From WIRED · Isabella Ward · · 3m
A mathematician cries over AI's math wins
Steven Strogatz on the week AI cracked problems his field spent decades loving, and what that costs the people who loved them.
In brief
Steven Strogatz on the week AI cracked problems his field spent decades loving, and what that costs the people who loved them. AI is answering the questions mathematicians built their lives around, and the open question is what human mathematicians are for once the answers arrive without them. Originally reported by…
The tears came first
Cornell professor Steven Strogatz starts to cry describing a single week of AI breakthroughs in mathematics. The science thrills him. The behavior around it does not.
“The science is thrilling, but there's a lot of human unpleasantness going with it.”
A 90-year-old problem falls
OpenAI said it threw tens of thousands of agents at a math problem carrying a $1 million prize, and solved it. The result still needs independent verification.
A credit fight breaks out
Mathematician Tristan Buckmaster says OpenAI rushed the problem after learning of his work with Anthropic researcher Levent Alpöge, and tried to shape who got credit.
Labs racing toward IPOs
Strogatz reads the rupture as a symptom: corporate labs chasing headlines ahead of enormous public offerings, with mathematics as the scoreboard.
Claude proves 29,500 theorems
Anthropic said Claude formalized an existing proof of Fermat's Last Theorem by proving 29,500 small theorems, work other mathematicians had been grinding at for years.
No longer at the frontier
Alex Townsend used ChatGPT to crack a decades-old linear algebra problem. He is also 15 years into a career whose peak skill just got passed.
“I'm the one with an AI agent, which feels very different actually. And I feel totally threatened by it.”
Marketing, not engineering
Strogatz is blunt about Navier-Stokes existence and smoothness: arcane, useless to a working engineer, and valuable mainly as proof one lab beat the other.
“Nobody cares about the Navier-Stokes singularity problem; only a tiny subset of pure mathematicians care about that.”
Who should get the million
He wants Córdoba and Martínez-Zoroa considered, and makes the case for Buckmaster, who posted solutions to three closely related problems days before OpenAI.
“They were on the trail, and I think they would've gotten there, but we don't know.”
Humans as translators
Machines produce proofs nobody can read. For now, explaining them back in human terms is a job only experts do well. Strogatz expects that to fall too.
“Are we going to be the great interpreters of the things that the machines do?”
Mess is the moat
The closer a field sits to the real world, the more it resists AI. Economics, sociology, international relations: safe for a while, because they are messy.
Who has taste
Infinitely many things are provable, but only some are interesting. Machines have shown no sign of mathematical taste yet, and no obvious reason they cannot learn it.
Devastated or democratized
The questions mathematicians love are getting answered. That is a loss if you loved the struggle, and an opening if you only ever wanted in the door.
“Pure math is being devastated or revolutionized, depending on your point of view.”
The canary in the coal mine
We still play tennis and chess badly and happily. But nobody funds a beautiful game, and math may be the first field to find out what losing understanding feels like.
“I feel like we're the first battleground here in math where we're on the brink of losing human understanding.”
The short version
AI is answering the questions mathematicians built their lives around, and the open question is what human mathematicians are for once the answers arrive without them.
