The Potential, Risks, and Regulatory Pathways of Generative AI in Education

In early 2023, the American AI research company OpenAI launched a generative artificial intelligence (AI) application called “Chat GPT” (Chat Generative Pre-trained Transformer), which possesses strong generative capabilities in language and text. Within just two months of its launch, the user count surpassed 100 million, while the domestic WeChat user base took 433 days to reach the same milestone. Although there are differences in logic and functionality between the two applications, the explosive popularity of Chat GPT highlights the significant influence of generative AI. Moreover, numerous domestic internet companies have publicly announced their acceleration in the research and development of generative AI, indicating that applications similar to Chat GPT are an important trend in the future development of society. With the rapid advancement of generative AI, society is accelerating its intelligent iteration, and countless industries, including education, will be affected to varying degrees.

Chat GPT exhibits exceptional performance in text generation, incorporating core technologies such as encoder-decoder-based fine-tuning, chain-of-thought techniques, and reinforcement learning based on human feedback, which will have a profound impact on the education sector. It may even trigger deep changes in education, from conceptual foundations to practical implementations. The long-standing limitations in personalized teaching customization may now have the potential for widespread development. Humanity is currently in an era of technological prosperity, and as we cannot isolate the impact of generative AI on education, we should adopt a proactive attitude towards embracing it. This article will explore the potential and risks of generative AI, represented by Chat GPT, in education at large and in music education specifically, and propose targeted regulatory pathways.

From a logical standpoint, Chat GPT employs a “big data + strong learning” model, rapidly scanning and receiving text through its vast information storage and learning capabilities, converting data and outputting it in human-like language. Currently, generative AI applications like Chat GPT demonstrate potential in three main areas within the education sector.

(1) From Scarcity to Abundance: Human-Machine Interaction Enhances Teachers’ and Students’ Understanding of Courses and Strategies

Chat GPT is supported by powerful computing capabilities and data, enabling it to perform bulk personalized analysis, producing different impacts on teachers and students.

From the perspective of teachers, human-machine interaction can enhance their understanding of courses. Generative AI presents a robust personalized analytical ecosystem that significantly improves teachers’ understanding of courses at different levels. In terms of detailed course theory, it stimulates teachers’ engagement in comprehending course levels. The Goodlad curriculum theory outlines five levels of curriculum, with the ‘understood curriculum’ occupying a crucial position. In simple terms, the “understood curriculum” can be understood as the implementation of teachers’ understanding of the curriculum in teaching design. In real teaching scenarios, the elements and creative points of teaching design are often difficult to exhaust, while the creative insights that teachers can capture are extremely limited. With the powerful computing power and openness of generative AI, the flexible mobilization of data and rapid generation of resources allow teachers to more easily acquire resources during educational activities. The model library of generative AI employs data storage and management technologies, enabling it to provide teachers with richer materials during the teaching design process, facilitating the acquisition of broader information, thereby increasing the likelihood of creating excellent teaching designs. Furthermore, this approach can enhance teachers’ cognitive engagement in the “understood curriculum” level, allowing them to identify deficiencies in their teaching designs and break free from misconceptions regarding curriculum understanding and teaching design, thereby engaging more actively in comprehending the curriculum.

In terms of specific content in music education, generative AI can continuously change text formats and content according to user needs, creatively meeting various requirements. For instance, when teachers design lessons, they often need to create melodies based on themes, match actions, or arrange instrumentation. The effectiveness of teachers’ creations primarily relies on their understanding of the curriculum. Generative AI possesses advanced natural language processing technology, enabling it to generate not only text but also corresponding sheet music, simple scores, rhythms, etc., based on requirements. It can even generate multi-part and multi-instrument combinations. It can provide teachers with more musical ideas for formal courses, promoting the development of the “implemented curriculum” and enhancing teachers’ understanding of courses through the integration of generative AI and music.

From the perspective of students, human-machine interaction can enhance their strategic understanding. The supercomputing power of AI can support personalized analysis of each student’s learning behavior in education, providing tailored learning suggestions and crafting learning plans, helping students choose learning strategies that better fit their actual situations. McCarroll and others have identified “resource management strategies” from a macro learning strategy perspective, where resources include time, spatial environments, and learning tools. By sending specific learning situations of students to Chat GPT, the algorithms can analyze and reason to provide effective resource management decisions.

For example, in music education, if a student sends a music learning evaluation form to Chat GPT, it can provide corresponding resource management strategy suggestions based on the student’s various abilities, such as “spend more time learning rhythm,” “enhance the use of learning tools (arrangement tools, notation tools),” and “immerse yourself in music spaces like concerts and performances to experience music.” In addition, generative AI can assist students in using “organizational strategies.” Organizational strategies emphasize integration, linking new knowledge with old knowledge to form new knowledge structures. By leveraging Chat GPT’s advanced natural language processing technology, students can send fragmented knowledge elements, and generative AI will quickly provide reasonable knowledge frameworks, helping students establish an orderly knowledge system from complex materials.

In summary, generative AI applications like Chat GPT can achieve precise analysis of teaching facts based on different users’ preferences, meeting personalized needs in teaching and providing convenience for teachers and students. For instance, it helps teachers understand the requirements of various course levels more deeply, promotes diverse teaching innovations, and assists students in learning according to their actual situations, thereby better mastering knowledge and improving learning efficiency. Thus, both teachers and students can enhance their engagement in teaching and learning with the help of generative AI.

(2) From Inefficiency to Efficiency: Information Generation Provides “Competency Support” for Teachers

In actual teaching, good teacher competency provides a guarantee for smoother implementation of teaching processes, referring to the value dimension of teachers’ teaching. In the intelligent era, data literacy has become one of the essential competencies for teachers. Data literacy primarily refers to the ability to reasonably utilize data, tools, and representations, ensuring that teachers can use intelligent courseware and new technology devices to assist in teaching. Specifically, teachers inevitably use the “Office Suite”—Word, Excel, and PPT—during the teaching process. These tools possess a certain level of professionalism; although the barriers have gradually lowered, teachers still need to invest effort to learn and spend time operating them after mastering them.

Currently, generative AI has integrated with professional tools like Word, Excel, and PPT, conveniently providing teachers with the teaching highlights they wish to showcase, including intelligent PPT generation, automatic addition of extracurricular materials, rapid literature classification, and intelligent formatting adjustments. During lesson preparation, AI acts as a “third hand” for teachers, providing stronger support for their data literacy, helping them optimize teaching, improve efficiency, and allowing them to focus more on creative labor, thus promoting the development of educational and research work.

In the field of music education, this is specifically manifested in score creation, audio production, and frequency spectrum presentation, enabling music teachers to devote more energy to creative practice, discover new concepts and methods in music education, and create more new musical works, thus driving the development of music education.

(3) From Narrow to Broad: Connectivity Across the Internet Expands Teachers’ and Students’ Cognitive Horizons and Resource Channels

Generative AI is backed by powerful interconnectivity, capable of accessing the entire internet, suitable for the retrieval and presentation of various educational resources, thus expanding the cognitive horizons and resource channels for students and teachers.

Firstly, connectivity across the internet can open up cognitive spaces for teachers and students. For example, Chat GPT possesses the capability for comprehensive information retrieval and presentation, significantly broadening the cognitive horizons of students and teachers. In real-world scenarios, it is almost impossible for any teacher to answer every question, but now, relying on the powerful algorithms and interconnectivity of generative AI, as well as the advanced technology of data information mobilization, existing educational information becomes readily accessible resources, allowing most questions from students and teachers to be answered through interaction with AI. Through dialogue with generative AI, the knowledge elements that students and teachers learn exhibit broad spatial and temporal characteristics, providing more opportunities to understand different knowledge elements and expand their cognitive horizons.

Secondly, the accurate results provided by generative AI can effectively avoid information overload. During use, generative AI can continuously identify the connotations of texts, maintaining constant learning to provide teachers and students with accurate and rich resource channels. Traditional search tools often present results mixed with a large amount of irrelevant information, such as advertisements, out-of-context related information, and distant recommendations, making it difficult for users to find effective information and key details. In contrast, generative AI can accurately identify user needs and provide the required information and educational resources, and the information provided by generative AI is free from advertisements and does not misinterpret user queries or recommend irrelevant information, significantly broadening the channels and effectiveness for teachers and students to access educational resources.

For example, in music education, many classical music audio recordings are outdated and difficult to access, or there may be issues with finding audio or video recordings. Chat GPT can provide the specific materials needed by music teachers based on different requirements, such as different versions, audio quality, and video quality, even down to the specific ensemble or conductor. Additionally, due to commercial development, results obtained from traditional search engines often contain numerous music advertisements, leading to a large amount of unnecessary information and confusion between similar-sounding works with entirely different content. The results provided by generative AI almost eliminate all redundant information, allowing for more accurate targeting of the desired results.

Despite the advantages of generative AI applications like Chat GPT, it cannot conceal the fact that it is a “double-edged sword.” Scholars have proposed ethical risks associated with AI from various dimensions, including subjectivity, relational aspects, algorithmic considerations, and resource dimensions. Specifically, these concerns highlight issues such as insufficient academic integrity, unstable information transmission, and unreliable ethical awareness. Overall, Chat GPT serves as both “Aladdin’s lamp” and “Pandora’s box,” with the latter identity primarily reflected in three aspects: subject dependence, information “black box,” and teacher-student relationships.

(1) Increasing Subject Dependence: Students’ Deep Learning is Replaced

The widespread application of AI products may deviate from original design concepts and negatively impact subjects. Over-reliance on generative AI may lead to outcomes that reflect deep learning concepts while the process is replaced by AI. If such situations arise, they may negatively affect the learning initiative and integrity of knowledge systems for the users.

On one hand, excessive reliance on AI will diminish users’ learning initiative. Prior to the popularity of generative AI, reports indicated that students’ reliance on question-searching tools led to declines in academic performance, as the process of solving problems with these tools effectively turned learning into a passive replication process. The current explosive functionality of generative AI surpasses traditional question-searching tools, producing more detailed and comprehensive answers through simple dialogues, inevitably impacting students’ learning initiative. Over time, students may develop a “technological inertia,” leading to situations where they struggle to start without AI, gradually becoming “feeding machines” devoid of independent thinking. If not addressed promptly, this could result in a vicious cycle of “dependence—decline in learning initiative—excessive reliance.”

On the other hand, over-reliance on generative AI may disrupt the integrity of students’ knowledge systems. Currently, students’ knowledge acquisition can be categorized into two broad directions: receptive and discovery learning. Knowledge gained through receptive learning is typically organized and logical, while discovery learning involves students finding patterns and connections under teacher guidance, forming relatively complete knowledge systems. However, information obtained from Chat GPT is often disordered or beyond cognitive levels, and during dialogues with generative AI, students frequently find themselves waiting for AI responses, leading to a decrease in knowledge exploration and hindering the establishment of a complete knowledge system. Furthermore, the basic logic of generative AI is based on human input, resulting in fragmented information output, which can lead to shallow memory of knowledge and hinder the formation of a complete knowledge system.

For instance, in music education, when students send numerous questions to AI, their brains receive a large volume of AI-generated answers. As the number of questions and answers increases, the complexity of the information received in a short time becomes overwhelming, making it difficult for the brain to accurately remember knowledge, potentially leading to confusion regarding music works. For example, mistakenly identifying Mozart as the composer of “Für Elise” is a low-level error. The information about musical works should be integrated with the historical context of composers or actual musical events (concerts) to construct appropriate knowledge frameworks for students. Given the vast number of musical works and different historical classifications, a simple Q&A format is unlikely to help students develop deep understanding and memory, often resulting in incorrect musical knowledge elements and loss of historical context.

(2) Reliability and Timeliness of Educational Materials are Weakened: The Technical “Black Box” Behind Information Delivery

Due to the opaque algorithmic logic behind generative AI applications like Chat GPT, there exists a technical “black box” that may generate false information on a large scale. Firstly, the reliability of educational materials presented by AI remains ambiguous. Users engaging with AI can only be sure of certain answers provided, but the algorithmic and cognitive processes remain unclear. This is primarily because generative AI is based on fixed algorithmic models designed by developers, and detailed technical parameters are often not transparent, potentially leading to errors and conceptual confusions in educational materials obtained from AI. Based on user experiences, AI can even fabricate non-existent references in formats like national standards or APA. Utilizing erroneous materials in education is akin to inserting a “Trojan horse” into students’ knowledge systems, misleading them.

Secondly, the timeliness of information pushed by generative AI is not strong. In education, the integration of “eternal” and current events is crucial. On one hand, “eternal” refers to the timeless curriculum philosophy, emphasizing the unparalleled significance of great works and thinkers, such as classics, famous quotes, and fundamental theorems. On the other hand, current events involve reflections on hot topics and educational reforms prompted by new educational philosophies. OpenAI has explicitly stated that Chat GPT’s knowledge is limited to September 2021, meaning that knowledge after that date has not been incorporated into the model, relying instead on information sent by users for learning. However, the veracity of user-submitted information is often difficult to discern, exacerbating the risks of misinformation, especially in subjects that require alignment with cutting-edge knowledge.

In music education, for instance, the increasing openness of education has led to a significant rise in the richness and production of musical works, making timeliness crucial. However, AI cannot quickly access the latest musical works, potentially providing outdated information that hinders music education. Additionally, copyright is vital in music, and the information provided by Chat GPT operates within a “black box,” which may infringe on music copyrights.

(3) Obscuring the Core of Interpersonal Communication: Diminishing Emotional Connections Between Teachers and Students

Fundamentally, the teacher-student relationship is a social relationship formed through mutual cognition, emotion, and interaction during shared educational activities. The widespread rise of generative AI may weaken the influence of core elements of interpersonal communication, disrupting normal teacher-student interactions and alienating these relationships. Firstly, from a cognitive perspective, it challenges teacher authority. The development of AI technology has made knowledge dissemination and acquisition more convenient. American educator John Dewey stated, “The method of acquiring knowledge in learning is more important than the knowledge itself.” The widespread use of generative AI transforms knowledge acquisition into a mechanical result, completely neglecting the importance of the process, leading some students to believe that having AI allows them to obtain more and better knowledge without the need for teachers. This situation arises from the disregard for the significance of the teaching process, undermining the professional status of teachers from a knowledge cognition standpoint.

Secondly, it increases the risk of emotional detachment between teachers and students. As previously mentioned, generative AI can answer most questions, diminishing the role of teachers in clarifying doubts, resulting in weakened emotional ties between teachers and students. When students view AI as a more capable tool than their teachers, they may reduce or even refuse to engage with teachers, endangering normal teacher-student relationships and rendering interactions mechanical.

In music education, for instance, students can more easily obtain results for melody creation, background information on works, author information, formal analysis, and harmonic structures from AI, making it difficult for them to patiently listen to teachers explain complex concepts like harmony. However, the essence of interpersonal communication lies in emotional empathy and intellectual exchange, and the emergence of generative AI reduces interpersonal communication to cold algorithms, inevitably leading to the alienation of teacher-student relationships.

In the future, the widespread development and application of generative AI like Chat GPT will be unstoppable. Establishing a good order of use and building a symbiotic bridge between education and AI is a crucial topic in current AI research.

(1) Suppressing Sole Reliance on AI: Promoting Educational Ecological Transformation and Implementing the Information + Teaching Concept

In the rapidly advancing era of AI, it is essential to promote the healthy development of educational ecology within the framework of “information + teaching.” Firstly, educational institutions such as schools should strive to transform educational ecology, advancing information-based and digital education. Currently, utilizing AI tools for intelligent learning and assisting in innovative tasks is likely to become a significant future direction in education. Therefore, schools should help students establish harmonious human-machine interaction models, fostering their survival skills in the information age and enhancing their ability to judge information, thereby minimizing the negative impacts of technological “black boxes.” Additionally, ethical and moral issues should be addressed, improving students’ self-discipline when using AI.

Secondly, General Secretary Xi Jinping emphasized in his important speech at the National Education Conference in 2018 that the essence of education is to address the question of “what kind of people to cultivate,” which is the most concentrated and vivid reflection of the essence of education. In light of the rapid development of AI, we should focus on “educating people” and enhance teaching reflection abilities through information technology, clarify individual competency levels, and strengthen the evaluation of “abilities” while downplaying purely knowledge and skill assessments, thus facilitating the smooth transformation of educational ecology.

In music education, for instance, while AI can assist students in melody creation and action arrangement tasks, stopping at this point poses risks, primarily reflecting insufficient creative components and educational value. By guiding students to delve deeper into creation based on AI-generated content, we can enhance students’ creative platforms, leading to the creation of more and better creative works and improving their creative practice competencies.

(2) Avoiding Solely Cognitive Approaches: Enhancing Embodied Learning Experiences and Utilizing Multiple Methods to Promote Deep Learning

As Merleau-Ponty stated, “The perceiving subject is not an absolute thinker; rather, it acts according to the connections between our body and the world, and between ourselves and our bodies.” In other words, our understanding of the world is embodied, embedded within it, rather than being purely cognitive as Descartes suggested. I believe that if education detaches from embodied experiences, it cannot deeply explore learning issues, nor can it discuss deep learning. Among the impacts of AI, there is a risk of leading learners towards “purely cognitive” states. Educators should enrich teaching methods related to embodied experiences, helping students gain more of these experiences. As early as 1916, Dewey proposed the educational concept of “learning by doing” in his book “Democracy and Education.” Over a century later, with the rapid development of AI, there is an increased risk of falling into “purely cognitive” approaches. Educators should continue to uphold the theory of “learning by doing” and enhance students’ embodied learning. For example, interdisciplinary thematic teaching and project-based learning can be widely implemented in the classroom.

In music education, teachers can conduct experiential teaching, such as allowing students to listen to symphonic music while teaching orchestration, or providing more hands-on opportunities for students to explore the physical principles behind music phenomena, such as constructing pitch-bearing instruments, further deepening their embodied understanding of theoretical knowledge. Activities like instrument-making can only be “informed” by AI on how to create them; they cannot replace the students’ hands-on experience. This not only strengthens students’ embodied learning experiences but also reduces their dependence on intelligence, returning the practice of deep learning to the students.

(3) Preventing the Decline of Interpersonal Relationships: Preserving the Core of Interpersonal Communication and Ensuring Harmonious Teacher-Student Relationships

Interpersonal relationships are social connections established through communication in production and life. Humans cannot exist outside of society, and thus cannot be separated from interpersonal communication. The three elements of education include educators, learners, and educational influences. Both educators and learners are subjects of interpersonal communication, making educational work essentially an interaction between subjects, a human-to-human activity. This activity encompasses knowledge transfer, emotional exchange, and intellectual collision.

We must recognize that technology is not a “panacea.” While technological advancement indeed compensates for certain aspects of human activities, such compensation typically addresses only a small portion of educational activities, such as knowledge and information. However, the work of educators involves more than just knowledge and information transmission; it is an emotional and creative labor form that far exceeds AI capabilities. The emotional and creative aspects of teachers can manifest in uncovering each student’s inner thoughts, understanding their personality traits, and focusing on shaping students’ values and character. This involves comprehensive communication with students, rather than solely addressing easily replaceable knowledge aspects. Moreover, teachers bear the responsibility of addressing issues related to AI, guiding students who excessively rely on AI out of their dilemmas and into reality, providing them with more opportunities for engagement and embodied experiences. This approach will better preserve the core elements of interpersonal communication, ensuring the harmonious and stable development of teacher-student relationships.

In the foreseeable future, generative AI will inevitably become a trend in the intelligent development of education, continuously influencing various aspects such as teachers and students, educational elements, and teaching models. Compared to “Internet+” and the “Metaverse,” generative AI offers more convenient usage, lower entry barriers, and richer intelligent forms. However, the emergence and application of any technology is not an isolated effect but rather the result of collaborative influences, and generative AI is no exception. Its powerful functionalities are built upon the foundations of “Internet+” and “Metaverse,” establishing educational interconnectivity and constructing a diverse educational ecology, leading to the widespread applications of generative AI today. Therefore, we must maintain the momentum of research on generative AI while also paying attention to the logical connections behind technology and the multidimensional collaborative effects involved.

Moreover, we should continue to explore ways to mitigate the risks brought by AI. The impact of AI on human society is a “double-edged sword”; whether it is “Pandora’s box” or “Aladdin’s lamp” depends on the mindset of the users. In the field of education, we are already in a highly intelligent era, having fostered many modern talents, educational philosophies, and tools, with AI playing an indispensable role. However, in the face of increasingly intelligent AI, we must maintain a critical perspective, keen thinking, and a rational attitude. On one hand, we should continue to uphold the core educational philosophy of “what kind of people to cultivate”; on the other hand, we must use AI judiciously, continuing to explore better teaching models and tools. Education is an essential component of national development, and we should not only focus on this aspect but also consider the overall development of various national endeavors. Therefore, users should ethically and socially responsibly enjoy the technological conveniences brought by AI, adopting a positive attitude to fully leverage AI for greater developmental momentum, thus providing an inexhaustible driving force for the prosperity of education in the intelligent era.

Author’s Institution: China Conservatory of Music; This article is adapted from “Chinese Music Education”