Anthropic的一位数学家称,他利用该公司的Claude AI模型证明了黎曼猜想,这是数学界最著名的未解难题之一。据彭博社报道,Levent Alpöge周五在其个人网站上发布声明提出这一主张。
Mmathematician at Anthropic says he has used the company’s Claude AI model to prove the Riemann hypothesis, one of the most famous unsolved problems in mathematics. Levent Alpöge made the claim in a note on his website on Friday, Bloomberg reported.
“我相信我已在Claude的协助下,发现了黎曼猜想的证明,”Alpöge写道。
“I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” Alpöge wrote.
该猜想最早于1859年提出,涉及素数的分布规律。它是克莱数学研究所设立的七大“千禧年大奖难题”之一,每个难题奖金为100万美元。Alpöge表示,他的证明还涵盖了一个更广泛的版本——广义黎曼猜想。
The hypothesis, first proposed in 1859, concerns how prime numbers are spread out. It is one of the Clay Mathematics Institute’s seven Millennium Prize Problems, each of which carries a $1m prize. Alpöge says his proof also covers a broader version, the generalised Riemann hypothesis.
他并未发布传统论文,而是以Lean编程语言发布了证明,该语言可逐步验证每一个逻辑步骤。据《科学》杂志报道,代码约达1700万行。
He did not publish a traditional paper. Instead, he released the proof in Lean, a programming language that checks every logical step. The code runs to about 17 million lines, Science reported.
麻省理工学院高级研究科学家Andrew Sutherland告诉彭博社,他已在自己的电脑上检查了该代码。他表示,至少还有三位专业数学家进行了核查。周五晚些时候,Sutherland对《科学美国人》表示,证明出错的概率大致为零。
Andrew Sutherland, a senior research scientist at MIT, told Bloomberg he had checked the code on his own computer. He said he had seen at least three other checks by professional mathematicians. Later on Friday, Sutherland put the chance that anything is wrong at roughly zero, Scientific American reported.
不过,其他数学家呼吁谨慎。在Lean中编译通过的证明仍需陈述正确的问题,且必须由人类确认其确实如此。
However, other mathematicians urged caution. A proof that compiles in Lean still has to state the right problem, and humans have to confirm that it does.
伦敦帝国学院教授Kevin Buzzard表示,目前尚无人理解该证明究竟显示了什么,半岛电视台报道称。该证明尚未被正式接受。Alpöge称,Claude完成了大部分工作。“这一次,在冲刺阶段,Claude远远超过了我,”他写道。
Kevin Buzzard, a professor at Imperial College London, said nobody yet understands what the proof shows, Al Jazeera reported. The proof has not been formally accepted. Alpöge said Claude did much of the work. “This time Claude far outpaced me near the finish line,” he wrote.
Anthropic拒绝就此向彭博社置评。这一主张出现的前一天,该公司刚完成创纪录的股市首秀。
Anthropic declined to comment to Bloomberg. The claim came a day after the company’s record stock market debut.
若得到证实,这将成为AI可能协助解决的第二个“千禧年大奖难题”。9月,OpenAI称其模型已产出纳维-斯托克斯方程问题的解。
If confirmed, it would be the second Millennium Prize Problem that AI may have helped solve. In September, OpenAI said its models had produced a solution to the Navier-Stokes problem.