Artificial intelligence teaching method and teaching robot based on large model

Abstract

According to the artificial intelligence teaching method based on the large model and the teaching robot, the generation type large model is used for learning the content needing to be taught, and the robot can be driven to carry out targeted teaching according to the content needing to be taught, the question content of students and the characteristics of the students. And meanwhile, the robot can be driven to make explanation expressions and explanation actions similar to those of human teachers, so that the robot can assist and even replace the teachers to teach courses.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence teaching method and a teaching robot based on a large model. Background Art

[0002] Robots assisting human teachers in teaching can help improve teaching standards.

[0003] In the process of realizing the present invention, the inventors found that there are at least the following problems in the prior art: the teaching robots in the prior art generally read from the textbook, but are unable to provide targeted explanations of the courses based on the students' questions.

[0004] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0005] Based on this, it is necessary to provide an artificial intelligence teaching method and teaching robot based on a large model to address the defects or shortcomings of the existing technology, so that the robot can provide targeted teaching according to the characteristics of students, their problems, and courses, and can simulate the teaching words and deeds of human teachers, thereby improving the efficiency and effectiveness of artificial intelligence robots in teaching courses and improving students' listening experience.

[0006] In a first aspect, an embodiment of the present invention provides a deep learning method, the method comprising:

[0007] Teaching content generation step: taking the course information and the questions raised by the students as the input of the course teaching model, and taking the output as the teaching content for the questions raised, and teaching the answers to the students who raised the questions by the robot;

[0008] Content video display steps: obtain course information, course photos, and the teaching content corresponding to the questions raised as input, and the output obtained by calculating the course visual model is used as the content video of the questions raised; the content video is displayed on the screen by the robot.

[0009] Preferably, the method further comprises:

[0010] Lecture video display step: obtain the lecturer selected by the student, the lecturer's information, the lecturer's photo, and the lecture content as the input of the lecturer's lecture video model, and the output is the lecture video of the lecturer corresponding to the lecture content;

[0011] The robot teaching action control steps are as follows: the facial image of the lecturer in the teaching video is displayed by the robot on the robot's head screen; the teaching video is input into the robot teaching model, and the calculated output is used as the action control instruction to control the robot's action.

[0012] Preferably, the method further comprises:

[0013] Course teaching big model pre-training step: taking course materials (including course information, detailed course introduction, lecturer information, lecturer's remarks, works, and other related data) as pre-training data of the generative big model, pre-training the generative big model, and obtaining the pre-trained course teaching big model;

[0014] Steps for fine-tuning the instructions of the course teaching model: Based on the pre-trained course teaching model, add M preset parameters, obtain the course information and questions raised as input, obtain the teaching content as the expected output, fine-tune the M preset parameters in the pre-trained course teaching model, and obtain the course teaching model after instruction fine-tuning;

[0015] The steps of aligning and fine-tuning the course teaching model are as follows: obtain the course information and questions raised as the input of the course teaching model after instruction fine-tuning, show the output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

[0016] Preferably, the method further comprises:

[0017] Steps for fine-tuning the instructions of the course visual big model: based on the generative big model to be transferred, add M preset parameters, obtain the course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, fine-tune the preset M parameters in the generative big model to be transferred, and obtain the course visual big model after instruction fine-tuning;

[0018] The steps of aligning and fine-tuning the course visual big model are as follows: obtain the course information, course photos, and teaching content as the input of the course visual big model after instruction fine-tuning, show the obtained output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course visual big model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course visual big model.

[0019] In a second aspect, an embodiment of the present invention provides an artificial intelligence system, the system comprising:

[0020] Teaching content generation module: takes the course information and the questions raised by the students as the input of the course teaching model, and the output is used as the teaching content for the questions raised. The robot teaches the answers to the students who raised the questions;

[0021] Content video display module: obtains course information, course photos, and the teaching content corresponding to the questions raised as input, and the output obtained through calculation of the course visual model is used as the content video of the questions raised; the content video is displayed on the screen by the robot.

[0022] Preferably, the system further comprises:

[0023] Lecture video display module: obtain the lecturer selected by the student, the lecturer's information, the lecturer's photo, and the lecture content as the input of the lecturer's lecture video model, and the output is the lecture video of the lecturer corresponding to the lecture content;

[0024] Robot teaching motion control module: The robot displays the facial image of the lecturer in the teaching video on the robot's head screen; the teaching video is input into the robot teaching model, and the calculated output is used as the motion control instruction to control the robot's motion.

[0025] Preferably, the system further comprises:

[0026] Course teaching big model pre-training module: taking course materials (including course information, detailed course introduction, lecturer information, lecturer's remarks, works, and other related data) as pre-training data of the generative big model, pre-training the generative big model, and obtaining the pre-trained course teaching big model;

[0027] Course teaching model instruction fine-tuning module: Based on the pre-trained course teaching model, add M preset parameters, obtain course information and questions as input, obtain teaching content as expected output, fine-tune the preset M parameters in the pre-trained course teaching model, and obtain the course teaching model after instruction fine-tuning;

[0028] Course teaching model alignment and fine-tuning module: obtain course information and questions as input of the course teaching model after instruction fine-tuning, display the output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

[0029] Preferably, the system further comprises:

[0030] Course visual big model instruction fine-tuning module: Based on the generative big model to be transferred, add M preset parameters, obtain course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, and fine-tune the preset M parameters in the generative big model to be transferred to obtain the course visual big model after instruction fine-tuning;

[0031] Course visual big model alignment and fine-tuning module: obtain course information, course photos, and teaching content as input of the course visual big model after instruction fine-tuning, display the obtained output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course visual big model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course visual big model.

[0032] In a third aspect, an embodiment of the present invention provides an artificial intelligence device, which includes a module of the system described in any one of the embodiments of the second aspect.

[0033] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method described in any one of the embodiments of the first aspect.

[0034] In a fifth aspect, an embodiment of the present invention provides an artificial intelligence robot, comprising a memory, a processor, and an artificial intelligence robot program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods described in the first aspect when executing the program.

[0035] The big model-based artificial intelligence teaching method and teaching robot provided in this embodiment use the generative big model to learn the content to be taught, and can drive the robot to provide targeted teaching based on the content to be taught, the content of the students' questions, and the characteristics of the students. At the same time, it can drive the robot to make explanation expressions and explanation movements similar to those of human teachers, so that the robot can assist or even replace the teacher in teaching the course. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flowchart of a deep learning method provided by one embodiment of the present invention;

[0037] Figure 2 A framework diagram of an artificial intelligence system provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present invention are described in detail below in conjunction with the embodiments of the present invention.

[0039] Basic Embodiments of the Invention

[0040] First, as Figure 1As shown, an embodiment of the present invention provides a deep learning method, the method comprising: a teaching content generation step; a content video display step. In a preferred embodiment, the method further comprises: a teaching video display step; a robot teaching action control step. In a preferred embodiment, the method further comprises: a course teaching large model pre-training step; a course teaching large model instruction fine-tuning step; a course teaching large model alignment fine-tuning step. In a preferred embodiment, the method further comprises: a course visual large model instruction fine-tuning step; a course visual large model alignment fine-tuning step.

[0041] Second, as Figure 2 As shown, an embodiment of the present invention provides an artificial intelligence system, the system comprising: a teaching content generation module; a content video display module. In a preferred embodiment, the system further comprises: a teaching video display module; a robot teaching action control module. In a preferred embodiment, the system further comprises: a course teaching large model pre-training module; a course teaching large model instruction fine-tuning module; a course teaching large model alignment fine-tuning module. In a preferred embodiment, the system further comprises: a course visual large model instruction fine-tuning module; a course visual large model alignment fine-tuning module.

[0042] Preferred embodiments of the present invention

[0043] There is an intelligent teaching robot in the teaching hall. It looks like a lecturer and can give lectures on the spot. It can also communicate with students and give instructions based on students' questions, just like a reborn human.

[0044] The course teaches the steps for initializing the big model: obtain the trained and tested AI question-answering big model (including the trained and tested GPT, Bert, LLAMA, etc.) as the generative big model to be transferred for learning.

[0045] Transfer learning steps for large models for course teaching: Through transfer learning, the large model can transfer the knowledge learned from a wider range of content texts to the large model for course teaching.

[0046] Course teaching model pre-training step: taking course materials (including course information, detailed course introduction, lecturer information, lecturer's speeches, works, and other related data) as pre-training data of the generative model, pre-training the generative model, and obtaining the pre-trained course teaching model. Course information includes course name and course introduction. Lecturer information includes lecturer's name and lecturer's introduction.

[0047] Pre-training can enable the big model to learn the knowledge of the course and prepare it for answering questions in the course.

[0048] The steps of fine-tuning the instructions of the course teaching model are as follows: fix the preset K layers in the pre-trained course teaching model, obtain the course information and the questions raised as input, obtain the teaching content as the expected output, and fine-tune the parameters of the unfixed layers in the pre-trained course teaching model to obtain the course teaching model after instruction fine-tuning; wherein, the questions raised and the teaching content include questions raised by students and the teaching content of the lecturer, and also include preset questions raised and manually annotated teaching content.

[0049] Or, based on the pre-trained course teaching model, add preset M parameters, obtain the course information and questions raised as input, obtain the teaching content as the expected output, fine-tune the preset M parameters in the pre-trained course teaching model, and obtain the course teaching model after instruction fine-tuning.

[0050] By fine-tuning the instructions, the large model can learn how to make teaching content based on the course information and questions raised.

[0051] The instruction fine-tuning steps of the course teaching model are as follows: fix the preset K layers in the course teaching model after instruction fine-tuning, obtain the student's characteristics, course information, and questions raised as input, obtain the teaching content as the expected output, and fine-tune the parameters of the unfixed layers in the course teaching model after instruction fine-tuning to obtain the course teaching model after instruction fine-tuning;

[0052] Or, on the basis of the course teaching model after fine-tuning the instructions, add preset M parameters, obtain the student's characteristics, course information, and questions raised as input, obtain the teaching content as the expected output, and fine-tune the preset M parameters in the course teaching model after fine-tuning the instructions to obtain the course teaching model after fine-tuning the instructions.

[0053] Through fine-tuning of instructions, the large model can learn how to make teaching content based on students' characteristics, course information, and questions raised.

[0054] The step of deep tuning the instruction of the course teaching model is as follows: fix the preset K layers in the course teaching model after fine tuning the instruction, obtain the information of the lecturer, the characteristics of the students, the information of the course, and the questions raised as input, obtain the teaching content as the expected output, and fine-tune the parameters of the unfixed layers in the course teaching model after fine tuning the instruction to obtain the course teaching model after deep tuning the instruction;

[0055] Or, on the basis of the course teaching model after fine-tuning the instructions, add preset M parameters, obtain the information of the lecturer, the characteristics of the students, the information of the course, and the questions raised as input, obtain the teaching content as the expected output, and fine-tune the preset M parameters in the course teaching model after fine-tuning the instructions to obtain the course teaching model after deep-tuning the instructions.

[0056] Through deep instruction tuning, the large model can learn how to make teaching content based on the lecturer's information, students' characteristics, course information, and questions raised.

[0057] The steps of aligning and fine-tuning the course teaching model are as follows: obtain the course information and questions raised as the input of the course teaching model after instruction fine-tuning, show the output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

[0058] Through alignment and fine-tuning, the large model can learn how to make the teaching content satisfy the users based on the course information and questions raised.

[0059] The steps for aligning and fine-tuning the course teaching model are as follows: obtain the student's characteristics, course information, and questions as the input of the course teaching model after instruction fine-tuning, show the output to the students or evaluators, obtain feedback from the students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

[0060] Through alignment and fine-tuning, the large model can learn how to make the teaching content satisfy the users based on the students' characteristics, course information, and questions raised.

[0061] The steps of aligning and deep tuning the course teaching big model are as follows: obtain the information of the lecturer, the characteristics of the students, the information of the course, and the questions raised as the input of the course teaching big model after instruction deep tuning, show the obtained output to the students or evaluators, obtain feedback from the students or evaluators, and perform reinforcement learning on the course teaching big model after instruction deep tuning based on the feedback to obtain the aligned and deep tuned course teaching big model.

[0062] Through alignment and deep tuning, the large model can learn how to make the teaching content satisfy the users based on the lecturer's information, students' characteristics, course information, and questions raised.

[0063] Steps for initializing the course visual big model: obtain the trained and tested artificial intelligence literary video big model (including the trained and tested Sora, MuseV, Latte, etc.) as the generative big model to be transferred learning.

[0064] Through transfer learning, the large model can transfer the knowledge learned from a larger range of cultural video datasets to the course visual large model.

[0065] Get the lecture video of the course and extract the lecture content from the audio of the lecture video.

[0066] Steps for fine-tuning the instructions of the course visual big model: fix the preset K layers in the generative big model to be transferred, obtain the course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, fine-tune the parameters of the unfixed layers in the generative big model to be transferred, and obtain the course visual big model after instruction fine-tuning;

[0067] Or on the basis of the generative big model to be transferred learning, add preset M parameters, obtain course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, fine-tune the preset M parameters in the generative big model to be transferred learning, and obtain the visual big model of the course after instruction fine-tuning.

[0068] Through instruction fine-tuning, the large model can learn how to generate teaching videos based on course information, course photos, and teaching content.

[0069] The steps of aligning and fine-tuning the course visual big model are as follows: obtain the course information, course photos, and teaching content as the input of the course visual big model after instruction fine-tuning, show the obtained output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course visual big model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course visual big model.

[0070] Through alignment and fine-tuning, the large model can learn how to generate teaching videos that satisfy users based on course information, course photos, and teaching content.

[0071] The lecturer's teaching video model initialization steps: obtain the trained and tested artificial intelligence literary video model (including the trained and tested Sora, MuseV, Latte, etc.) as the generative model to be transferred learning.

[0072] Transfer learning steps for the lecturer teaching video model: Through transfer learning, the model can transfer the knowledge learned from a larger range of cultural video datasets to the lecturer teaching video model.

[0073] Obtain the lecturer's lecture video and extract the lecture content from the voice in the lecture video.

[0074] Steps for fine-tuning the instructions of the lecturer's teaching video model: fix the preset K layers in the generative model to be transferred, obtain the lecturer's information, the lecturer's photo, and the teaching content as input, obtain the lecturer's teaching video corresponding to the teaching content as the expected output, fine-tune the parameters of the unfixed layers in the generative model to be transferred, and obtain the lecturer's teaching video model after instruction fine-tuning;

[0075] Or, on the basis of the generative large model to be transferred learning, add preset M parameters, obtain the lecturer's information, the lecturer's photo, and the teaching content as input, obtain the lecturer's teaching video corresponding to the teaching content as the expected output, fine-tune the preset M parameters in the generative large model to be transferred learning, and obtain the lecturer's teaching video large model after instruction fine-tuning.

[0076] Through instruction fine-tuning, the large model can learn how to generate teaching videos based on the lecturer's information, lecturer's photo, and teaching content.

[0077] The steps of aligning and fine-tuning the lecturer's teaching video model are as follows: the lecturer's information, the lecturer's photo, and the teaching content are obtained as the input of the lecturer's teaching video model after instruction fine-tuning, the obtained output is displayed to students or evaluators, feedback from students or evaluators is obtained, and reinforcement learning is performed on the lecturer's teaching video model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned lecturer's teaching video model.

[0078] Through alignment and fine-tuning, the large model can learn how to generate a teaching video that satisfies the user based on the lecturer's information, the lecturer's photo, and the teaching content.

[0079] Robot teaching model initialization steps: Get the deep learning model as the robot teaching model.

[0080] The training and testing steps of the robot teaching model are as follows: control the robot to perform various movements; obtain the robot's action control instructions (including mouth control instructions, facial expression control instructions and limb movement control instructions) and their corresponding action videos; use the action video as input and the action control instructions corresponding to the action video as expected output to train and test the robot teaching model.

[0081] The robot obtains information steps: the robot obtains the voice location of the student who asked the question, drives the robot to turn to the student who asked the question, performs face recognition on the student, and obtains the student's characteristics, which include age, gender, or occupation and major.

[0082] The steps of generating teaching content are as follows: obtaining the course information referred to by the student who asked the question through the robot; if the student who asked the question did not refer to a course, obtaining the course information closest to the student who asked the question; using the course information, the question asked, or the student's characteristics, the course information, the question asked, or the lecturer's information, the student's characteristics, the course information and the question asked by the student who asked the question as the input of the course teaching big model (for example, the course teaching big model after deep alignment); the output obtained is used as the teaching content for the question asked, and the teaching content is taught by the robot to the student who asked the question.

[0083] Content video display step: obtain course information, course photos, and the teaching content corresponding to the question as input, and calculate the output of the course visual macro model (such as the course visual macro model after alignment and fine-tuning) as the content video of the question. The content video is displayed on the screen by the robot. The screen includes the screen on the robot body.

[0084] Through robot control, "the teaching content is taught by the robot to the students who asked the questions" and "the content video is displayed by the robot on the screen" are synchronized.

[0085] Teaching video display steps: obtain the lecturer selected by the student, the lecturer's information, the lecturer's photo, and the teaching content as the input of the lecturer's teaching video model (the lecturer's teaching video model, for example, the lecturer's teaching video model after alignment and fine-tuning), and the output is the lecture video of the lecturer corresponding to the teaching content.

[0086] The robot's teaching action control steps are as follows: the robot displays the facial image of the lecturer in the teaching video on the robot's head screen; the teaching video is input into the robot's teaching model, and the calculated output is used as the action control instruction, and the robot's action is controlled by the action control instruction, so that the robot's expression and action when teaching are the same as those of humans.

[0087] Through robot control, "the facial image of the lecturer in the lecture video is displayed by the robot on the robot's head screen" and "the robot's movements are controlled through action control instructions" are synchronized.

[0088] Through robot control, "the teaching content is taught by the robot to the students who ask questions" and "the content video is displayed by the robot on the screen" and "the facial image of the lecturer in the teaching video is displayed by the robot on the robot's head screen" and "the robot's movements are controlled through motion control instructions" are synchronized.

[0089] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. An artificial intelligence method, characterized in that: The method comprises: Teaching content generation step: taking the course information and the questions raised by the students as the input of the course teaching model, and taking the output as the teaching content for the questions raised, and teaching the answers to the students who raised the questions by the robot; Content video display steps: obtain course information, course photos, and the teaching content corresponding to the questions raised as input, and the output obtained by calculating the course visual model is used as the content video of the questions raised; the content video is displayed on the screen by the robot.

2. The artificial intelligence method according to claim 1, characterized in that: The method further comprises: Lecture video display step: obtain the lecturer selected by the student, the lecturer's information, the lecturer's photo, and the lecture content as the input of the lecturer's lecture video model, and the output is the lecture video of the lecturer corresponding to the lecture content; The robot teaching action control steps are as follows: the facial image of the lecturer in the teaching video is displayed by the robot on the robot's head screen; the teaching video is input into the robot teaching model, and the calculated output is used as the action control instruction to control the robot's action.

3. The artificial intelligence method according to claim 1, characterized in that: The method further comprises: Course teaching big model pre-training step: taking course materials (including course information, detailed course introduction, lecturer information, lecturer's remarks, works, and other related data) as pre-training data of the generative big model, pre-training the generative big model, and obtaining the pre-trained course teaching big model; Steps for fine-tuning the instructions of the course teaching model: Based on the pre-trained course teaching model, add M preset parameters, obtain the course information and questions raised as input, obtain the teaching content as the expected output, fine-tune the M preset parameters in the pre-trained course teaching model, and obtain the course teaching model after instruction fine-tuning; The steps of aligning and fine-tuning the course teaching model are as follows: obtain the course information and questions raised as the input of the course teaching model after instruction fine-tuning, show the output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

4. The artificial intelligence method according to claim 1, characterized in that: The method further comprises: Steps for fine-tuning the instructions of the course visual big model: based on the generative big model to be transferred, add M preset parameters, obtain the course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, fine-tune the preset M parameters in the generative big model to be transferred, and obtain the course visual big model after instruction fine-tuning; The steps of aligning and fine-tuning the course visual big model are as follows: obtain the course information, course photos, and teaching content as the input of the course visual big model after instruction fine-tuning, show the obtained output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course visual big model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course visual big model.

5. An artificial intelligence system, characterized in that: The system comprises: Teaching content generation module: takes the course information and the questions raised by the students as the input of the course teaching model, and the output is used as the teaching content for the questions raised. The robot teaches the answers to the students who raised the questions; Content video display module: obtains course information, course photos, and the teaching content corresponding to the questions raised as input, and the output obtained through calculation of the course visual model is used as the content video of the questions raised; the content video is displayed on the screen by the robot.

6. The artificial intelligence system according to claim 5, characterized in that: The system further comprises: Lecture video display module: obtain the lecturer selected by the student, the lecturer's information, the lecturer's photo, and the lecture content as the input of the lecturer's lecture video model, and the output is the lecture video of the lecturer corresponding to the lecture content; Robot teaching motion control module: The robot displays the facial image of the lecturer in the teaching video on the robot's head screen; the teaching video is input into the robot teaching model, and the calculated output is used as the motion control instruction to control the robot's motion.

7. The artificial intelligence system according to claim 5, characterized in that: The system further comprises: Course teaching big model pre-training module: taking course materials (including course information, detailed course introduction, lecturer information, lecturer's remarks, works, and other related data) as pre-training data of the generative big model, pre-training the generative big model, and obtaining the pre-trained course teaching big model; Course teaching model instruction fine-tuning module: Based on the pre-trained course teaching model, add M preset parameters, obtain course information and questions as input, obtain teaching content as expected output, fine-tune the preset M parameters in the pre-trained course teaching model, and obtain the course teaching model after instruction fine-tuning; Course teaching model alignment and fine-tuning module: obtain course information and questions as input of the course teaching model after instruction fine-tuning, display the output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course teaching model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course teaching model.

8. The artificial intelligence system according to claim 5, characterized in that: The system further comprises: Course visual big model instruction fine-tuning module: Based on the generative big model to be transferred, add M preset parameters, obtain course information, course photos, and teaching content as input, obtain the teaching video corresponding to the teaching content as the expected output, and fine-tune the preset M parameters in the generative big model to be transferred to obtain the course visual big model after instruction fine-tuning; Course visual big model alignment and fine-tuning module: obtain course information, course photos, and teaching content as input of the course visual big model after instruction fine-tuning, display the obtained output to students or evaluators, obtain feedback from students or evaluators, and perform reinforcement learning on the course visual big model after instruction fine-tuning based on the feedback to obtain the aligned and fine-tuned course visual big model.

9. An artificial intelligence robot, comprising a memory, a processor, and an artificial intelligence robot program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images