#503 – AI Crisis in Math: Navier-Stokes, Riemann Hypothesis & Complexity Theory

Lex Fridman hosts mathematicians Ankur Moitra, Steven Brunton and Alex Kontorovich to examine AI’s potential to change mathematical discovery, proof and understanding. The show notes frame the discussion as a crisis moment, asserting that AI has solved Navier-Stokes, the “quasi-Riemann hypothesis,” the unique games conjecture and hundreds of other problems. However, they provide no supporting papers, proof details or independent verification, so those assertions should be treated as the episode’s framing rather than established results.

The guests bring complementary expertise: Moitra, at MIT and visiting the Institute for Advanced Study, works in theoretical computer science; Brunton, at the University of Washington, specializes in fluid dynamics; and Kontorovich, at Rutgers, specializes in number theory. The outline connects their fields through discussions of the Riemann hypothesis, Navier-Stokes, the Unique Games Conjecture and complexity theory. Lean and machine-checked proofs introduce verification as another dimension of AI’s role in mathematics, alongside the questions of whether AI can solve existing problems or create new mathematics.

The episode matters now because it places questions about AI capability alongside questions about what mathematical progress actually means. Its agenda moves from headline problems to how mathematicians think, how nature computes and why large language models work so well. That structure separates several issues often conflated in AI discussions: finding an answer, producing a checkable proof, creating a new idea and understanding the reasoning behind it. The notes identify these discussion topics but do not document the guests’ conclusions or specific disagreements.

The open questions include whether AI will “solve” mathematics, whether it can resolve P vs NP and which problems remain beyond its capabilities. The discussion also turns to whether mathematics is invented or discovered, Grigori Perelman and advice for young people. Without the transcript or cited technical results, the notes support a map of the debate—not a determination that the named mathematical problems have been settled.