💡 TL;DR：AI 在数月内给出数百道困扰数学家几十年的难题的解法，他们一边用上它，一边担心「数学末日」：证明的信任、审稿的意义、数学家这个职业还剩什么。

‘Hostile takeover’: mathematicians reckon with an AI revolution

作者：Tom Whipple（科学记者、特派记者）、Rhys Blakely（科学编辑） 发表：《泰晤士报》（The Times），2026-10-09 18:01 BST 链接：https://www.thetimes.com/uk/science/article/mathematicians-ai-solving-problems-bv3nzhgcf 副题：Welcome to the ‘mathpocalypse’, when hundreds of solutions to problems that had stumped experts for decades seemed to be solved by artificial intelligence 备注：文中 Matt Zaremsky 关于「恐龙骨架 + TNT」的段落即中文网络流传的那段。以下为英文原文 + 中文全文翻译。

---

原文

When it happened, James Maynard, an Oxford University professor and a Fields medallist, was just about to go to bed.

Instead, sitting in front of his computer late on Tuesday night, he tried to make sense of what might well be the most significant day in the 10,000-year history of maths.

He felt “childish excitement”, he said, but also extreme concern. A single PDF had announced hundreds of solutions to problems that had stumped mathematicians for decades, some to problems he never dreamt he would see solved.

However, Maynard, who won the mathematical version of a Nobel prize in 2022, said he could not ignore that the source was not a human but AI. “Everything is moving so fast, I don’t know how the community will adapt.”

Hugo Duminil-Copin, a French Fields medallist, saw the same update from OpenAI and was a little less equivocal. He compared it to a truck driving through his work. He said: “Dozens of papers deal with topics I was working on. Between results that beat you to the finish line and thousand-page proofs, I don’t even know where to look any more.

“The shock is immense. I am paralysed. Tomorrow, we will find a way forward … but for today I don’t have the energy.”

His reaction is not unusual. Words such as “shock wave”, “cataclysm” and “mathpocalypse” have been used among mathematicians. But, also, a little more rarely, “thrilling” and “stupendous”.

Late on Tuesday night, OpenAI put more than 700 papers on the internet. Some constituted incremental advances. Others appeared to be breakthroughs in problems that had stumped humans for decades. Many were poorly written, rambling and incomprehensible — at first — to humans. But all told the same story: mathematics will never be the same again.

Mathematicians had anticipated this day would come. Now they were trying to understand what it meant.

Henry Wilton of Cambridge University said the problems OpenAI had tackled were far more than a collection of unsolved puzzles. Some, such as the quasi-Riemann hypothesis, which relates to the distribution of prime numbers, may mark significant steps towards solving some of the biggest problems in mathematics. Others have served as signposts, guiding researchers towards promising lines of inquiry.

Built up over decades, these problems have helped mathematicians to decide which questions were worth pursuing and organise research around shared aims. “They provide structure, a yardstick for progress and long-term goals,” he said.

Solving them all at once, he argued, would strip away much of the intellectual framework that gives the field its direction.

“If OpenAI wanted to destroy the mathematical community, this would be a great way to go about it,” Wilton said.

“The advent of AI was always going to be a seismic shock, but this huge dump of papers is a tsunami that we have no time to prepare for.” That was particularly true because many of the documents were confusingly written. One mathematician likened it to “reading the ramblings of a crackpot.”

Ana Caraiani, from Imperial College, called the OpenAI release “more a demonstration of power than of scholarship”. A problem she had made significant progress on, then left when she felt it needed “one more trick”, had been solved. It had indeed needed one more trick. That, she said, was exciting. But, not the way it had been revealed.

She said: “It feels like a hostile takeover of the community … It’s much bigger than mathematics. Anyone who does any intellectual work will face this.”

Thomas Fink, director of the London Institute for Mathematical Sciences, said worries were understandable, but they were also clouding the fact that this was a great day for mathematics.

He added: “We feel like we’re on crack cocaine. This is an exciting time; we are glued to the calculations, making great progress. So many things will fall.” He thinks it is plausible that most open problems in maths — that are solvable — could be solved in a couple of years. What then for mathematicians? He said they would still have a vital role, but that there might not be so many who can fulfil it.

He said: “All these conjectures, a hundred years of smart people saying, ‘This is what I think the next step is, go in this direction’? AI has done that. But who is going to say what are the next directions? Choosing which problems to focus on is critical. Someone has to find the next layer. And that’s humans.

“There will be people saying life sucks, why bother? But it’s not as dark as it seems. Look for something more visionary. The people who will survive need to be visionaries.” And those who don’t? “There’s too many scientists. Too many researchers.”

Mathematicians are still trying to digest the papers and to understand the decades of apparent progress that arrived overnight. In some, they have found flaws.

In surveying those that have been tackled, they have also found notable absences. There are open problems in cryptography of potentially comparable difficulty that are absent from the list, but which could have huge ramifications for internet security if solved.

Sophie Maclean, a final year PhD student at King’s College London, said that many junior colleagues were simply “taking a breather, going outside, and not thinking about maths”.

She said: “I personally am quite down. Maths will still survive, and mathematicians will still survive in some form, but no one knows what form yet. The human bits, the stories, the struggle — I worry those are the bits that are going to suffer.

“The thing I find saddest is it’s taking the bits that I love about maths away. The collaborating with people, the finding the limits of human achievement. The solving puzzles.”

For some, it is more immediately personal. On the other side of the Atlantic, Matt Zaremsky learnt by text that years of work had been scooped, after putting his four-year-old twins to bed. A colleague had messaged to say a problem Zaremsky had been working on in group theory had been “killed by AI”. How did that feel?

“Well, it feels like we’ve been working on an archeological dig for four years, gradually discovering more and more of a really cool-looking dinosaur skeleton, and then a trillion-dollar company showed up and just blasted the whole thing with TNT, handed us the whole skeleton, and walked away. So, I guess I’d say, ‘not great’.”

---

中文全文

“恶意收购”：数学家们面对一场 AI 革命 副题：欢迎来到“数学末日”——数百个困扰专家数十年的问题，似乎被人工智能解决了

事情发生的时候，牛津大学教授、菲尔兹奖得主詹姆斯·梅纳德正要上床睡觉。

结果，周二深夜，他坐在电脑前，试图弄明白这可能是数学一万年历史上最重要的一天。

他说，他感到“孩子气的兴奋”，但也极度担忧。一份 PDF 宣布了数百个困扰数学家数十年的问题的解答，其中有些问题他从未想过能看到被解决。

然而，2022 年拿下“数学界诺贝尔奖”的梅纳德说，他无法忽视一件事：答案的来源不是人，是 AI。“一切都太快了，我不知道这个共同体要怎么适应。”

法国菲尔兹奖得主雨果·迪米尼尔-科潘看到了 OpenAI 的同一份更新，态度没那么含糊。他把它比作一辆卡车碾过自己的工作。他说：“几十篇论文都在处理我正在做的题目。一边是别人先你一步撞线，一边是上千页的证明，我都不知道该看哪儿了。”

“冲击是巨大的。我瘫住了。明天，我们会找到出路……但今天，我没有力气。”

他的反应并不罕见。“冲击波”“灾变”“数学末日”这样的词在数学家里流传。但也有一些更少见的词：“刺激”“惊人”。

周二深夜，OpenAI 把 700 多篇论文放上了网。有些是渐进式的推进。另一些看起来是困扰人类数十年的问题的突破。很多写得糟糕、散乱，一开始根本看不懂。但它们都在讲同一件事：数学再也不会和以前一样了。

数学家早就预料到这一天会来。现在他们在弄明白它意味着什么。

剑桥大学的亨利·威尔顿说，OpenAI 处理的那些问题，远不只是“一堆没解出来的谜题”。有些问题——比如与素数分布有关的“准黎曼假设”——可能是通向数学中几个最大问题的重要一步。另一些问题则像路标，把研究者引向有希望的方向。

这些问题积累了几十年，帮数学家判断哪些问题值得追，也把研究组织在共同的目标周围。“它们提供结构、衡量进步的尺子，还有长期目标。”他说。

他认为，一次性把它们全部解掉，会剥掉这个领域赖以确定方向的很大一部分智识框架。

“如果 OpenAI 想摧毁数学共同体，这是一个很好的办法。”威尔顿说。

“AI 的到来注定是一场地震，但这次论文的海量倾倒，是一场我们来不及准备的海啸。”这一点尤其因为很多文档写得令人费解而加剧。一位数学家把它比作“读一个疯子的呓语”。

帝国理工学院的安娜·卡拉伊亚尼说，OpenAI 这次发布“更像是在展示权力，而不是在展示学术”。她曾在一个问题上取得重大进展，后来觉得“还差一个技巧”，就放下了——那个问题被解决了。它确实只差一个技巧。她说，这很令人兴奋。但不是以它被公开的方式。

她说：“感觉像是对这个共同体的恶意收购……这件事远不止数学。任何做智识工作的人都会面对它。”

伦敦数学科学研究所所长托马斯·芬克说，担忧可以理解，但这些担忧也遮住了一个事实：这对数学来说是伟大的一天。

他补充说：“我们感觉自己像在吸可卡因。这是个激动人心的时刻；我们盯着计算，进展巨大。很多东西都会倒下来。”他认为，数学中大多数可解的开问题，有可能在几年内被解决。那数学家呢？他说他们仍然会有至关重要的角色，但能胜任的人可能不会那么多。

他说：“所有这些猜想，一百年来聪明人一直在说，‘我认为下一步是这个，往这个方向走’？AI 已经做完了。但谁来指出下一个方向？选择聚焦哪些问题，是关键的。总得有人去找下一层。而那是人的事。”

“会有人说，生活真烂，何必呢？但事情没有看上去那么黑。去找更有远见的东西。能活下来的人，必须是有远见的人。”那没有远见的人呢？“科学家太多了。研究者太多了。”

数学家还在消化这些论文，试图理解一夜之间到来的、看起来是几十年的进展。在一些论文里，他们找到了漏洞。

在梳理已经被处理的问题时，他们也发现了明显的缺席。密码学里有一些难度可能相当的开问题不在名单上——而这些问题一旦被解决，对互联网安全的影响可能是巨大的。

伦敦国王学院最后一年博士生索菲·麦克林说，很多年轻同事干脆“歇一口气，出门走走，不去想数学”。

她说：“我个人挺低落的。数学还会活下来，数学家也会以某种形式活下来，但没人知道是什么形式。人的部分，那些故事，那些挣扎——我担心受损的是这些。”

“我觉得最难过的是，它把我爱数学的那些部分拿走了。和人合作，触碰人类成就的边界。解谜。”

对一些人来说，这件事更直接地扎在个人身上。大西洋另一边，马特·扎雷姆斯基把四岁的双胞胎哄睡之后，从短信里得知自己多年的工作被人抢先了。一位同事发消息说，扎雷姆斯基在群论里做的一个问题已经被“AI 干掉了”。那是什么感觉？

“这么说吧，感觉就像我们做了四年的考古挖掘，一点一点挖出一副特别酷的恐龙骨架，然后一家市值万亿美元的公司出现了，直接用 TNT 把整个现场炸平，把整副骨架递给我们，然后走了。所以，我想我会说：‘不太好。’”