendation algorithms;

[0053] (3) From the perspective of user security: not only is the security of user information guaranteed by encryption, but also the security of expert certificate information is guaranteed by selective leakage. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A schematic diagram of the overall architecture logic of a shared intelligent AI learning assistant method and system provided by the present invention;

[0056] Figure 2 A schematic diagram of the user registration logic for a shared intelligent AI learning assistant method and system provided by the present invention;

[0057] Figure 3 A schematic diagram of the user sharing module of a shared intelligent AI learning assistant method and system provided by the present invention;

[0058] Figure 4 A schematic diagram illustrating the expert endorsement strategy logic of a shared intelligent AI learning assistant method and system provided by this invention;

[0059] Figure 5 A logical diagram illustrating the distributed storage strategy of a shared intelligent AI learning assistant method and system provided by the present invention;

[0060] Figure 6 A schematic diagram of the Prompt sharing logic of a shared intelligent AI learning assistant method and system provided by the present invention;

[0061] Figure 7 A schematic diagram of the user learning logic of a shared intelligent AI learning assistant method and system provided by the present invention;

[0062] Figure 8 A schematic diagram of the Prompt selection logic of a shared intelligent AI learning assistant method and system provided by the present invention;

[0063] Figure 9 A flowchart illustrating the self-supervised learning module of a shared intelligent AI learning assistant method and system provided by this invention;

[0064] Figure 10 A schematic diagram of the architecture of an existing learning assistant provided by the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example 1

[0067] like Figures 1 to 9 As shown, this invention provides a shared intelligent AI learning assistant method and system, wherein, reference... Figure 1 The overall architecture diagram shown illustrates that this invention provides a shared intelligent AI learning assistant system, which includes, but is not limited to, a view layer module, a business layer module, a model layer module, and a storage layer module.

[0068] Preferably, the view layer module includes a front-end interface, which is connected to the business layer module, the model layer module, and the storage layer module respectively. The business layer module is also connected to the model layer module and the storage layer module. The front-end interface is used to display information of the above-mentioned modules and output operation control commands for the above-mentioned modules, so that users can obtain information and operate and manage the system.

[0069] Preferably, the business layer module includes, but is not limited to, a user verification unit, a user incentive unit, a user sharing unit, a knowledge sharing unit, a knowledge recommendation unit, and a knowledge verification unit. The business layer module provides multiple functions, including but not limited to user registration, user sharing, user learning, prompt sharing, and user incentive functions, facilitating user editing and management, knowledge sharing, and self-supervised learning. Specifically, the self-supervised learning refers to the automatic analysis of user B's knowledge weaknesses and error set by the backend based on user A's knowledge sharing and user B's learning and problem tracking after user A shares knowledge.

[0070] Preferably, the model layer module includes, but is not limited to, AI deep learning models, which include problem diffusion models, incentive mechanism models, user sharing models, MOE large language models, recommendation learning models, and endorsement strategy models. The model layer module is used to construct various learning assistant models, which include, but are not limited to, AI deep learning models, and also include federated learning models involved in privacy computing.

[0071] Preferably, the storage layer module includes, but is not limited to, a short-term memory database, a long-term memory database, and a blockchain ledger, and is used to store, manage, and track various learning assistant data.

[0072] In this embodiment, reference Figures 2 to 9 The present invention also provides a method for sharing an intelligent AI learning assistant based on the AI ​​learning assistant described above. This AI learning assistant method includes, but is not limited to, a method for implementing user registration, a method for implementing user sharing, a method for implementing Prompt sharing, and a method for implementing user learning.

[0073] In this embodiment, reference Figure 2 The user registration logic diagram shown below illustrates that the user registration function implementation method includes, but is not limited to, the following steps:

[0074] a1. Create a blockchain network;

[0075] b1. The user authorizes the learning assistant. Here, "authorization" is a proxy authorization, that is, the user authorizes the learning assistant to verify relevant information.

[0076] c1. The learning assistant creates a DID digital identity on the blockchain network;

[0077] d1. Regulatory agencies need to verify identity information. For example, public security organs will verify the authenticity of a user's identity, and certificate authorities (such as universities) will verify the authenticity of certificates. This ensures that various regulatory agencies guarantee the authenticity of user identities and verify whether users are genuine experts.

[0078] e1. Users upload their certificate information to the blockchain network, generating a proof digest of their DID. Once user information is on the blockchain, its immutability is ensured. Specifically, the certificate and user are added to the proof digest of the DID identity; the learning assistant's encryption system ensures that information stored on the blockchain is kept private and cannot be leaked.

[0079] f1. The certificate verification party will verify the validity of the certificate;

[0080] g1. User identity generation, i.e., the formal generation of the DID identity. As the user's skills improve, the user will obtain new skill certificates. The user simply updates the skill certificates to the proof digest of the DID. This ensures the immutability of the information.

[0081] The user registration function mentioned above requires clarification as follows:

[0082] (1) Users can generate student and teacher identities based on their registration information and identification information;

[0083] (2) Using DID digital identity and blockchain networks can ensure that user information is tamper-proof;

[0084] (3) Using encrypted authentication and selective leakage can ensure that users can verify the authenticity of teacher users' certificates and view public information about teacher certificates.

[0085] In summary, the logical architecture of the AI ​​learning assistant system described above ensures that the user can guarantee the authority of their knowledge in a specific field. Furthermore, it selects experts in that field based on the authority of their certificates and obtains their endorsement.

[0086] In this embodiment, reference Figure 3 The diagram shown illustrates the logic of the user sharing module. The method for implementing the user sharing function includes, but is not limited to, the following steps:

[0087] a2. Verify user identity;

[0088] b2. Users share knowledge;

[0089] c2. Use a knowledge classification model to classify knowledge, such as classifying knowledge into English test questions, and then classifying the question types, such as cases that focus on testing the word 'a'.

[0090] d2. Endorsement by domain experts;

[0091] e2. Store the knowledge and expert endorsement information on the blockchain for evidence.

[0092] f2. Update the distance relationships of domain knowledge. For example, by updating the distance relationships, all questions in the English test that test the word 'a' will be very close in distance.

[0093] g2. Store knowledge in distributed storage memory for users to retrieve.

[0094] The following points need to be noted regarding the aforementioned user sharing function:

[0095] (1) Verifying user identity can ensure that the knowledge they share has the corresponding authority of the author in the field;

[0096] (2) Use the “expert endorsement” strategy to endorse the knowledge shared by users and ensure that the shared knowledge is more accurate;

[0097] (3) Knowledge classification can be used to identify which experts in the field endorse the knowledge;

[0098] (4) Knowledge distance relationship update is to ensure the similarity between the knowledge shared by users and other knowledge in storage;

[0099] (5) The understanding of knowledge classification is: to divide the shared knowledge into subject type and question type.

[0100] For details, please refer to Figure 4 The diagram shown illustrates the logic of the expert endorsement strategy. In this strategy, the certificate information provided by the user during registration can guarantee whether the expert has the status of an expert, and the experts are ranked according to the value of their certificates.

[0101] The following points need to be clarified regarding the aforementioned expert endorsement strategy:

[0102] (1) The domain expert network consists of domain experts who hold relevant certificates, and 2n+1 experts are selected (ensuring that the number is odd) based on the value of the certificate information.

[0103] (2) 2n+1 experts verify and reach consensus on the knowledge shared by the user (they need to verify that the knowledge is correct on their own expert network nodes), and then endorse the knowledge.

[0104] (3) The final output is the domain knowledge and expert network endorsement information. You can see which experts are endorsing the domain and their publicly available certification information through the endorsement information;

[0105] (4) Verify the knowledge through a domain phishing network that uses expert consensus (i.e., users holding domain certificates);

[0106] (5) If the consensus is not reached immediately, the system will re-enter the expert network for consensus. If this situation occurs three times, the knowledge will be discarded.

[0107] (6) If approved, users who share knowledge and experts who endorse the network will receive token rewards.

[0108] (7) If the phishing network fails, the experts in the network will be rewarded with tokens, while the users who share knowledge and the experts who endorse the network will be penalized with tokens.

[0109] For details, please refer to Figure 5 The diagram illustrates the logic of the distributed storage strategy, which employs a Retrieval-Enhanced Generation (RAG) scheme for distributed storage memory. RAG retrieves relevant knowledge through retrieval and integrates it into the large model's Prompt, allowing the large model to refer to the relevant knowledge and provide a reasonable answer. Therefore, the core of RAG can be understood as "retrieval + generation." "Retrieval" primarily utilizes the efficient storage and retrieval capabilities of the vector database to recall target knowledge; "generation" mainly utilizes the large model and the Prompt project to rationally utilize the recalled knowledge and generate the target answer. The data preparation phase of the distributed storage record application process is as follows: data extraction -> text segmentation -> vectorization (embedding) -> data storage.

[0110] The following points need to be explained regarding the aforementioned distributed storage strategy:

[0111] (1) This part uses various methods such as distributed vector databases to store vectors and maintain the distance relationship between vectors;

[0112] (2) Based on the prompt of the user's question and the knowledge retrieval results, the MOE expert big language model will return the accurate answer to the user.

[0113] In this embodiment, the Prompt of the present invention can be considered from the following two categories, thereby avoiding the situation where too many tasks can be performed on the same sentence, leading to the Prompt being unsure which task to instruct the model to perform. The two categories considered by the Prompt are as follows:

[0114] (1) Answer hint type Prompt: Design the Prompt according to the task objective and the type of answer, and use the Prompt to elicit the answer;

[0115] (2) Task prompt type: reminds the model what task to do.

[0116] refer to Figure 6 The diagram shown illustrates the Prompt sharing logic. The implementation method of the Prompt sharing function includes, but is not limited to, the following steps:

[0117] a4. Verify user identity;

[0118] b4. Verify whether the user has the skills and certification information to write excellent Prompts;

[0119] c4. Select the knowledge domain category and question type; when performing this step, it is necessary to verify whether the user has a certificate in the knowledge domain and whether the user has a certificate in the skills to write a Prompt.

[0120] d4. Verify the Prompt through POS consensus. This step mainly involves ranking the value of the certificate in the relevant field plus the value of the Prompt skill certificate. The higher the score, the greater the benefits. Therefore, POS here is a ranking of the benefits of both skills. Experts in different fields will have different experts in their POS.

[0121] e4. Check if the verification is successful; This step mainly involves verifying the POS consensus achieved in step d4 by having relevant domain experts endorse it.

[0122] f4. Perform on-chain evidence storage, which mainly involves storing the user-written Prompt and expert endorsements on the blockchain.

[0123] g4. Open for use.

[0124] The above-described method for implementing the Prompt sharing function provides a way for users to share Prompts during their learning process, avoiding situations where experts in each field lack the skills to write excellent Prompts and therefore cannot use them.

[0125] Specifically, the following points need to be explained regarding the aforementioned Prompt sharing function:

[0126] (1) DID authentication can verify whether a user has the skill to write a Prompt;

[0127] (2) The explanation of domain knowledge classification and question classification is as follows: Domain knowledge includes, but is not limited to, classifications such as Nth grade Chinese, Nth grade mathematics, postgraduate entrance examination English, etc.; and the questions in each knowledge classification are further divided into several question types;

[0128] (3) Use POS consensus to ensure that this Prompt is recognized by the most excellent experts. If a user uses this Prompt when learning, the user will be given token rewards. As the number of tokens increases, the POS rights and interests of the user become greater;

[0129] (4) After opening it for external use, users can choose to use the Prompt for learning.

[0130] In this embodiment, referring to Figure 7 the user learning logic schematic diagram shown, the method for implementing the user learning function includes, but is not limited to, the following steps:

[0131] a3. The user asks a question;

[0132] b3. The Agent decomposes the question;

[0133] c3. Retrieve the decomposed atomic questions in the distributed vector library;

[0134] d3. Feed the Prompt provided by the user and the retrieval context into the MOE large language model;

[0135] e3. The language model analyzes in the way of "chain of thought" and answers the question through the "chain of action";

[0136] f3. Convert the text answering the question into the voice of an authoritative expert;

[0137] g3. Synthesize the voice of the authoritative expert + the video (or picture) recorded by the authoritative expert into a high-definition video;

[0138] h3. Send the explanatory video to the streaming media server;

[0139] i3. The user learns through the video.

[0140] Among them, it should be noted that the above user learning function:

[0141] (1) The MOE model here supports multi-modal, which can be voice, text, image, video, etc. For example, if the user inputs "The torrent dashes down three thousand feet", the model will generate a diffusion model picture with a waterfall to deepen the understanding of student users;

[0142] (2) Students can get different response formats by using different prompts. If there is no publicly available option in the selection list, students can use the image scan provided by the system to automatically generate their own prompt from the case templates in the textbook;

[0143] (3) The generated video will be generated by an authoritative teacher through a problem-solving thinking chain.

[0144] For details, please refer to Figure 8 The diagram shown illustrates the Prompt selection logic. The steps of the Prompt selection module include, but are not limited to, the following:

[0145] (1) Users select the knowledge domain category and question type;

[0146] (2) Select the Prompt for this type of question;

[0147] (3) If a publicly available Prompt is selected, the Prompt and the problem will be sent to the downstream task together; (4) If a publicly available Prompt is not selected, the exercise cases in the reference book can be scanned and the assistant model can be used to automatically generate the Prompt corresponding to the problem.

[0148] (5) Send the newly generated Prompt and issue to the downstream task.

[0149] The following points need to be noted regarding the Prompt selection module mentioned above:

[0150] (1) The Prompt list displays two columns: Prompt and Number of Uses. Since the number of uses is displayed for each user, users can choose based on the publicly available Prompts. They can either default to the most effective one or choose the most frequently used one.

[0151] (2) Users can scan the question sample templates in the reference book, and the system will automatically generate the user's prompt based on the AI ​​model in the system.

[0152] Specifically, the steps of the knowledge tracking module include, but are not limited to, the following:

[0153] (1) Locating user profile information, that is, since the user profile is basically determined when the user registers, the user's knowledge profile can be located through the user's self-learning.

[0154] (2) Analyze the user's learning problems and classify and statistically analyze them. That is, the system will automatically track the user's problem history, use the model to share the user's knowledge classification and problem types, and then update their profile.

[0155] (3) Use recommendation algorithms to recommend this type of problem to users, that is, after determining the user profile, push the user's weak points + further problems to the user;

[0156] The following points need to be noted regarding the knowledge tracking module mentioned above:

[0157] (1) When a user shares knowledge, the distance relationship between that knowledge and other knowledge has been updated. Therefore, the knowledge with the closest distance relationship is of the same type.

[0158] (2) This module allows users to categorize and organize the errors they have made previously;

[0159] For details, please refer to Figure 9 The diagram shown illustrates the logic of the self-supervised learning module. This module primarily organizes large amounts of data based on the questioning history of users and students, and then sends it to a large MOE (Multiple Objects) model for training.

[0160] The self-supervised learning module mentioned above needs to be explained as follows:

[0161] (1) The system will automatically perform data analysis based on user behavior. For example, it will clean the user's dialogue dataset and then train the MOE language model; it will categorize and organize the user's questions and answers to form the user's error set; and it will recommend exercises that the user has not participated in based on the user's set ratio.

[0162] This invention utilizes an AI learning assistant system, which possesses intelligent learning characteristics that are shareable, learnable, and traceable.

[0163] Specifically:

[0164] (1) Establish a bridge between learning assistants in various fields so that they can obtain the courses in the desired field through only the learning assistant;

[0165] (2) Provide information verification for contributors to domain knowledge to ensure that their knowledge is not copied;

[0166] (3) Use domain knowledge-based expert endorsement strategies to improve the accuracy of content;

[0167] (4) Using self-learning, self-searching and self-tracking methods can ensure improved learning effectiveness;

[0168] (5) Use retrieval enhancement methods to address the limitations of knowledge, the illusion of large models, and data security issues;

[0169] (6) Using DID authentication and selective leakage can help domain experts improve the quality of their knowledge and prevent the leakage of important information about them.

[0170] (7) Achieve full-process management of the user's "thinking chain" to "action chain".

[0171] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method and system for sharing intelligent AI learning assistant, characterized in that, It includes but is not limited to view layer module, business layer module, model layer module, storage layer module; The view layer module includes front-end interface, and the front-end interface is connected with the business layer module, the model layer module and the storage layer module respectively, the business layer module is also connected with the model layer module and the storage layer module, and the model layer module is connected with the storage layer module; Through the front-end interface, the information of the above various modules is displayed, and the operation control instruction of the above various modules is output, so that the user can obtain information and operation, management system; The business layer module includes but is not limited to user authentication unit, user incentive unit, user sharing unit, knowledge sharing unit, knowledge recommendation unit and knowledge authentication unit, and the business layer module has multiple functions, which include but are not limited to user registration function, user sharing function, user learning function, Prompt sharing function and user incentive function, so that the user can conveniently edit and manage business, share knowledge and self-supervised learning; The model layer module includes but is not limited to AI deep learning model, the AI deep learning model includes problem diffusion model, incentive mechanism model, user sharing model, MOE large language model, recommended learning model and endorsement strategy model, and the model layer module is used to build various learning assistant models, and the learning assistant model includes but is not limited to AI deep learning model, and also includes federal learning model related to privacy calculation; The storage layer module includes but is not limited to short-term memory database, long-term memory database and blockchain ledger, and the storage layer module is used to store, manage and track various learning assistant data; It also includes but is not limited to user registration function implementation method, user sharing function implementation method, Prompt sharing function implementation method and user learning function implementation method.

2. The shared intelligent AI learning assistant method and system of claim 1, wherein: The user registration function implementation method includes but is not limited to the following steps: a1, creating a blockchain network; b1, the user authorizes the learning assistant; c1, the learning assistant creates a DID digital identity on the blockchain network; d1, the supervisory authority needs to verify the identity information; e1, the user uploads the information of the certificate to the blockchain network and generates the proof digest information of DID; f1, the certificate verification party verifies the validity of the certificate; g1, the user identity is generated.

3. The shared intelligent AI learning assistant method and system of claim 1, wherein: The user sharing function implementation method includes but is not limited to the following steps: a2, verifying the user identity; b2, the user shares knowledge; c2, using a knowledge classification model to classify knowledge; d2, the field expert endorses; e2, the knowledge and expert endorsement information are stored on the chain; f2, update the field knowledge distance relationship; g2, store the knowledge in distributed storage memory for user retrieval.

4. The shared intelligent AI learning assistant method and system of claim 1, wherein: The Prompt sharing function implementation method includes but is not limited to the following steps: a4, verify the user's identity; b4, verify whether the user has the skills and certificate information to write good prompts; c4, select the knowledge field classification and question type; d4, verify the prompt through POS consensus; e4, check whether it passes the verification; f4, perform on-chain storage; g4, open for use.

5. The shared intelligent AI learning assistant method and system of claim 1, wherein the user learning function implementation method comprises the following steps: a3, the user raises a question; b3, the agent decomposes the question; c3, search the decomposed atomic question in the distributed vector library; d3, feed the user-provided prompt and search context to the MOE large language model; e3, the language model analyzes according to the "thinking chain" mode and answers the question through the "action chain"; f3, convert the text of the answer to the question into the voice of an authoritative expert; g3, synthesize high-definition video of the voice of the authoritative expert + video or pictures recorded by the authoritative expert; h3, send the explanation video to the streaming media server; i3, the user learns through the video. ​

Images