When Computers Write Proofs, What's the Point of Mathematicians?

The video challenges the classic view of mathematics as a self-contained edifice of certainty, arguing that proof ultimately rests on unprovable philosophical foundations. It explores how AI tools like Lean are transforming mathematical practice by verifying proofs step-by-step, acting like a persistent, questioning colleague—as demonstrated when it probed Peter Scholze's uncertainty. The central question is whether machines will move from assisting proofs to leading them, with computer-generated proofs still nascent but promising. This prospect raises deep anxieties about mathematicians' identity, purpose, and values, potentially reducing their role to hoping the computer verifies their ideas rather than achieving certainty themselves.

• ▶ 0:19 The classic view of mathematics as an unshakable edifice built solely on axioms is a beautiful "fantasy" that is "not even close to true."
• ▶ 0:29 Top mathematicians now ask "massive questions" about proof—what we want from it, what we believe when something is proved, and how AI changes that practice.
• ▶ 1:55 Aristotle's idea of "primitives" and self-evident axioms shows that proof ultimately rests on unprovable foundations, blurring the line between mathematical certainty and philosophical assumption.
• ▶ 2:50 Traditional mathematical verification relies on a library model: published claims are checked by consulting stored books, with verification meaning matching recorded knowledge or revising it.
• ▶ 3:14 Modern AI proof assistants like Lean embed verified results in the program itself, letting mathematicians input proofs that Lean then checks step-by-step against its own rules.
• ▶ 3:42 Lean acts like an obnoxious colleague who won't let go, persistently asking clarifying questions—and in the Peter Scholze example, it questioned exactly the parts Scholze was unsure about, confirming that the tool was asking the right questions.
• ▶ 4:44 The central question is whether machines will change mathematics—moving from merely following or suggesting proofs to leading proofs themselves.
• ▶ 4:58 Computer-generated proofs are still in their infancy and have not yet done much of consequence, though exciting new ideas suggest genuinely interesting machine-generated proofs may soon emerge.
• ▶ 5:21 This prospect is frightening for mathematicians’ identity: if machines handle proofs, it raises deep questions about professional purpose, training, and values—and risks turning mathematicians into “more like physicists” who simply hope the computer verifies their ideas.

• ▶ 0:00 The Axiomatic Fantasy and the Nature of Proof (0:00 - 2:53) - - Explores the undergraduate ideal of axioms, philosophical doubts about proof, Granville's graphic novel project, and Aristotle's primitives.
• ▶ 2:53 Traditional and AI Proof Verification (2:53 - 4:44) - - Contrasts traditional library-based verification with modern AI proof assistants like Lean, and notes the value of being challenged.
• ▶ 4:44 Machines and the Future of Mathematics (4:44 - 6:35) - - Asks whether machines will transform mathematics, considers threats to mathematicians' identity, and admits the future is hard to predict.