Fwd: AI in mathematics: updates

[Siddhartha Gadgil]

Just forwarding something I sent to the maths department.

Siddhartha

Dear Friends,

        A lot is happening in AI for mathematics in recent times, even just this month (February 2026). Some highlights:

• Aletheia:  This framework using Gemini has done some non-trivial mathematics, see https://arxiv.org/abs/2602.10177, especially the case studies
• First proof: This was a test set up by a group of leading mathematicians with the tasks being non-trivial lemmas in their research they had proved but which were not public. An excellent account of this and more was written by Daniel Litt: https://www.daniellitt.com/blog/2026/2/20/mathematics-in-the-library-of-babel
• A quote from that essay: "In March 2025 I made a bet with Tamay Besiroglu, cofounder of RL environment company Mechanize, that AI tools would not be able to autonomously produce papers I judge to be at a level comparable to that of the best few papers published in 2025, at comparable cost to human experts, by 2030. I gave him 3:1 odds at the time; I now expect to lose this bet."
• In my opinion, Daniel Litt is both very knowledgeable and honest, but his public estimates of AI capabilities tend to be at the low end of those from people who are honest and knowledgeable (many mathematicians will estimate lower but that is usually ignorance and sometimes disingenuity). In any case here is a quote from Bartosz Naskręcki about this Daniel Litt essay: "Please spend some time to read this excellent entry by one of the most prominent algebraic geometers of the modern generation. I agree with all the predictions of Daniel. Actually, I am a bit sad, because I thought Daniel's predictions were much more reluctant and conservative than my own. In 5 years we will see that most human mathematicians will not participate in mathematical research as we know it today. Only a few, if any, will remain in the competitive game of designing and proving new theorems."
• At the other end of the spectrum from Daniel Litt is Christian Szegedy (again restricting to those in whom I have some faith). On the Daniel Litt article:  "A fair assessment. It looks like the gap in our timelines is narrowing. I also agree that there is more to math than just solving well-defined problems. I also think that it will take until 2027/28 until all other aspects of math research will be performed by AI at a human expert level."
• Christian Szegedy's timelines: "What we will see this year is that all puzzle-solving aspect of math will be delegated to AI within the next two years. By the end of the decade, all other aspects (conjecturing, modelling, theory building) will be taken over by AI. This year I expect AI to become as good or better than most but the top 100 mathematicians in the world in terms of proving and solving problems."

Addendum: quoting a young mathematician Ashvin Swaminathan (ashvin-swaminathan.github.io/home, postdoc at Harvard) on twitter reacting to Daniel Litt's essay:

"This is an excellent take on the current state and future direction of AI in math. As someone about to go on the job market, I'm left with a lot of questions about the longevity of research mathematics as a profession outside of the most elite institutions. I think mathematicians with tenured positions at elite universities will be fine for a long time to come. These people will serve as curators for the field, guiding AIs toward problems of interest. Crucially, they are the most likely to be able to secure funding for their work. At a time where the public has started to question the idea of taxpayer-funded research, where pure math in the US has lost > 50% of govt funding, private money is of the essence. And very few people can walk into the office of wealthy donor and convince them to fund their work. Most of us rely on govt grants to attend grad school and support our research. How long will that last? The general public seems disinclined to fund research they don't understand, especially research of no discernible practical value that may very soon be largely automatable. One might ask whether research mathematicians can support their work through teaching. But that ignores the dramatic impact that AI is having on teaching. A university degree may very soon be worthless, given how many students use AI to do their work for them. There's a reason why the number of tutoring requests I've received, and the number of students who attend office hours, has decreased dramatically. There's a reason why students do incredibly well on homeworks and far worse on exams. Many kids are graduating with no real skills. All this is to say that us young mathematicians without the security of permanent positions have a lot to weigh in deciding what to do with our lives. The dream of a stable university job may very quickly become a thing of the past."

Regards,

Siddhartha