Terence Tao’s “Math 2.0” speech warns that AI may flood mathematics with proofs humans cannot understand

Fields Medalist Terence Tao used an Oct. 9 talk at the California Institute of Technology’s Beckman Auditorium to argue that mathematics is entering a new phase he calls “Math 2.0.” The speech, originally planned under the title “Machine-Assisted Proof,” was rewritten after a burst of recent AI progress in mathematical research. The article says Tao’s central concern is not whether AI can produce correct results, but whether those results remain intelligible to human researchers. The piece lists several developments cited as signs of a turning point: OpenAI’s reported work on the Erdős unit distance problem in May; a Sept. 8 claim that an internal model coordinated roughly 10,000 AI agents over 88 hours to produce a finite-time singularity proof for the Navier-Stokes equations and formalize it in Lean; and an Oct. 6 GitHub release containing more than 700 AI-generated mathematical manuscripts covering 372 results. It also mentions Anthropic’s Claude and Google DeepMind’s AlphaProof. Tao’s answer is not to push humans out of mathematics. Instead, he argues that AI should be used to strengthen human understanding, explain difficult results more clearly, connect mathematical knowledge into evolving networks, and help researchers discover new principles rather than merely generate more opaque proofs.