字号·· | 护眼
南华早报

AI解决了黎曼猜想?这一说法在中国举行的200周年诞辰活动上引发轰动Has AI solved the Riemann hypothesis? Claim makes waves at 200th birthday event in China

点「原文对照」整页切到原文,或双击某段只看那段的原文。

美国一位数学家声称利用人工智能(AI)解决了近170年前的黎曼猜想,引发学界震动。

Chinese Fields medallist hails ‘new era’ after mathematician claims that AI solved 167-year-old problem from Bernhard Riemann.

该消息于周五在中国南部深圳的一场大型会议上公布。会议汇聚了世界顶尖数学家,持续至周一,以纪念德国先驱伯恩哈德·黎曼诞辰200周年——黎曼是该领域最著名的未解难题之一的提出者。

这一声明在数学界掀起冲击波,一位与会学者表示他有98%的把握认为该证明是正确的。

Shing-Tung Yau, director of the Yau Mathematical Sciences Centre at Tsinghua University and chair of the bicentennial conference’s scientific committee, says he learned of the feat at the event. Photo: Nanjing Centre for Applied Mathematics A US-based mathematician has stunned the community by claiming to have solved the nearly 170-year-old Riemann hypothesis using artificial intelligence (AI).

菲尔兹奖得主、清华大学丘成桐数学科学中心主任、此次两百周年会议科学委员会主席丘成桐在接受《北京新闻》采访时称,他正参加会议时得知了这一成果。

The news broke on Friday as a major conference opened in southern China’s Shenzhen, where world-leading mathematicians have gathered until Monday to mark the 200th anniversary of the birth of Bernhard Riemann – the German pioneer behind one of the field’s greatest unsolved problems.

他表示,这一宣布表明数学“已经进入了一个新纪元”,该论文于周日报道。

The claim sent shock waves through the mathematics community, with one attending scholar saying that he was 98 per cent certain the proof was correct.

丘成桐说,会议上有多位学者研究黎曼猜想,其中包括演讲者、同为菲尔兹奖得主的詹姆斯·梅纳德。

牛津大学教授梅纳德在该假设研究方面做出过重要贡献,他在宣布结果后不久接受《纽约时报》采访时表示,他认为这份 AI 证明很可能是正确的。据引述,他说:“我有 98% 或 99% 的把握。”

Fields medallist Shing-Tung Yau, director of the Yau Mathematical Sciences Centre at Tsinghua University and chair of the conference’s scientific committee, told The Beijing News that he was at the event when he learned of the feat. He said the announcement indicated that mathematics “ has entered a new era, ” the paper reported on Sunday.

这份证明由土耳其裔美国数学家、Anthropic 公司研究员莱文特·阿尔波盖于周五公布。他在个人网站上表示,利用该美国公司的 AI 模型 Claude 解决了这一问题。

Yau said several scholars at the conference had studied the Riemann hypothesis, including speaker and fellow Fields medallist James Maynard.

独立数学家随即开始对 Claude 在六周内生成的基于 Lean 的代码进行核查,目前已有部分数学家确认其数学准确性。

Maynard, an Oxford University professor who has made important contributions to the hypothesis, told the New York Times soon after the announcement that he believed the AI proof was likely to be correct.

尽管同行给出了早期认可,但将解决方案的底层逻辑转化为人类可读的论文、并获得正式认可,仍可能需要一段时间。

“I’d be 98 or 99 per cent certain,” he was quoted as saying. The proof was unveiled on Friday by Levent Alpoge, a Turkish-American mathematician at Anthropic.

伯恩哈德·黎曼(1826-1866)于 1859 年提出该假设。图片:手稿黎曼生于 1826 年 9 月,于 1859 年首次提出该假设。这一问题在长达 167 年的时间里未被正式证明,被广泛视为数学界最重要的未解难题之一。

He said on his personal website that he had solved the problem using the American company’s AI model, Claude. Independent mathematicians immediately began running checks on the Lean-based code – generated by Claude over a period of six weeks – with some now confirming its mathematical accuracy. Despite the early nod from peers, translating the solution’s underlying logic into a human-readable paper – and securing formal recognition – could take some time.

该问题与素数分布相关,构成现代数论的基石。即使没有正式证明,数学家们长期以来一直假定该假设为真,并以此为基础推导出无数其他定理。

Bernhard Riemann (1826-1866) introduced his hypothesis in 1859. Photo: Handout Riemann, born in September 1826, first proposed the hypothesis in 1859. Formally unproven for 167 years, it is widely regarded as one of the most important unsolved problems in mathematics.

阿尔波盖并未发布完整的可读手稿,而是发布了约 1700 万行用 Lean 编写的代码。Lean 是一种开源编程语言和交互式定理证明器。

The problem is tied to the distribution of prime numbers and forms a cornerstone of modern number theory. Even without formal proof, mathematicians have long assumed the hypothesis to be true, relying on it to derive countless other theorems.

阿尔波盖在周五随代码发布的说明中写道:“试图发布某种大规模、不可读、糟粕般的‘垃圾’显然是不合适的,因此我决定仅展示 Lean 代码。”

Alpoge did not publish a full readable manuscript, but rather released around 17 million lines of code written in Lean, an open-source programming language and interactive theorem prover.

菲尔兹奖得主、加州大学洛杉矶分校数学家陶哲轩 于当天在博客文章中作出回应。

“It being obviously inappropriate to attempt to post some massive unreadable travesty of ‘slop’, I decided to only present Lean,” Alpoge said in a note posted on Friday alongside the code.

“我想消化这个过程需要一些时间,不过至少在这个案例中,我们有一位人类作者可以沟通,以协助完成这项任务,”陶哲轩写道。

Terence Tao , Fields medallist and mathematician at the University of California, Los Angeles, responded in a blog post the same day.

黎曼猜想是克莱数学研究所2000年提出的七个千禧年大奖难题之一,每个题目的解决者可获得100万美元奖金。目前只有一个题目已被正式解决。

“I imagine the digestion process will take some time, though in this case at least we have a human author to communicate with to assist with this task,” Tao wrote. The Riemann hypothesis is one of seven Millennium Prize Problems proposed in 2000 by the Clay Mathematics Institute, with each solution carrying a US$1 million prize. Only one has been officially resolved so far.

上个月,另一个AI里程碑在数学界引起轰动,当时OpenAI发布了数百份手稿,涵盖广泛问题——其中也包括千禧年难题列表上的其他题目。

Last month, another AI milestone took the mathematics community by storm when OpenAI published hundreds of manuscripts covering a wide range of problems – including others also on the Millennium list.

据丘成桐说,AI的使用可能会使数学研究转向生物学、物理学和化学等领域常用的模式,即先找到结果或答案,再仔细加以审视,而不是采用数学家传统使用的逐步推导方法。

According to Yau, the use of AI could shift mathematics towards a mode of research more often used in fields like biology, physics and chemistry, where the result or answer is found first and carefully examined afterwards, rather than the step-by-step approach traditionally used by mathematicians.

“至于人类智能,我认为机器仍无法相比,”丘成桐对《新京报》表示,并补充说机器的解题能力仍强于其提出问题的能力。

“As for human intelligence, I think machines still cannot compare,” Yau told The Beijing News, while adding that the problem-solving ability of machines was still stronger than their ability to pose questions.

勒文特·阿尔波格发布了约1700万行用Lean编写的代码,以支持他的主张。图片:受访者提供阿尔波格在周五的说明中表达了相同的观点。

Levent Alpoge has released around 17 million lines of code written in Lean in support of his claim. Photo: Handout Alpoge echoed the sentiment in his Friday note.

“这个时代让我们处于一种相当非凡的境地,即我在掌握其数学周围的大量领域之前,就向世界呈现了一个完整的论证,”他说。

“This puts us in the rather extraordinary situation this age has made possible, in which I am presenting the world with a complete argument before having mastered a massive neighbourhood around its mathematics,” he said.

时机引发了猜测,因为黎曼猜想证明发布于Anthropic在纳斯达克首次公开募股(IPO)之后不久。

The timing has triggered speculation, given that the Riemann proof was released right after Anthropic’s initial public offering on the Nasdaq.

阿尔波格在说明中回应了这一问题,称这种时机“很糟糕”,但坚持认为如果可能他会更早完成。

Alpoge addressed the issue in his note, describing the timing as “horrible” but maintaining that he would have finished earlier if he could have.

“存在一种显而易见的愤世嫉俗的看法,与时机有关,但这是错误的,而且我很高兴我处于一种让这种看法显而易见是错误的境地。”

“There is an obvious cynical view, to do with the timing, but it is false, and I am pleased that I am in circumstances where it is obviously false.”Sections: News / China / Science | Topics: Science, United States, Turkey, Germany, Shing-Tung Yau, Anthropic

栏目:新闻 / 中国 / 科学 | 话题:科学、美国、土耳其、德国、丘成桐、Anthropic短链接: 在《南华早报》阅读原文

阅读媒体原文 ↗