AI in the Classroom

“It’s time for an AI competition in answering questions. My question is: how to prove the equivalence of the second law of thermodynamics?” said Lu Diannan, the instructor of the Chemical Engineering Thermodynamics course, with a smile.

On the screen, the AI pondered for a few seconds before outputting an answer; below, students diligently answered questions on the ‘Rain Classroom’ platform, taking photos to upload. This scene occurs three times each class in Tsinghua University’s Chemical Engineering Thermodynamics course.

“On one hand, it optimizes students’ learning experiences and engages them in the classroom; on the other hand, by comparing the problem-solving approaches of AI and students, it highlights the differences in thinking and deepens students’ understanding of the essence of knowledge,” Lu explained.

In recent years, AI has quietly entered more classrooms at Tsinghua University, creating vivid scenarios that witness the deep integration of AI with teaching.

“Artificial intelligence technology is profoundly changing the inherent patterns of classroom teaching, student learning, and educational evaluation. In the wave of technological change, higher education bears an important mission and responsibility,” said Qiu Yong, Secretary of the Tsinghua University Party Committee. The university is actively promoting the deep integration of AI with education, reshaping the knowledge system for innovative talent and reforming talent cultivation models to convert technological development benefits into actual improvements in educational equity and quality.

AI as a Learning Companion

“Please introduce the current development of commonly used thermodynamic equations.”

“Here is a detailed analysis of commonly used equations and their current state…”

After class, Li Xuerui, a student in the Chemical Engineering department, asked the 24-hour intelligent learning companion for the Chemical Engineering Thermodynamics course to help with his homework. This was not about copying answers but engaging in project-based learning. Inspired by AI, he focused his topic on a deeper direction: how to improve existing state equations?

Li refers to AI as his learning companion, believing that it can help him conduct independent learning without replacing his own thinking and judgment. “In the age of artificial intelligence, the ability to learn independently remains core; AI is just a support and supplement, but human-AI collaborative learning should be an essential skill for every student,” Li stated.

In the fall semester of 2023, Tsinghua University launched a teaching reform plan empowered by artificial intelligence. Lu Diannan’s Chemical Engineering Thermodynamics course was included in the first batch of eight pilot courses for AI-assisted teaching. “AI can autonomously handle most basic knowledge questions, effectively improving learning efficiency and allowing teachers to focus on cultivating abilities and values behind knowledge, which is the most important task in cultivating innovative talent,” Lu reflected after more than two years of practice.

Data corroborated this perspective. In last semester’s Chemical Engineering Thermodynamics course, students interacted with the AI teaching assistant for an average of around six hours. A comparison showed that students who used AI for pre-class autonomous learning performed better in subsequent class tests or assignments than those who did not.

Before the “University Physics A” course, students asked their AI companions questions, and the system analyzed high-frequency questions from the entire class in real-time, generating a “Q&A card” to push to the teacher. In programming courses, AI acted as a teaching assistant, answering common issues like syntax errors and debugging thoughts in real-time. Currently, over 450 courses at Tsinghua University have integrated AI, realizing ten functional scenarios such as AI companions, AI teaching assistants, and lesson preparation assistants, covering pre-class, in-class, and post-class activities. This technology continually drives students to engage in innovative learning and interdisciplinary research, with personalized learning efficiency steadily improving.

As technology penetrates more classrooms, the core goals of educational reform have become clearer in practice. According to Peng Gang, Vice President of Tsinghua University, courses and teaching in the AI era need to rethink “what to teach, how to teach, and for whom to teach.” “The core is to make the growth experiences of every teacher, every course, and every student ‘irreplaceable,’ allowing universities to fulfill their educational value.”

Building a Multi-layered Training System

“How ‘perfect’ can a potato chip be?”

In the spring semester of 2025, Professor Mi Haipeng from the Academy of Arts opened a general course titled “Artificial Intelligence and Art Design.” A piece of work provided an answer. Three students defined a concept using DeepSeek, generated an image with Dream AI, and built a 3D model using Tripo AI, ultimately creating a piece named “The Most Perfect Potato Chip in the World”: thickness precisely at 0.88±0.02mm, porosity controlled at 32.7%.

The students also constructed a complete “hype ecosystem” around the potato chip work: creating an AI-driven conceptual artist, designing a virtual currency system called “Chip Coin,” and even planning a complete art auction.

“The characteristic of this course is to guide students from simple tool usage to deeper reflection, which is the core goal of general education,” Mi Haipeng noted with surprise, as students began to ponder the boundaries of AI creation and the impact of technological advancement on society.

By 2026, Tsinghua University had established an AI general education course system covering five directions and 57 courses, and built an “AI course matrix” of 162 courses, allowing students from various backgrounds in humanities, sciences, engineering, and medicine to find accessible entry points. Each course has its focus: “Artificial Intelligence and Law” explores data governance, algorithm governance, and AI supply chain security across six major modules, directly addressing cutting-edge regulatory issues in the intelligent era; “Robot Cognition and Practice” integrates advanced technologies from multiple disciplines, providing students with a systematic understanding, hands-on practice, and deep insights.

To enable students to truly bring their ideas to life, in 2025, Tsinghua distributed 1000 yuan worth of computing power vouchers to each student, providing the “fuel” for their AI sparks.

The multi-layered training system is gradually improving, laying a stepping stone for students with different aspirations. Starting in the fall semester of 2025, Tsinghua University will offer AI minor degrees and AI course certificate programs, breaking down departmental barriers and allowing students with extra capacity to systematically build AI skills beyond their major.

More specialized training will be carried out by the “Wuqiong Academy.” In 2025, the academy welcomed its first cohort of 171 students, aiming to cultivate the most innovative AI leaders through project-based learning and dual mentor support.

Establishing AI Infrastructure

“I want to learn about the development and challenges of phosphorus recovery technology in wastewater.” After class, environmental science student Sang Peiyang opened the “Beyond Classroom” platform, inputting his needs and starting a journey of autonomous learning.

Clicking “Generate Learning Plan,” a knowledge map unfolded before Sang: mainstream phosphorus recovery technologies, emerging phosphorus recovery technologies… eight knowledge points with clear connections. Each knowledge point is accompanied by detailed introductions and corresponding test questions, while tracking the learning progress.

“Beyond Classroom” is not just an ordinary platform; it is a powerful subject knowledge engine. To ensure AI is scientifically and efficiently integrated into education, starting in the spring of 2024, Tsinghua University has focused on building a subject knowledge engine, proactively proposing a three-layer decoupled architecture: model layer, engine layer, and application layer. This systematizes the collection and structural organization of vast subject knowledge, transforming general large models from “generalists” into subject “specialists” to achieve a problem or task-oriented deep learning model.

In the model layer, teachers and students can switch between various large models like DeepSeek and Zhipu Qingyan based on course needs; in the engine layer, training models with teaching materials uploaded by teachers builds a “subject knowledge engine” to address accuracy issues in specialized vertical fields; in the application layer, an AI workstation is created based on the “Rain Classroom” platform, allowing teachers to utilize ten AI functional scenarios without changing their teaching habits.

The “Beyond Classroom” that Sang Peiyang used is a dynamic knowledge base integrating all teaching materials accumulated since the establishment of the environmental science department and global publicly available research findings, constructing a cross-disciplinary knowledge map covering over 50,000 effective nodes and more than 100,000 relationships. If you focus on “sponge cities,” the system automatically associates it with fluid mechanics, water treatment technologies, and other interdisciplinary content, generating a personalized knowledge network just for you.

“Previously, credits could only be earned based on complete physical classroom attendance; now, autonomous learning driven by questions and intelligent navigation can achieve the same,” said Professor Yue Dongbei from the environmental science department. By guiding learning objectives, AI can plan differentiated learning paths based on individual backgrounds and dynamically track and adjust them, aiding in the precise cultivation of talent.

In May 2025, the first batch of subject knowledge engines for integrated circuits, industrial engineering, environmental engineering, and other disciplines were officially released, with related construction work in 20 departments gradually underway. Currently, the subject knowledge engine has signed agreements for co-construction and sharing with 80 universities nationwide, including Peking University and Nanjing University, aiming to transform outstanding results from individual schools and teachers into valuable resources serving multiple schools and teachers.

While promoting technology empowerment, Tsinghua has not overlooked the establishment of ethical boundaries. In 2025, the university formulated the “Guidelines for the Application of Artificial Intelligence in Education at Tsinghua University” and established a full-process ethical review mechanism for AI-related matters, proceeding with a “proactive yet cautious” attitude to steadily advance in balancing technology and humanities.

On the foundation of strengthening its own base, Tsinghua University further pushes its concepts and practices globally, aiming to foster international consensus on the governance of AI education. In December 2025, at the World MOOC and Online Education Conference, it released another important consensus in the global higher education community—the “Mexico City Declaration,” further focusing on the trends of higher education reform in the AI era, adhering to five principles: student-centered, quality-first, inclusive equity, ethical safety, and collaborative exchange, advocating five actions to jointly create the future of intelligent education.

“Cultivating virtue and nurturing talent is the foundation of a university’s existence. We hope to further stimulate teachers’ intrinsic motivation based on the explorations of the previous stage, focusing on how to leverage artificial intelligence to expand new possibilities in education, allowing students to gain deeper and more enlightening learning experiences, and promoting technology to serve talent cultivation more precisely, contributing to the digital transformation of higher education in the new era,” said Li Luming, President of Tsinghua University.