Grant Sanderson – AI and the future of math
Summary
Dwarkesh Patel interviews Grant Sanderson about AI’s rapid progress in mathematics, noting that while models now excel at benchmarks like the IMO, true AGI requires more than solving existing problems. They discuss whether AI can generate novel conjectures and definitions, using Galois theory as an example of insights that took decades to verify. The conversation explores if AI proofs will be understandable or alien, and how the role of mathematicians may shift from proving theorems to consolidating and explaining AI-generated insights.
Topics discussed
Introduction: AI progress and the International Math Olympiad
AI performance on IMO categories and the nature of mathematical creativity
The Unit Distance Conjecture and the difficulty of benchmarking conjecture generation
Historical case study: Abel, Galois, and the invention of Group Theory
The 'Fall of the Theorem Economy': Automating proofs vs. finding insights
AI as a connector: Finding links between disparate mathematical fields
Technical challenges: Verifiers, Lean, and the credit assignment problem
Automated theorem proving and the role of formal verification
AI limitations: Theory of mind, social intelligence, and learning
The future of mathematicians: Curation, teaching, and economic value
Listen ad-free on Castria