Lex Fridman Podcast Lex Fridman Podcast

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

Oct 26, 2026 · 3h 42m

Summary

This emergency episode features mathematicians Stephen Brunton, Ankur Moitra, and Alex (number theorist) discussing the recent AI breakthroughs in solving major mathematical problems like Navier-Stokes and the Quasi-Riemann Hypothesis. They debate whether these results signal a crisis or a golden age, exploring the impact on graduate students, the value of human insight versus machine computation, and the future of "Centaur" mathematics.

Topics discussed

Introduction: The AI math crisis and recent breakthroughs Sponsorships and host's personal state Magnitude of the moment and Fields Medal implications Impact on students, postdocs, and intellectual loneliness Engineering perspective: Hope vs. uncertainty in applied math Motivation from known solutions and the 'solved axiom' Human progress vs. AI: Generalizability and abstraction Centaur math: Interacting with AI for beautiful theories The Doorman Fallacy and the value of asking questions Reworking institutions and the limits of AI proofs Adversarial examples and the space of discoverable math New ideas vs. convex hull: Depth of AI-generated proofs Impact of known truths on research organization Economic impact of polynomial-time factoring The nature of mathematical artifacts and lectures Mathematician independence and the joy of research talks Mathlib, Lean, and formalizing mathematics Formal definitions vs. intuition in calculus Explaining the Riemann Hypothesis and prime distribution Speed of progress and formalization challenges Human discourse, disagreement, and the Labor Day reaction Prime number mysteries and the interplay of addition/multiplication Technical limits of the quasi-RH proof methods Protecting ideas in the age of AI exploration Navier-Stokes equations and fluid dynamics intuition Blowup in Navier-Stokes and discovery of new phenomena Nature vs. technology in solving complex problems Formulation of the Navier-Stokes blowup problem 2D vs 3D flows and engineering insights from nature Fundamental unknowns in fluid equations and AI potential Learning fundamental principles from data for fluid systems Algorithmic lens and complexity in natural systems Optimization formulation and flow control challenges Black hole analogies and forcing in vortex dynamics SAT, P vs NP, and the nature of computational hardness Unique Games conjecture and the OpenAI announcement Remembering curiosity and the future of collaboration Post-victory depression and the Andrew Wiles documentary Thinking processes and the role of AI in learning Insights from natural systems and cross-field transfer Complexity of life vs. current AI capabilities Is math real? The existence of numbers and zero The professionalization of infinity in the 19th century Worst-case analysis pitfalls and beyond worst-case complexity Online algorithms, paging, and finding where to dig Nature abhors hard problems and LLM structure Subliminal learning and the Ouija board analogy AI dependency, verification, and cellular automata effects Practical ease of SAT problems and Terence Tao's reaction Centaur math, learning speed, and visual examples Advice for PhD students: Learn difficult abstract math Advice for young learners: Fundamentals and curiosity Navigating uncertainty and finding purpose in a changing world The value of stumbling and community in the AI era

More from Lex Fridman Podcast

#505 – GTA 6, Crime, Jewel Heists, Prison, Corrupt Cops, Mafia, Miami & Money
Nov 19, 2026
#504 – Russia-Ukraine War: Putin, Zelenskyy, Drones and the Future of Warfare
Nov 10, 2026
#502 – Psychiatry, Insane Asylums, Mental Illness, ECT, Lobotomies, Freud & Jung
Sep 17, 2026
#501 – DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux
Aug 26, 2026
#500 – Khabib Nurmagomedov: Dagestan, MMA, UFC, Islam, Conor, Fedor & Football
Aug 12, 2026
Listen ad-free on Castria