A small episode from this summer keeps coming back to me.
At the National Academies meeting Organizing Mathematical Knowledge in the Age of AI and Formalization, David Spergel discussed Japan’s participation in the Department of Energy’s Genesis Mission. According to contemporary MIT-hosted notes, Japan became a partner “on condition that AI for math be included as a priority area.”
The words “on condition” are interesting. Apparently mathematics did not simply have its place at the table. Someone had to insist on it.
Let me speculate. Institutions such as the Department of Energy naturally think in terms of missions, engineering goals, infrastructure, and deliverables. There is nothing wrong with that. But in my experience, many engineers have felt that they already knew the mathematics they needed. They were not always particularly interested in mathematics as an independent intellectual force.
AI makes a stronger version of this attitude possible: Why involve mathematicians in setting the direction? AI will do the mathematics for us.
Whether AI can actually do “all the math” is almost beside the point. If the people making institutional decisions believe that it can, mathematicians may be invited only after the important decisions have been made to check results, formalize arguments, curate data, or maintain infrastructure. The questions, goals, and standards will have been chosen elsewhere.
At the same time, the widespread political concern about AI may offer an unexpected opportunity. “Humans must lead AI” is becoming a plausible political principle. In our field, this should mean something more specific:
Mathematicians must lead mathematical AI.
This opportunity may not remain open for long. Governments, companies, and engineering organizations are already creating programs and deciding priorities. The Japanese condition shows that these priorities are still negotiable but also that mathematics will not necessarily be included unless somebody speaks for it.
I hope our learned societies the AMS, the DMV, and their counterparts elsewhere are not asleep at the wheel. This is a moment for them to become active: to formulate standards, represent mathematical values, defend the education of young mathematicians, and insist that mathematicians participate in deciding what mathematical AI is supposed to become.
Unfortunately, withdrawal currently seems to be en vogue among many mathematicians. It is presented as a defense of human authorship, mathematical understanding, and intellectual independence. These concerns are serious and deserve serious discussion. But refusing to participate in shaping mathematical AI will not prevent its development. It will merely leave the consequential decisions to technology companies, engineers, government agencies, and others whose priorities may be quite different from those of the mathematical community.
I am occasionally reminded perhaps unfairly of the Judean People’s Front Crack Suicide Squad in Monty Python’s Life of Brian: it arrives at the decisive moment with admirable conviction and promptly removes itself from the action.
The joke makes a serious point. We should not defend human mathematics by leaving the field. If mathematicians withdraw, the field will not remain empty. Others will decide what mathematical AI is for, what standards it follows, and what place remains for human mathematicians.
The opportunity is here. It would be unfortunate if the mathematical community squandered it.
Disclosure. The argument and observations in this post are mine. I used OpenAI Codex to help organize and edit the text, check the cited sources, and draft portions of the prose. I reviewed and revised the final text and take responsibility for it.
An extended version was posted on the arXiv.
Received 21 September 2026.
Add to the discussion