Can AI do every PhD problem in a weekend?
Alessio Figalli won the 2018 Fields Medal and teaches at ETH Zurich. In September, he and 24 other Fields Medalists signed a declaration, A Severe Misalignment of AI in Mathematics, warning that AI labs’ race to solve math problems as benchmarks could harm the field. I sat down with him in late October, two days after OpenAI announced its Yang-Mills proof.
The moment
Alessio says AI can now do every basic PhD problem in hours or days. So the question is no longer whether machines can prove theorems. It’s who will still understand them. A proof nobody understands can be worse than no proof.
We also talked about why Jim Simons got stuck in dimension 8, how a technique for minimal surfaces cracked the problem of melting ice, what happens when 10,000 AI agents race human mathematicians, and why AI could empty the PhD pipeline.
Take this with you
The next time AI hands you an answer, try explaining it to someone else. If you can’t, you don’t own it yet.
Until next time,
Brian
Missed Monday? Aristotle was a genius. But he couldn’t count teeth. What an ancient authority can teach us about trusting AI. Read this week’s musing
Â
Advertisement
Tomorrow’s prize goes to a discovery made decades ago. Here’s how the winners made it: Focus Like a Nobel Prize Winner!
In Into the Impossible, Volume 2, I sat down with Nobel Prize winners to find out how they actually work: how they protect their attention, pick problems worth decades of their lives, and keep going when nobody else believes the result. Their answers are full of practical lessons you can use to concentrate your creativity and ignite your career, whether or not Stockholm ever calls.
Get your copy on Amazon and learn from the laureates themselves.