Fields Medalist Yu Deng Shares How AI Assists in Mathematical Proofs
Mathematician Yu Deng revealed that he uses AI tools like GPT to assist in his research, noting that GPT recently helped him resolve a specific mathematical sub-problem that he had been stuck on for days. He emphasized that while AI makes research more convenient, it cannot replace independent thinking and verification. This highlights the growing role of Large Language Models (LLMs) in advanced scientific research, demonstrating that even top-tier mathematicians are adopting AI to accelerate proof exploration. It underscores a future of human-AI collaboration where humans design the conceptual framework and AI assists with technical derivations. Although the proof generated by GPT for the special case could not be generalized to the main proposition, it provided valuable insights that helped Deng move forward. Deng cautioned early-career researchers and students not to blindly trust AI-generated proofs without performing rigorous independent checks.
## BACKGROUND
The integration of artificial intelligence in scientific discovery, often called AI for Science (AI4S), has accelerated with the rise of LLMs. In mathematics, automated theorem proving has traditionally relied on formal logic systems, but modern AI tools are now being used to assist with informal reasoning, generating lemmas, and exploring proof pathways.