AI is already changing mathematical research, but not replacing mathematicians

APPSO · wechat · 2026-07-23

This long article argues that AI is already reshaping mathematics work, but not replacing mathematicians.

It opens with the 2025 Fields Medal, noting that two China-born mathematicians, Deng Yuhan and Wang Hong, were awarded in the same year for the first time.
The article uses that backdrop to ask what mathematicians do when AI can already solve problems, check proofs, and search large theorem libraries.
A key distinction is drawn between writing down an answer and producing a beautiful proof: shorter code or proofs do not automatically mean deeper understanding.
Interviewed mathematicians say AI has moved from being a tool for solving exercises to a daily collaborator in research workflows.
The piece argues that AI is strong at pattern search, proof exploration, and formal verification, but still weak at judging which problems matter, explaining why a proof is elegant, and turning isolated results into theory.
It concludes that the next generation of mathematics will likely be done with AI in the toolset, while the rare human skill becomes choosing the right questions and interpreting what answers mean.

More from AGI Musings