"我认为世界正在改变。我们现在所熟知的数学职业……我不认为它还会以现在这个样子存在。"
说这句话的时候，38岁的Jacob Tsimerman刚刚接过菲尔兹奖奖章。7月23日，2026年国际数学家大会在美国费城开幕，Tsimerman成为加拿大机构首位菲尔兹奖得主——也是当天最受瞩目的获奖者之一。记者会上有人问他未来打算，这位新科菲尔兹奖得主没有任何铺垫，直接摊牌：很快入职OpenAI，方向是AI安全。
没有过渡，没有"考虑一下"，领奖当天直接宣布加盟一家AI公司。这在十年前是不可想象的——菲尔兹奖得主的归宿通常是普林斯顿高等研究院、MIT数学系这种象牙塔顶端。但在2026年，连数学界最高荣誉的获得者都选择了直接跳进AI浪潮。
他是谁？为什么重要？
Tsimerman今年38岁，是多伦多大学数学系教授，因在o-minimality（o-极小性）和数论交叉领域的突破性工作获奖。他16岁上大学，24岁拿到博士学位，师从著名数学家Peter Sarnak，是当代数论和模型论领域最顶尖的学者之一。
更有意思的是，他的学生刚用Anthropic的Fable 5模型推翻了一个悬置87年的数学猜想。量子位在报道中提到了这一细节——这不是巧合。Tsimerman本人早已深度关注AI在数学研究中的应用，甚至可能已经在日常研究中使用大模型辅助证明。他对AI的理解，不是外行的好奇，而是一线研究者的亲身体验。
菲尔兹奖每四年颁发一次，授予40岁以下最杰出的数学家，被称为"数学界的诺贝尔奖"。历史上，菲尔兹奖得主几乎全部留在学术界终老。上一个有类似"跨界"举动的是Edward Witten——但他是从物理转向数学，还是在象牙塔内部。Tsimerman直接去了一家商业AI公司，而且是去做AI安全，不是去做AI for Science——这背后的信号比他个人的选择更重要。
顶级数学家为什么流向AI安全？
你可能会问：一个做o-极小性和数论的数学家，去AI公司能做什么？
答案是：AI安全恰恰需要最顶尖的数学头脑。当大模型越来越强、Agent越来越自主，一个核心问题变得越来越紧迫：我们如何从数学上证明一个AI系统是安全的？如何保证它不会在长时间运行中偏离目标？如何验证它的决策符合人类意图？这些问题本质上都是数学问题——涉及形式化验证、模型论、可解释性、博弈论、复杂系统等领域。
"当最聪明的数学头脑都开始认真对待AI安全，这说明这个问题的难度和重要性已经被严重低估了。"—— 一位AI安全研究员的评论
OpenAI不是第一次招数学家。过去两年，OpenAI已经从顶尖大学挖走了数十名数学和物理博士，超级对齐（Superalignment）团队虽然解散了，但AI安全研究反而在加速。Tsimerman的加入标志着一个新高度：连菲尔兹奖得主都来了。
这背后有两个趋势值得关注。第一，AI公司能提供的研究资源已经远超任何大学。在OpenAI，Tsimerman可以访问世界上最强大的计算集群、和最顶尖的AI研究者合作、看到最前沿的模型能力——这些是任何大学数学系都给不了的。第二，AI安全正在从"哲学讨论"变成"工程问题"，需要真正的数学工具来解决，而不只是写论文讨论伦理。
王虹、邓煜两位中国籍数学家也在本届菲尔兹奖获奖名单中——这是历史上首次有中国籍数学家获得菲尔兹奖。他们的选择同样值得关注。Tsimerman去了OpenAI，其他顶级数学家会去哪？Google DeepMind？Anthropic？还是继续留在象牙塔里？无论答案如何，有一点是确定的：AI正在虹吸整个科学界最聪明的头脑，数学只是开始。
明天见。
"I think the world is changing. The profession of mathematics as we know it… I don't think it will exist in its current form."
Thirty-eight-year-old Jacob Tsimerman had just received his Fields Medal when he said those words. On July 23, at the opening of the 2026 International Congress of Mathematicians in Philadelphia, Tsimerman became the first Fields Medalist from a Canadian institution — and one of the most closely watched laureates of the year. When a reporter asked about his plans at the press conference, the newly minted medalist didn't hedge. He came right out with it: he would be joining OpenAI soon, to work on AI safety.
No transition period, no "I'll think about it" — announcing a move to an AI company on the very day he received the highest honor in mathematics. A decade ago, this would have been unthinkable. Fields Medalists typically ended up at the Institute for Advanced Study in Princeton, MIT Math, or similar ivory-tower pinnacles. But in 2026, even winners of math's highest prize are jumping straight into the AI wave.
Who Is He, and Why Does It Matter?
Tsimerman, 38, is a professor at the University of Toronto, awarded the Fields for breakthrough work at the intersection of o-minimality and number theory. He started university at 16, earned his PhD at 24 under renowned mathematician Peter Sarnak, and is one of the leading scholars in contemporary number theory and model theory.
More interestingly: his student just used Anthropic's Fable 5 model to resolve an 87-year-old unsolved conjecture. QbitAI mentioned this detail in its coverage — and it's no coincidence. Tsimerman has been deeply engaged with AI applications in mathematical research for some time, possibly already using LLMs to assist with proofs in his daily work. His understanding of AI isn't amateur curiosity — it's hands-on experience from a frontline researcher.
The Fields Medal is awarded every four years to the most outstanding mathematicians under 40, often called the "Nobel Prize of mathematics." Historically, Fields Medalists have almost universally remained in academia for their entire careers. The last comparable "cross-over" was Edward Witten — but he moved from physics to mathematics, staying within the ivory tower. Tsimerman is going directly to a commercial AI company, and specifically to work on AI safety — not AI for science. The signal here matters more than the individual choice.
Why Are Top Mathematicians Flowing to AI Safety?
You might wonder: what can a mathematician who works on o-minimality and number theory actually do at an AI company?
The answer: AI safety is precisely the kind of problem that demands the sharpest mathematical minds. As LLMs grow more capable and agents become more autonomous, a central question grows urgent: how do we mathematically prove that an AI system is safe? How do we guarantee it won't drift from its objectives over long horizons? How do we verify its decisions align with human intent? These are fundamentally mathematical problems — spanning formal verification, model theory, interpretability, game theory, and complex systems.
"When the sharpest mathematical minds start taking AI safety seriously, it means the problem's difficulty and importance have been drastically underestimated."— An AI safety researcher's observation
This isn't the first time OpenAI has hired mathematicians. Over the past two years, OpenAI has poached dozens of math and physics PhDs from top universities; the Superalignment team may have been dissolved, but AI safety research has, if anything, accelerated. Tsimerman's arrival marks a new threshold: even a Fields Medalist is coming aboard.
Two trends underlie this. First, the research resources AI companies can offer now vastly outstrip any university's. At OpenAI, Tsimerman will have access to the world's most powerful compute clusters, collaboration with leading AI researchers, and exposure to the most advanced model capabilities — none of which any university math department can match. Second, AI safety is shifting from "philosophical debate" to "engineering problem," requiring genuine mathematical tools rather than ethical discussion papers.
Two Chinese mathematicians, Wang Hong and Deng Yu, also received Fields Medals this year — the first time Chinese nationals have won the prize. Their choices will be worth watching too. Tsimerman went to OpenAI. Where will other top mathematicians go? Google DeepMind? Anthropic? Or stay in the ivory tower? Whatever the answer, one thing is certain: AI is siphoning the brightest minds across the entire scientific world. Mathematics is just the beginning.
See you tomorrow.
我们现在所熟知的数学职业……我不认为它还会以现在这个样子存在。
—— Jacob Tsimerman，2026菲尔兹奖得主
The profession of mathematics as we know it… I don't think it will exist in its current form.
— Jacob Tsimerman, 2026 Fields Medalist
菲尔兹奖 · Jacob Tsimerman · OpenAI · AI安全 · 数学家流动 · o-minimality · 数论 · 顶级人才 · 多伦多大学
Fields Medal · Jacob Tsimerman · OpenAI · AI safety · mathematician talent flow · o-minimality · number theory · top talent · University of Toronto
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 9 个源信号生成，经编辑部人工审核。素材来源：量子位、IT之家、国际数学联盟。
This article was generated by the Dawn Vision cognitive engine processing 9 source signals, with human editorial review. Sources: QbitAI, IT Home, International Mathematical Union.
