Tao and Gowers on AI in Math: Lacking the 'pruning' ability for deep search trees
新智元 · wechat · 2026-08-17
Mathematicians Timothy Gowers and Terence Tao analyze the role of AI in mathematics following OpenAI's model Astra results.
Key Insights:
- Blurred Line between Counterexamples and Theorems: AI excels at finding counterexamples (e.g., disproving the Erdős conjecture), which Gowers argues is structurally similar to proving theorems, differing mainly in psychological expectation.
- Search Strategy Divergence: AI relies on massive trial-and-error (trying standard examples, random attempts), effective for pattern-rich problems. However, humans excel at 'pruning' deep search trees using intuition—a capability AI currently lacks.
- Missing 'Natural Friction': Tao notes that AI-generated proofs are too polished, erasing the 'natural friction' found in human proofs that highlights difficulty, preventing models from learning true judgment.
- Evaluation Criteria: Gowers posits that the real milestone will be when AI proposes a method that shifts the field's workflow, like the 2016 cap-set solution.
Related event: Fields Medalists Weigh In on AI's Strengths and Limits in Math(2 posts)→
More from AGI Musings
- Testing alongside an agent in a p2p room feels oddly real, says dev — sull · 2026-10-16
- Yudkowsky's 2008 takeoff scenarios resurface as AI-advances-AI takes hold — jd_pressman · 2026-10-16
- Guardrails can make risky delegation work — but it takes a full-time team, says iandanforth — iandanforth · 2026-10-16
- e/acc founder: agentic models are autocomplete for digital actions — beffjezos · 2026-10-16
- What can you truly delegate? Only tasks with unambiguous, verifiable results — iandanforth · 2026-10-16
- Casper: Understanding 90% of neurons ≠ understanding the network — StephenLCasper · 2026-10-16