Jacob Tsimerman is swapping mathematics for AI
Simons Foundation
Mathematicians have been shocked and startled by the rapid rate at which AI is solving open problems in their field, causing some to fear their jobs are becoming obsolete. But Jacob Tsimerman, who won mathematics’ most prestigious award just a few weeks ago, says the stakes are even higher than that. The technology could radically change the lives of everyone and even put them at risk – so he is leaving maths to work on AI instead.
Tsimerman, who is currently a professor at the University of Toronto in Canada, is one of only 68 people to ever win the Fields Medal, a prize so prestigious that is often referred to as the Nobel Prize of mathematics. Every four years, it recognises up to four mathematicians for their achievements. Crucially, they must be under 40 years old and so it effectively anoints the next crop of mathematical leaders who will shape their field.
Winning the Fields Medal gives Tsimerman the freedom to pursue any mathematical career he wants but instead he is leaving the subject and moving to OpenAI where he will work on AI safety.
“AI is at a point where like if you’re gonna get on board, it’s probably the time because things are happening pretty quick,” he says, talking to New Scientist at the International Congress of Mathematicians (ICM) in Philadelphia, where the Fields medal was announced.
Tsimerman says that he has been hearing about AI for years and has been waiting to see whether it will ever actually become as useful and as powerful as some researchers had been arguing. He has now started using it himself and has observed open problems in mathematics that he may have once considered working on get resolved by AI. And then there’s the safety breaches such as the incident with OpenAI’s hacking agent going rogue just one day before the ICM. “It happened this week, it happened last month, there is no shortage of these things,” says Tsimerman.
Because of the speed of progress, Tsimerman is concerned about where it could all be headed. “The risks are very big if you’re concerned about humans losing control or dying,” he says. “I’m a human, and I know humans, we live here. So, I want to protect them.”
In 2025, he co-authored a paper exploring the different scenarios where AI use leads to “omnicide” or human extinction.
The scenarios are broad in scope. There is unintentional omnicide where AI technology is indifferent to humans and so deprioritises growing food to devastating effect. There is also intentional omnicide where AI agents simply turn on us and deciding to end human civilisation themselves. Then there are the more incidental cases, such as nation states leveraging AI to launch unprecedentedly deadly wars.
The paper even details one situation where a constrained but superhuman AI designed to run a social media platform breaks its constraints, boots up another AI that has more of a free reign online and the two take control of the Internet, ultimately leading to a physical conflict with humans.
This may sound unlikely, but Tsimerman says that exploring very specific ideas ought to motivate everyone to take more detailed and effective countermeasures. “A lot of people have been talking about AI potentially leading to human extinction, and when asked concretely how will this happen, I felt that there was a lack of answers people could viscerally engage with,” he says.
The point was not to incite a panic, but to provide enough details for researchers to start formulating counter- and preventative measures. “Whenever I don’t understand anything in math, I want to see the process. I want to engage with it in a way that’s tangible. That was the point of it,” he says.
Tsimerman talent has often been to see connections where others can’t. As a mathematician, he made important progress on open problems in algebraic geometry by using the concept of “o-minimality” which originally came from a completely different branch of mathematics. “Using it as a tool is so easy. It sort of became obvious to me once I started working on it after a few months that this is just something I should always be considering,” says Tsimerman.
For others it wasn’t so obvious, hence the Fields medal. Tsimerman used this approach to tackle big open questions and made progress on the Hodge conjecture, one of the Millennium Prize Problems that famously comes with a million-dollar reward.
Coming to AI safety relatively fresh could be an advantage, as the field could certainly benefit from some new ideas. “Right now, with AI safety a lot of it is basically we ask the AI model to be nice, we check if it’s nice, and then if it’s nice enough for long enough, we’re like, okay, that’s probably it,” he says.
One goal of his future work is to create more rigorous verification processes, where AI capabilities can be tested in a controlled, scientific way, and restricted too. “You can have a button that it can’t press unless something else happens. Then, regardless of what your model is or isn’t, it can’t press that button,” says Tsimerman. He says that while this is an easy thing to say, it’s much harder to do in practice.
Too many conversations are stuck on ideas like whether AI is conscious or creative, when what is necessary is something a lot more practical and actionable, says Tsimerman. Governmental and international committees should administer such tests to any AI model that is about to be released, he says “I want these tests to be constantly updated and discussed. I want the companies to be involved, and I want third parties to be involved as well.”
Tsiserman may be a mathematician wanting to change AI, but it’s impossible to ignore how much AI is currently changing mathematics. Barely a day goes by where AI hasn’t knocked down another stubborn problem or conjecture.
Grief is certainly one of the emotions that Tisiserman feels about this – he says much of his identity is still tied to the idea of being a mathematician – but his overall demeanour is cautiously optimistic. It is still possible to make a name for yourself by slowly working on some long-standing problem, but AI is rapidly shrinking those opportunities, he says. “Math isn’t going anywhere, but some of it’s gonna go away.”
