On May 12, 2025, "Teacher · China" NO.63 Academic Salon, organized by Center for Teacher Education Research, Beijing Normal Univeristy (CTER-BNU), was successfully held in Room 623 of the Yingdong Building. The event featured Associate Prof.  Lu Yu from the School of Educational Technology, Faculty of Education, Beijing Normal University, who shared insights on the topic "Generative AI-Driven Educational Innovation and Practical Research." The session was chaired by Associate Professor Liao Wei from CTER-BNU.

Generative artificial intelligence (AIGC), as one of the most groundbreaking frontier technologies in AI, is profoundly reshaping the educational landscape. Prof. Lu Yu provided an in-depth analysis from three dimensions: technical principles, educational applications, and cutting-edge research.

1. Understanding Generative AI: Core Concepts and Development
Prof. Lu began by using ChatGPT as an example to explain the fundamental concepts and core technical architecture of generative AI in an accessible manner. The development and application of AIGC can be divided into two stages: pre-training and transfer learning, with the latter being more critical in education. He then compared two major AI models—ChatGPT and Deepseek—to illustrate the current state of large-scale AI development.

For instance, ChatGPT evolved rapidly from GPT-1 to its current iteration, characterized by massive parameters and high computational demands. Regarding Deepseek, Prof. Lu emphasized the need for a balanced perspective—while it offers a high-performance multilingual model accessible to the public and reduces localization costs, its advantages should be evaluated objectively.

2. Innovative Applications in Education
Prof. Lu highlighted how generative AI is transforming both K-12 and higher education. Noting that two of OpenAI’s six flagship applications are education-related, he pointed out that, unlike past technologies, generative AI is proactively integrating with educational scenarios. This is due to AIGC’s unique capabilities: Inspirational content generation, Sequential task execution, Contextual dialogue understanding, Programming language interpretation.

Through case studies, Prof. Lu demonstrated how these capabilities are reshaping educational demands, prompting educators to reconsider teaching philosophies, objectives, and curricula in the AI era. He also showcased practical applications, such as using "Zhipu Qingyan" to create teaching agents, assist lesson planning, and support academic research, allowing attendees to explore these tools firsthand.

3. Latest Research Trends and Future Directions
In the final segment, Prof. Lu discussed emerging research in AI in Education (AIED), including key academic journals and conferences. He outlined seven major research directions for AIED in 2025 and shared his team’s work in three areas: Adapting machine learning techniques (e.g., fine-tuning, prompt engineering) for educational tasks;Addressing ethical and fairness concerns in AIED;Designing AI curricula for schools and supporting AI-specialized teacher development.

Challenges and Opportunities
Prof. Lu concluded by noting that while large AI models are currently used mostly in general domains, future focus should be on specialized applications, such as tailored models for education. He also cautioned against overestimating AI’s capabilities, emphasizing the need for rational and regulated adoption.

The salon concluded on an inspiring note, with Prof. Lu’s comprehensive presentation—from technical foundations to real-world applications and future trends—highlighting the vast potential of AIGC in education. His insights encouraged educators to embrace new technologies responsibly, fostering innovation to cultivate future-ready talent.