Terence Tao explores the dilemma of AI-generated proofs that no human understands

QGallouedec · x · 2026-08-20

Renowned mathematician Terence Tao discusses the profound impact of AI on mathematics in a new paper. He notes that there are already AI-generated proofs verified as correct that no human—including those who prompted the AI—truly understands. Tao warns that we may soon face a verified proof of a major result that no human understands well enough to explain, raising questions about what constitutes a contribution to the field if correctness is insufficient without human comprehension.

Related event: Terence Tao's "Mathematics in the Age of AI": Assuming AI Can Already Do Research-Level Math(6 posts)→

Original post →

More from AGI Musings

AGI Musings channel →