After OpenAI announced their solution to the Navier-Stokes problem, one of the Millennium prize problems, many mathematicians and mathematical researchers are feeling an existential crisis. For example, Fields medalist and celebrity mathematician Cédric Villani said after the announcement of OpenAI's solution to the millennium problem: “I was shaken. An atmosphere of the end of history. It's a cataclysm unlike anything mathematics has ever known”1.
Fields medal winner James Maynard gone from being an AI skeptic to absolutely frightened: “The people who sound like they're talking from a sci-fi novel are the people who've been making much much more accurate predictions. The risks posed by artificial intelligence are an emergency that should not be dismissed as hype”2.
Thanks for reading! Subscribe for free to receive new posts and support my work.
SubscribeEven the most well-known mathematician of our times, and generally enthusiast of any new tool, Terence Tao calls for a slowdown in AI for Math in his talk: “we have to slow down AI. the pace is insane, and there's no reason to be this fast — no reason at all”3.
AI has come for the forest, and the great trees are falling first.
However while the great trees fall, the small ones can still be there, even flourish. Perhaps folks, who have been bestowed the highest honor of their field by the time they are 40 (which, in the the 21st century, is really the new 20) do not know how to wither a storm?
On the other hand, there are people who love mathematics for its beauty alone, with no hope of a reward ever. They fail often, and are devalued, insulted, defeated. Rejection is a Monday for them. Can they request the greatest minds of our times to not react too fast, not opine every week, not go from denial to doomsday cultism in matter of days? You do not have to open your heart to every twitter announcement. Let’s wait and observe.
Some great things have happened. And we can objectively look at them.
- The interest in research mathematics is all time high. So many people have learned about Hodge conjecture in the last few days.
- Companies like OpenAI and Anthropic employ great mathematical minds. It is far from just button pushing.
- Everyone needs a lesson in humility.
See, these three things are uncontroversially true.
Solving a great mathematical conjecture may not have even long-term bottomline value to human condition. In short term, nothing will change even if we are provided with solutions of most of the cherished math research problems by an oracle. The AI companies are interested in this to show that their frontier models have high reasoning capabilities, and thus keep their valuation. Ultimately, if they cannot show a bottomline utility they will be out of business.
We should all cheer OpenAI and Anthropic to get a Nobel prize in medicine, and cure cancer. Only a negligible fraction of humanity is affected by their research math adventures.
Scientists and engineers all need college-level math knowledge, and AI is not replacing all of them in a short time. Math educators are actually in quite safe position, i.e., no more in danger than others disciplines.
In research math, we have already started to value understanding and exposition4. It is true that credit assignment and evaluations of abilities will be much difficult going forward. But we will certainly not solve that by panicking. It is a much bigger problem in other fields where AI slop is much easier to produce5.
The next few months will be as exciting as ever in math. I hope the little trees and weeds and fallen leaves will play a big role in carrying life to a new era.
p.s. Tao’s blog is the platform where multiple mathematicians/thinkers are expressing their opinions. It is highly recommended, but let’s not get too depressed by some posts.
1https://www.lemonde.fr/sciences/article/2026/09/18/les-avancees-de-l-ia-provoquent-une-crise-inedite-chez-les-mathematiciens_6776810_1650684.html
234https://terrytao.wordpress.com/2026/09/18/if-math-is-more-than-proof-we-need-to-better-celebrate-the-rest-of-it/
5https://www.reddit.com/r/MachineLearning/comments/1wks0dv/iclr_2027_submission_50kd/