AI Solves 70% of Research Math Problems at Low Cost

Terence Tao warns of a mathematical crisis as AI generates publishable proofs for 7/10 problems at $10-$1000, shifting scarcity from discovery to human verification and explanation.

Woofun AI reports that mathematician Terence Tao addressed the International Congress of Mathematicians, asserting that AI will soon handle significant research-level tasks affordably. Recent First Proof tests showed AI systems producing publishable solutions for seven of ten new problems, with computing costs between $10 and $1,000 per task.

Tao predicts a shift from "proof scarcity" to "proof abundance", where the critical bottleneck becomes human verification and elucidation. He argues that the ability to assess significance and clearly articulate results is now more valuable than initial problem-solving, insisting that unarticulated AI-generated proofs should not be published regardless of verification status.