What it Means to Be a Mathematician When AI Does the Math
TL;DR
IEEE Spectrum frames AI in math as a shift from tool to co-researcher: DeepMind and OpenAI reached IMO gold-level performance, Aletheia produced Ph.D.-level results, and OpenAI disproved a combinatorial geometry conjecture. Proof assistants such as Lean, Isabelle and Rocq are becoming the trust layer as LLMs help turn informal proofs into checkable formal code.
Nauti's Take
The achievement list has a PR-race smell, but the sober core is real: math gives AI a cleaner verification layer than most fields. That makes it a proving ground for serious human-machine work.
The bad reflex would be to automate away every slow stretch of thinking. That friction is where taste forms, and taste decides which proof matters, which shortcut is suspicious, and which question deserves years.
Briefingshow
Mathematics is a sharp test case for AI because results can be formally checked and mistakes cannot hide behind taste. If AI can contribute there, the same pressure will hit many knowledge jobs: value moves to choosing problems, understanding outputs, auditing proofs and training people. Powerful models and formal libraries could concentrate research capacity.
Video
Sources