#503 – AI Crisis in Math: Navier-Stokes, Riemann Hypothesis & Complexity Theory
Summary
Lex Fridman hosts an emergency episode with mathematicians Stephen Brunton, Ankur Moitra, and Alex to discuss the crisis sparked by AI solving major problems like Navier-Stokes and the Quasi-Riemann Hypothesis. The guests debate whether this marks the end of human mathematics or a new "centaur" era, analyzing the impact on graduate students and the shifting value of asking questions versus proving theorems. They explore the "doorman fallacy" to argue that while AI excels at execution, human insight remains crucial for defining meaningful problems and understanding deep mathematical structures.
Topics discussed
Introduction: AI crisis in mathematics
Sponsors and host's personal state
Magnitude of AI breakthroughs in math
Impact on students and the 'solved axiom'
Human role: asking questions vs. solving
Centaur math and the doorman fallacy
Adversarial examples and AI limitations
New ideas vs. convex hull of known results
Impact of proofs on research organization
Mathematician independence and research talks
Lean, Mathlib, and formal verification
Explaining the Riemann Hypothesis
Personal reaction to AI progress speed
Navier-Stokes equations and fluid dynamics
Blowup problem and existence of solutions
Complexity, vortices, and AI in fluids
P vs NP and Unique Games Conjecture
The gold medalist problem and curiosity
Thinking processes and natural systems
Philosophy of math and history of infinity
Worst-case analysis and natural data
Subliminal learning and AI safety
Verification, cryptography, and Terence Tao
Advice for students in the AI era
Purpose, uncertainty, and human community
Listen ad-free on Castria