From ChatGPT to Sora: Paradigm Innovation in AIGC Education

The education field is undergoing unprecedented changes, with the emergence of rapidly iterating large models like Sora, GPT-4, and Claude3 revealing the potential and challenges of AIGC in modern education. AIGC refers to “Artificial Intelligence Generated Content,” which learns from a vast amount of data samples, understands the rules and structures of content generation, and automatically creates new and original content. If ChatGPT emphasizes natural language processing capabilities, Sora focuses more on enhancing teaching and learning efficiency and effectiveness through interactive teaching content or data analysis. From ChatGPT to Sora, the continuously iterating AIGC technology aims to assist education in forming a four-ability educational innovation from low ability to high ability, single ability to multiple abilities, multiple abilities to super abilities, and super abilities to extraordinary abilities, while also processing teaching outcomes, industrializing innovative thinking, and normalizing knowledge emergence. This study, based on the intrinsic value of AIGC, proposes the four-ability education aimed at AIGC, further exploring the picture of AIGC transforming the entire educational process, and proposes a paradigm innovation in education aimed at AIGC based on a review of the current status of AIGC education both domestically and internationally, aiming to provide theoretical and practical references for the development of a Chinese-style modern education system.

Currently, AIGC has wide applications in various fields such as art creation, game development, and education generation. This study explores the intrinsic value of AIGC because it emphasizes AI’s capabilities in creative tasks, such as the automated generation of text, images, and videos. This creative ability has a disruptive impact across the entire chain of educational subjects, content, methods, and technical platforms. The algorithmic foundation of AIGC’s creativity originates from the paper “Attention is All You Need” published by the Google team in 2017, which introduced the “attention mechanism” of the Transformer model for the first time^[1], significantly enhancing the model’s ability to process and understand language.

The intrinsic value of AIGC is reflected in aspects such as non-linearity, multi-scale, self-organization, self-adaptation, irreducibility, signal transmission, feedback loops, and historical dependence. AIGC can handle complex problems, where its outputs and inputs are not a simple linear relationship but can create unexpected solutions or content. Through deep learning and understanding of a large number of training texts, it achieves rich knowledge embedding, accurately introducing facts, concepts, theories, and viewpoints when generating text. Its contextual adaptability ensures that the generated content is highly consistent with the dialogue or text context, while its diversity allows it to cover various topics, styles, and emotions, enriching the diversity of communication. AIGC’s innovative thinking and ability to handle complex dialogues enable effective understanding and response to various user interactions. Through semantic deepening, AIGC can also handle and generate deep semantic texts containing complex discussions and arguments. At the same time, inclination regulation and self-monitoring mechanisms ensure that the generated content meets predetermined inclinations while avoiding inappropriate or harmful information, reflecting AIGC’s comprehensive value in intelligence, safety, and ethics. In summary, the intrinsic value of AIGC lies in its strong adaptability, creativity, and self-evolution capabilities.

Therefore, in the field of education, AIGC integrates four intelligences, driving educational momentum and demonstrating enormous value: 1) Generative intelligence can provide customized learning materials and innovative forms of knowledge presentation for students, stimulating learning motivation; 2) Experiential intelligence can provide real-time feedback and personalized learning experiences by understanding students’ learning environments and needs; 3) Understanding intelligence enables AI to deeply understand educational content and student feedback, promoting more effective teaching and learning strategies; 4) Consciousness intelligence, although still in the exploratory stage, suggests that future AI can participate more deeply in the educational process, providing more humanized support and interaction. By integrating different intelligences, AIGC not only achieves innovation in knowledge transfer in education, but its “emergent” creativity also brings tremendous changes to educational participants, content, teaching methods, and technical platforms.

1 The Basic Logic, Value Goals, and Current Challenges of Education

The basic logic of education is to prepare individuals for effective participation and contribution in social and professional life by imparting knowledge, cultivating skills, shaping values, and promoting holistic personal development^[2]. The value goal of education lies in promoting learners’ cognitive development, emotional maturity, and social adaptation, stimulating innovative thinking and critical thinking, as well as cultivating lifelong learning abilities and good moral qualities, ultimately facilitating personal self-realization and social progress and prosperity.

Currently, the education field faces multifaceted challenges, including teaching content and quality, technology integration, regional educational resource allocation, educational equity, and internationalization. The rapid development of large models has created a new contradiction between technological-led innovation and lagging practical mechanisms in education and teaching^[3]. How to effectively integrate advanced technologies like AIGC into the education system to facilitate learning has become a significant challenge^[4]. At the same time, technical support for teachers’ professional development also urgently needs to be synchronized, and research institutions must provide necessary training and resources to ensure that teachers can effectively utilize these technologies. Furthermore, in the context of global knowledge dissemination, the internationalization of education is also one of the challenges facing education in our country, requiring the education system to be more international in content, technology, and platforms to cultivate students’ global perspective and competitiveness.

2 Proposal of the Four-Ability Education Theory of AIGC

Combining the current development of AIGC technology, this study proposes the AIGC Quadruple Ability Education theory, which states that AIGC helps learners achieve comprehensive development from low ability to high ability, from single ability to multiple abilities, from multiple abilities to super abilities, and from super abilities to extraordinary abilities through assisting learning and personalized education, facilitating interdisciplinary learning and comprehensive skill cultivation, enhancing deep learning and higher-order thinking ability improvement, and expanding cognitive boundaries and innovative thinking modes. The theoretical model architecture of AIGC’s four abilities is shown in Figure 1. Starting from defining teaching needs, AIGC assists learning and personalized education, promotes interdisciplinary learning and comprehensive skill cultivation, enhances deep learning and higher-order thinking ability improvement, and expands cognitive boundaries and innovative thinking modes, combined with teaching research technology presentation for teaching research practice, ultimately completing the educational paradigm innovation.

Figure 1 The Theoretical Model Architecture of AIGC’s Four Abilities

1) In the first stage, from low ability to high ability, AI assists learning and personalized education. Through personalized education, AI technology is used to help learners rapidly transition from a low ability state to a high ability state. By analyzing learners’ learning habits, knowledge mastery, and feedback, AI can enhance the adaptability of teaching materials, customizing learning content and teaching strategies suitable for each student. For example, for beginners, AI can recommend foundational and detailed teaching materials, establish an interactive teaching process, and gradually enhance students’ understanding and application abilities through practice and feedback on digital content. An English learner can receive personalized vocabulary, grammar, and listening/speaking training based on their learning pace and interests through an AI learning platform, significantly improving their English proficiency in a short time, that is, AI customizes teaching content and methods according to each learner’s needs and strengths. At this stage, large models as tools not only replace traditional teaching tools but also change the teacher-centered knowledge transfer approach. Through effective personalized interaction and feedback mechanisms, the learning experience becomes active and deep^[5], achieving the transition from low ability to high ability and cultivating learning interest.

2) In the second stage, from single ability to multiple abilities, AI promotes interdisciplinary learning and comprehensive skill cultivation. AI technology is used to help learners develop from a state of possessing a single skill to having multiple mature skills. AI can integrate knowledge from various fields, providing learners with a wider range of learning resources to help them master multiple skills and develop comprehensive skills. For example, a student proficient in mathematics can begin to learn how to apply mathematical knowledge in computer programming and data analysis through AI guidance, gradually forming interdisciplinary problem-solving abilities; an engineer can learn knowledge about marketing and human resource management through the AI platform, thereby showcasing a more comprehensive capability in the workplace. At this stage, students utilize large models to promote collaborative learning boundaries, achieving the integration of interdisciplinary heterogeneous knowledge, expanding the skill knowledge tree, and broadly exploring the “human encyclopedia,” resulting in cross-cultural learning and interdisciplinary collaborative effects. Teachers can reasonably adopt methods for constructing multimodal teaching resources empowered by GPT in teaching practice, further promoting the digitalization of multimodal teaching resources, endowing students with higher academic and research capabilities through more timely, richer, and more exciting teaching content.

3) In the third stage, from multiple abilities to super abilities, AI enhances deep learning and higher-order thinking ability improvement. Through deep learning and higher-order thinking ability cultivation, AI technology is used to help learners develop from a state of possessing multiple mature skills to a super high ability state. AI can guide learners to deeply study specific fields by providing high-quality, specialized knowledge systems, improving their innovative ability, critical thinking ability, and problem-solving ability. Through the application of large model technology, AI can also simulate complex real-world problems for students to solve, or provide experimental conditions through advanced simulation laboratories for scientific research. For example, a medical student can conduct advanced biomedical research through a virtual laboratory, enhancing their professional abilities and innovative thinking in the medical field. At this level, AI begins to change the essence of teaching, focusing more on students’ deep active participation in teaching practice, emphasizing students’ deep learning and thinking abilities as well as interdisciplinary collaborative thinking. Currently, the ChatGPT large language model has launched a multi-user document collaboration mode, and video creation models like Midjourney can also serve multiple people in various scenarios by creating new asset groups, allowing teachers to cultivate students’ super high innovative thinking and technical abilities using these tools.

4) In the final stage, from super abilities to extraordinary abilities, AI expands cognitive boundaries and innovative thinking modes. AI technology is used to help learners develop from a super high ability state to an extraordinary ability state. The former refers to the acquisition, satisfaction, and enhancement of abilities based on the existing human knowledge system, while the extraordinary ability state refers to a technological leap beyond existing human capabilities. New content such as anti-physical aesthetics, intelligent illusion aesthetics, and infinitely approaching aesthetics generated by AI represents a significant subversion of the existing physical world laws and the construction of imagination and aesthetics in the educational process. For example, students can learn biological principles using AI tools and apply them to sustainable design, creating entirely new design works. Additionally, AI educational large models can be applied in teaching, creating previously unimaginable new scenarios and tasks, forming learning environments that seamlessly integrate physical and network spaces^[6]. Students can “visit” Cleopatra’s palace or enter the interior of a whale’s skeleton to examine its structure through AI-constructed simulation spaces in a classroom of limited size. The use of technology far exceeds the basic routines of displaying lecture notes on projection screens and entering assessment scores in digital databases, expanding the cognitive boundaries of educators and students. These teaching experiences continuously improve teaching effectiveness through real-time feedback and assessment. On this basis, AI assists learners in exploring their potential cognitive abilities, developing unprecedented “extraordinary abilities” to meet the challenges and changes of future society.

For four-ability education, in the multi-ability domain, AI can quickly enable individuals to master new skills, such as composing, painting, and video production. However, the key lies in how to transition from multiple abilities to super abilities or even extraordinary abilities, promoting differentiated development of the younger generation’s capabilities. Therefore, education should take AIGC-generated content as a new starting point, facilitating the integration of heaven, man, and artificial intelligence from unity of heaven and man to unity of heaven, man, and intelligence. Through process-oriented and endogenous innovation, cognitive outsourcing can be transformed into opportunities for capability enhancement.

1 Learners: Full-Age Coverage of “Self-Production” and “Self-Adaptation” Learning

AIGC’s multimodal content “self-production” capability provides rich and diverse learning resources and environments for students at different developmental stages, supporting their growth in knowledge breadth and depth. In elementary education, AIGC can stimulate students’ interests through various modalities (text, images, audio, etc.), helping them identify points of interest and build interdisciplinary knowledge systems. In secondary education, AIGC promotes students’ cognitive development and thinking ability improvement by deepening multimodal analysis of knowledge and providing materials for critical thinking training. In higher education, AIGC supports students in conducting in-depth research in their professional fields, leveraging its powerful data processing and analysis capabilities to accelerate research progress and providing practical learning opportunities through simulated experimental environments.

AIGC tests individuals’ knowledge blind spots through the integration of powerful content generation, rich knowledge resources, and intelligent interactive tools, significantly enhancing students’ “self-adaptive learning” abilities. The so-called “self-adaptive learning” refers to a personalized, intelligent learning model that covers individuals across all age groups. While enhancing self-adaptive learning abilities, AIGC also provides fertile ground and motivation for the development of students’ creative thinking and problem-solving skills through its efficient content generation capabilities and deep understanding of knowledge. AIGC, leveraging its excellent content generation mechanisms, vast knowledge base, and question-answer systems, exercises students’ imagination and creativity while broadening their cognitive boundaries, further promoting their ability to think independently and solve problems. At the same time, the design of personalized learning paths and real-time interactive feedback mechanisms stimulates students’ desire for exploration, fostering their self-directed learning. It is evident that AIGC enhances students’ imagination, creativity, aesthetic ability, and judgment through a “superhuman” learning approach^[7].

The “intelligent teaching integration” of AIGC not only optimizes the teaching process but also lays the foundation for students’ personalized and lifelong learning, achieving optimal resource allocation in education by precisely adapting to each student’s learning needs and preferences^[8]. This transformation in educational models fundamentally promotes learners’ initiative and creativity.

2 Educators: From Teaching to Mutual Growth

For traditional educators, the impact of AIGC is one of “transformation” and “reconstruction.” It not only changes the way teaching content is created and presented but also enhances teaching efficiency and the potential for personalized teaching. Teachers can utilize AIGC to generate personalized teaching materials and make data-driven teaching decisions, allowing them to focus more on strategy formulation and student interaction.

At the same time, the application of AIGC in education also signifies a transformation in teaching roles. The role of teachers is shifting from traditional knowledge transmitters to guides and facilitators of learning. This transformation means that teachers are no longer the sole source of knowledge in the classroom but become partners assisting students in navigating the knowledge exploration process. Teachers will focus more on stimulating students’ interests, guiding their thinking and exploration, and promoting interaction and cooperation among students^[9]. This shift helps break down the traditional top-down teaching model, achieving equality and interaction in the teaching environment, thus promoting mutual growth in teaching.

Moreover, once AIGC is combined with humanoid robots in teaching, some repetitive and low-efficiency teaching tasks will be replaced^[10]. For example, AI can automatically grade assignments, provide students with learning analysis reports, or undertake some basic knowledge teaching, allowing teachers to have more time and energy to focus on teaching design and student interaction. This not only enhances the innovation of teaching activities but also improves teaching efficiency and quality. However, the professional development of teachers also faces new challenges. In the face of technological changes, teachers need to continuously adapt and learn, enhancing their information technology application abilities, which includes understanding and mastering the use of AI educational tools and how to effectively integrate these tools into teaching.

3 Innovative Applications in Teaching Concepts, Resources, and Technical Platforms

In terms of teaching concepts, AIGC introduces the concepts of personalized and adaptive learning, making education more student-centered. The traditional “unified” teaching model is being replaced by more flexible and personalized teaching strategies, with the concept of teaching according to students’ individual needs being given new significance by AIGC’s rich teaching content. In the Middle East, research based on the SAMR theory has summarized the innovative aspects of AIGC in educational concepts into four levels: Substitution, Augmentation, Modification, and Redefinition. AIGC provides a layered approach to integrating technology into daily teaching activities^[11]. This transformation encourages students to learn proactively, promoting their critical thinking and creative problem-solving abilities while providing teachers more space to focus on students’ individual differences and teaching quality.

In terms of resources, the introduction of AIGC technologies like Sora indicates that AI-generated teaching resources and content will become the new norm, thereby reconstructing traditional textbook preparation methods. For example, AI can instantly generate teaching explanation images for the four seasons in the Northern and Southern Hemispheres without the cumbersome manual collection. These teaching resources not only cover traditional texts and images but also include interactive simulations, virtual laboratories, and augmented reality/virtual reality multimedia elements. Research shows that AIGC creates entirely new teaching experiences, providing students with immersive learning experiences that fundamentally alter traditional teaching methods^[12]. Additionally, the interactive features of AI can adaptively identify and test learners’ knowledge blind spots, accelerating cognitive enhancement.

In terms of technical platforms, AIGC drives the evolution of intelligent platforms, making them more intelligent and interactive. These platforms not only support the creation and presentation of multimodal content but also can adjust and optimize in real-time based on students’ interactions and learning progress. Large models like Sora and ChatGPT, although still in their infancy, have begun to change the ways educational resources are created and accessed. Kasneci et al.^[13] explored the opportunities and challenges of large language models in educational applications, emphasizing their potential in creating educational content. The integration and application of these platforms provide possibilities for seamless online and offline learning experiences, cross-regional resource sharing, and collaborative learning, breaking the time and space constraints of traditional educational models.

In summary, for learners, AIGC technology achieves personalized and multimodal teaching in full-age education through its self-production and self-adaptive learning models, effectively promoting cognitive growth and capability enhancement at different developmental stages. For educators, AIGC triggers transformations in teaching roles and methods, thereby improving teaching efficiency and fostering a more interactive and personalized teaching environment. Meanwhile, the innovative applications of AIGC in teaching concepts, resources, and technical platforms provide richer and more dynamic learning resources for education, promoting the evolution of educational models towards greater flexibility, interactivity, and personalization. In the short term, the application of large AIGC models in education and training has realized “what is thought is what is seen,” meaning that what people think can be visualized through artificial intelligence, facilitating instant capability training. In the long run, the human-AI symbiotic AI environment will automatically identify and supplement the deficiencies in human knowledge structures. The ability of the human brain to receive new information and generate innovative ideas reflects the emergence of human wisdom and inspiration, which is crucial for the learning and creative process. This strategy not only reshapes the teaching process but also promotes continuous innovation in teaching methods and content, ensuring that educational practices can keep pace with the times and better meet the needs of modern education.

1 Research Paradigm Innovation of AIGC and Educational Large Models

To effectively apply AIGC’s four-ability education in practice, it is necessary to explore the research laws and current status of AIGC technology in depth, and based on this, systematically innovate and construct the educational large model. The technical foundation of AIGC is the large model, which refers to artificial intelligence systems with training parameters ranging from billions to hundreds of billions or more. These systems result from the collaborative development of deep learning technology, GPU hardware acceleration, and large-scale datasets. The transition from general large models to specialized large models in the education field is an inevitable trend in the deepening development of artificial intelligence large model technology^[14]. AI educational large models are artificial intelligence systems designed specifically for educational environments, possessing massive training parameters. The introduction of deep learning, especially architectures such as Generative Adversarial Networks (GAN) and Variational Autoencoders (VAE), provides strong generative capabilities for creating AI educational large models^[15]. Currently, there is no systematic educational large model platform globally, and educational large models need to be constructed based on general large models. Table 1 lists representative general large models and their applications in China and abroad. For example, ChatGPT and Sora represent the technologies of text generation for text/images and text generation for videos, respectively, with the latter providing richer information due to its intuitiveness and the inherent rhythm and multidimensional value (visual, social, emotional, temporal-spatial) of videos. Sora’s video generation capability optimizes the training of various human skills, enhancing abstract thinking and imagination. In the future, video content is expected to dominate daily life.

Table 1 Examples of Representative General Large Models

Starting from general large models and combining the educational goals and processes discussed earlier, it can be seen that single-modal content generation cannot meet educational needs. Educational large models are no longer limited to single content generation but are moving towards multimodal and interactive creation. Model-driven generative artificial intelligence is changing the attributes of its technical tools^[16]. Currently, countries around the world are placing high importance on the application of educational large models in education. On the practical level, students participating in Tsinghua University’s course on “Metaverse Development and Challenges” completed an exploratory experiment using AI for writing assignments, questioning the boundaries of human-machine interaction^[17]. Through this experiment, students not only mastered the application of AI technology in academic writing but also deeply experienced the potential and limitations of AI-assisted learning, promoting their profound reflection on the interaction between future technology and society, as well as ethics. On the policy level, the U.S. Department of Education’s technology office released a policy report on “Artificial Intelligence and Future Teaching,” emphasizing the role of humans in the loop and designing artificial intelligence according to modern learning principles^[18]. The United Nations Educational, Scientific and Cultural Organization (UNESCO) aims to utilize artificial intelligence technology to achieve the 2030 Education Agenda, developing publications to help education policymakers prepare for the use of artificial intelligence^[19]. The UK Department of Education has released guidance on the use of generative artificial intelligence in education, including large language models (such as ChatGPT and Google Bard), emphasizing the application potential of generative AI in education while pointing out the risks and challenges that need attention when using these technologies^[20].

This study uses CiteSpace to visually present the knowledge domain of all scientific research literature related to large models, education, and other aspects from the WoS database from 2018 to 2024, to understand the global research progress of AIGC education. Using the Web of Science (WoS) core database as the literature source, advanced searches were conducted using keywords such as “education,” “large language model,” and “content generation,” with a time frame from January 1, 2018, to February 28, 2024. Through a combination of precise matching and manual coding methods, 381 samples were annotated. Based on knowledge nodes (as shown in Figure 2), theme clustering (as shown in Figure 3), and keyword statistics (as shown in Table 2), this study finds that the combination of large models with artificial intelligence and education, educational innovation, and educational technology is closely related.

Table 2 Keyword Statistics for Education and Large Models from 2018 to 2024 (Top 20)

The data indicates that the combination of AIGC and education is an inevitable trend in the future development of education. The paradigm innovation and framework construction of educational large models based on language large models and video large models are urgent. Currently, the high frequency of the term artificial intelligence indicates that it is the core field of research on educational large models. The frequent appearance of the keyword large language model and its related terms (such as large language model and language model) emphasizes the importance of large models in teaching research. Keywords such as generative AI, education technology, and students indicate that AIGC has the conditions to create new learning resources (such as simulation experiments, teaching cases, etc.) in the field of education, with a close connection between education and educational large models. For example, research shows that the application of AIGC in education can enhance the learning process^[21], empowering teaching practices through three-dimensional simulations of human organs, highlighting the collaborative potential of generative AI and educational technology. Moreover, keywords such as medical education, higher education, and diagnosis reflect the application of educational large models in specific educational fields: theoretically, learning through simulated cases has improved the interactivity and practicality of medical education; practically, the application of AIGC in medical imaging has promoted innovation in clinical applications and played an important role in helping students understand and analyze medical images^[22].

In summary, the global picture shows that the application of AIGC technology in the field of education is becoming more widespread and in-depth. Research from various countries has clarified the importance of ensuring the credibility of generated content and the successful integration of generative AI tools in educational environments, with AIGC integrating to promote the intelligent transformation of the traditional education industry. The theoretical construction of educational large models aimed at AIGC not only helps clarify the role and goals of AI in education but also ensures the directional and effective application of technology, avoiding blindly pursuing technological innovation while neglecting the practical needs of educational practice. Additionally, the construction of educational large models aimed at AIGC can better integrate into educational links such as teaching design, curriculum content, assessment methods, and learning analysis, thereby promoting personalized learning, enhancing teaching interactivity, and improving educational quality.

Figure 2 Research Map of Education and Large Models

Figure 3 Theme Clustering of Education and Large Models Research

2 Construction of AIGC Educational Large Models

The emergence of large model content generation platforms reveals the potential and challenges of AI content generation in modern education. Generally speaking, educational large models rely on vast and diverse datasets for training, which cover extensive information within the model application field. To construct AIGC educational large models, it is necessary to preprocess and conduct feature engineering training on educational data, including learning behavior data, educational resources, assessment outcomes, and interaction records, to enhance data quality and optimize the learning efficiency and performance of the model. The construction process of AIGC educational large models also includes model architecture, optimization algorithms, evaluation metrics, and training strategies, which ultimately need to be integrated into practical educational applications. All these elements together constitute the core of AIGC educational large models. This study proposes the AIGC educational large model paradigm shown in Figure 4, which includes the underlying architecture layer, model training layer, data processing layer, application layer, user interface layer, data security and privacy layer, and feedback and iterative optimization layer, covering all aspects from underlying technology to application, aiming to provide efficient, precise, and personalized services for education and other fields.

Figure 4 AIGC Educational Large Model Paradigm

In practice, the overall paradigm of AIGC educational large models is driven by innovation through “four plus five,” where the model backend includes cloud computing, foundational large models, standardized frameworks for educational content, and alignment of values, while the frontend interface encompasses student learning assistants, teacher teaching assistants, a repository of teaching experts, a wisdom cluster of top teachers, and user entities. This paradigm demonstrates how AIGC educational large models can flexibly adapt to the ever-changing educational needs and technological advancements, starting from foundational data and technical architecture, through complex data processing and model training, ultimately providing personalized and highly interactive educational services at the application layer.

1 Innovations of the Era of “Education + AIGC”

The future AIGC educational paradigm innovation is both a technological revolution and a profound transformation of educational concepts and methods. In this paradigm, the application of artificial intelligence tools such as ChatGPT and Sora is likely to transcend the boundaries of traditional teaching, leading education towards a more personalized, interactive, and intelligent direction. The educational applications of AIGC will achieve a shift from information transmission to knowledge construction, from memory understanding to innovative thinking, emphasizing the subjectivity and creativity of learners. The emergence of large models essentially signifies the industrialization of innovation and thinking; traditional industrialization refers to the industrialization of visible tangible products, while the emergence of AIGC large models completes the industrialization of human innovative thinking. At the same time, AIGC’s four-ability education can assist researchers in analyzing large and complex datasets, revealing potential scientific laws, accelerating and leaping personal capabilities, leveraging 1% of human effort to unlock 99% of AI content production power, and ultimately transitioning from the unity of heaven and man to the unity of heaven, man, and intelligence, thus accelerating the processes of scientific discovery and knowledge innovation. The intelligent teaching integration of AIGC can be summarized as a “pan-L1-L5 theory” level, divided into L1 Assisted Automation, L2 Partial Automation, L3 Conditional Automation, L4 High Automation, and L5 Full Automation. In the future, the powerful content generation, data processing, and learning capabilities of AIGC technology will provide unprecedented support for research and educational practice, emphasizing the comprehensive enhancement of technology, cognition, learning, and innovative abilities. Based on artificial general intelligence (AGI), highly complex machine learning models, deep logical reasoning, and autonomous learning and adaptation capabilities, AIGC will promote the gradual progression of the entire education process in complexity, independence, and innovative capabilities.

On May 29, 2023, General Secretary Xi Jinping emphasized during the fifth collective study session of the Political Bureau of the CPC Central Committee that “building a strong educational nation is a strategic precursor to comprehensively building a modern socialist country, an important support for achieving high-level technological self-reliance, an effective way to promote common prosperity for all people, and a foundational project for advancing the great rejuvenation of the Chinese nation with Chinese-style modernization.”^[23] General Secretary Xi’s important speech emphasizes the core position of education in the national modernization process, particularly pointing out the need to accelerate the construction of world-class universities and advantageous disciplines with Chinese characteristics and to promote scientific research innovation. In this context, promoting scientific research and educational innovation and constructing the AIGC education framework as an important outcome of emerging technologies is of great significance. It not only plays an important role in accelerating research processes, promoting interdisciplinary integration, advancing educational modernization, and building a strong educational nation but also provides strong technological support for educational innovation.

2 Future Risks and Recommendations

As a probabilistic model, the inherent AI hallucination phenomenon of large models leads to the possibility of errors, which contradicts the high requirements for accuracy in the education field. Therefore, “education + AIGC” should focus on improving the reliability of models. In specific teaching practices, firstly, there exists an essential difference between data model technology and the physical rules of the real world^[24], which may lead to erroneous knowledge transfer and subsequently affect students’ learning outcomes. Secondly, over-reliance on technology may cause students to lack understanding and mastery of foundational knowledge; the imitation of styles by AIGC cannot replace human artistic practice and the continuous accumulation and evolution of styles in practice^[25]. Therefore, education in foundational knowledge and skills remains indispensable. Furthermore, regional inequalities in technology and education are becoming increasingly prominent^[26]. Additionally, issues such as data privacy and security are also worthy of attention. In the face of these risks and challenges, it is necessary to train teachers and students in data science awareness during the use of large models to ensure the scientific application of AIGC technology. At the same time, relevant parties must continuously monitor and evaluate the effects of technology applications to avoid excessive reliance on AIGC technology. Moreover, an open-ended AI capability test for students can be included in the educational assessment system. As AI technology is increasingly applied in the field of education and the AI hallucination problem is effectively mitigated, cultivating students’ AI creative abilities will become increasingly important. This paradigm innovation not only promotes the innovation of educational content and methods but also provides students with opportunities for skill cultivation that keep pace with the times.

The integration of AIGC with educational subjects, content, technology, and platforms is an inevitable direction for future educational development. On one hand, it relies on the support of large models, transforming all materials and teaching environments into fully multimodal and AI-driven formats, with the explosion of information and intelligent content generation stimulating students’ imagination and creativity. On the other hand, AIGC achieves the “four-ability education” by assisting learning and personalized education, promoting interdisciplinary learning and comprehensive skill cultivation, enhancing deep learning and higher-order thinking ability improvement, and expanding students’ cognitive boundaries and innovative thinking modes. AIGC also drives teachers’ roles to shift from traditional “instructors” to “guides of learning.” AI will serve as a dynamically developing intermediary role, producing unprecedented profound impacts on the education field, with its role evolving with technological advancements, from teaching assistants to professors, even surpassing the existing academic boundaries of knowledge. As AI capabilities continue to grow, its position in the educational process will correspondingly rise, potentially becoming a leading learning model. This technological advancement fosters lifelong learning as a universal phenomenon, providing strong support for interactions between individuals and AI.

This study, based on the intrinsic value of AIGC technology, proposes the four-ability education theory aimed at AIGC, analyzing the impact of AIGC on the modern educational process, including the disruptive innovation of content and educational subjects, and the redefinition of technology and platforms. Based on a review of global educational and large model research frontiers, it further constructs the AIGC educational large model from foundational educational data architecture, complex learning model training, interactive teaching layers to innovative educational services. This study provides theoretical basis and practical guidance for utilizing advanced technologies to promote educational innovation, helping to cultivate innovative talents that meet the future societal needs of our country.