Terence Tao – How the world’s top mathematician uses AI

Summary

The video is a conversation between Dwarkesh Patel and mathematician Terence Tao about the history of scientific discovery, from Kepler to Newton and Darwin, and how AI is changing mathematics and science. Tao argues AI has made idea generation cheap but shifted the bottleneck to verification, validation, and evaluating partial progress. The discussion covers current AI math performance, formal proof tools like Lean, the future of hybrid human-AI research, and the difficulty of measuring scientific progress.

• The episode uses Kepler, Copernicus, Newton, and Darwin to illustrate how scientific verification and persuasion can lag idea generation for decades or centuries.
• Tao argues AI has reduced idea-generation costs to near zero, but verification, validation, and evaluation are now the main bottlenecks.
• Current AI tools solved roughly 50 Erdős problems, but pure AI one-shot progress has plateaued after low-hanging fruit.
• Tao sees near-term mathematics as hybrid human-AI collaboration, with AI strong at breadth and humans strong at depth.
• Formal proof systems such as Lean are enabling atomic proof analysis, refactoring, and possible future AI-assisted strategy languages.
• The discussion includes conditional scientific scenarios around the Riemann hypothesis and prime-based cryptography but makes no explicit tradable market calls.