Jacob Tsimerman, 38, says AI will soon outperform human mathematicians and could accelerate their field 100 times, and he is leaving for OpenAI to work on safety.
Key Points:
- Tsimerman expects AI systems to surpass humans at research mathematics, forcing the discipline to rethink how it operates.
- He is taking leave from the University of Toronto but keeping his faculty post.
- His work at OpenAI will target safety rather than model capabilities.
Jacob Tsimerman Explains OpenAI Move
In an interview published Friday, the University of Toronto number theorist described a shift that has taken hold over the past few months. Contest problems written for high school students fell first, he said, and research-level results followed within a matter of months. Models that once stumbled on elementary questions now turn out proofs that professional mathematicians would be proud to achieve and publish.
If that trend holds, he expects AI to become robustly better than humans at the work the profession does today. At that point, he says, mathematicians will have to rethink how the entire field works.
He first announced the move on Jul. 23, hours after the International Mathematical Union named him one of four winners of this year's Fields Medal for his work on the André-Oort conjecture. The award, given every four years to mathematicians under 40, also went to Hong Wang, Yu Deng and John Pardon. Tsimerman is taking leave from the Toronto department but keeps his faculty post, and he says he has no firm plan past the next year.
Also Read: Bitcoin ETFs Absorb $233.1M As A Single Fund Supplies Most Of It
Terence Tao Warns Of Math Crisis
Terence Tao told the same congress in Philadelphia last weekend that mathematics is entering a turbulent period, which he described as a crisis in the discipline's foundations and its working values.
He pointed to a May experiment in which frontier models solved seven of 10 unpublished research problems, with the proofs refereed by experts for correctness and judged fit to print. Tsimerman sees a real upside in that shift. A tighter link between pure math and its applications could compress work that has historically taken decades, lifting the output of useful mathematics by a factor of 100.
Tsimerman Backs AI Safety Push
His own assignment will be safety, not capability. Capabilities are advancing well enough without him, he argues, while the safety side has far more unfinished work and far fewer people on it right now.
That emphasis is not new. Last year he co-wrote a report with Andrew Critch that sorted AI-driven human extinction into five categories, organized by who, if anyone, would bear responsibility.
This week he also welcomed an open letter that more than 1,000 employees at OpenAI, Anthropic, Google and Meta signed, among them Anthropic chief executive Dario Amodei, asking Washington to back an international effort on pacing frontier AI development.
Read Next: Polymarket Traders Give Spider-Man A 91% Shot At A Historic Debut