Sam Altman 同情数学家处境，但表示 OpenAI 的数学研究攻势绝不会停止

在连续数月用源源不断的 AI 生成成果冲击数学界之后，OpenAI 首席执行官 Sam Altman 表示，他同情那些正努力理解这一切的研究人员。但他丝毫没有停下来的打算——并暗示该公司的科研储备库中可能还会涌现出更多发现。

“归根结底，我认为世界理应理解宇宙是如何运作的，”Altman 在周四发布的播客《Cheeky Pint》一期节目中说道，“但我确实对这种研究方式抱有同理心。”

这标志着这位 OpenAI 掌舵人首次直接公开承认该公司对数学家带来的冲击。此前，OpenAI 推出了一系列真正令人印象深刻的 AI 生成结果，但其混乱且沟通欠佳的发布方式已经让研究人员感到困惑迷茫、疲于应付——即使学界普遍对 AI 作为科学发现工具的潜力感到兴奋。

Altman 辩称，公众获取新科学知识的权利“超过”了 OpenAI 发布这些成果所带来的混乱，但他承认公司需要改进传达研究成果的方式。“在这一点上我的态度非常明确：人类理应获得基础知识，我们不应该把一项重大的基础科学发现捂在手里，”他说，“这是全世界应得的。”

接受《The Verge》采访的数学家曾屡次批评 OpenAI 公布成果的方式，指出其署名不充分、缺乏上下文背景且缺少充分验证。研究人员表示，这导致他们在 OpenAI 不断抛出新成果的同时，不得不承担繁复吃力的核验与解读工作。单在过去一个月里，OpenAI 就发布了数百项数学成果，但往往仅附带简略的技术说明，被研究人员评价为文笔粗糙且难以理解。

然而，OpenAI 现在却正因为其做出了恰恰与 Altman 所言相反的行为而面临批评：将具有重大意义的潜在发现秘而不宣。在本周接受《The Verge》采访时，多位数学家与科学家表达了对OpenAI 学术把关行为的沮丧。此前该公司宣布在数学、物理和计算机科学领域取得三项重大突破，但其中一项涉及现代粒子物理学数学基础——杨-米尔斯理论（Yang-Mills theory）的成果细节却未公开发布。OpenAI 称这是根据其专门挑选的一个专家顾问团的建议作出的决定，顾问团就如何最妥善地发布该成果提供了意见。这一做法与该公司此前咨询过的另一组数学家的建议背道而驰，但也构成了该公司声称的成果沟通新模式的一部分。

在播客中，Altman 还暗示 OpenAI 手头未公开的数学发现远不止已发布的这些。当被问及公司除已公布的数百个证明之外是否还有更多储备时，Altman 回答道：“确实有。那是当然的。”

Altman 同时坦承，OpenAI 对数学问题的攻关在短期内确实“失去了控制”，但他表示这是刻意为之，也是为了探测其 AI 模型的极限所必需的。他透露，公司内部曾设立过一个名为“Model Name-Hold My Beer”的 Slack 频道，员工们“直接把模型丢去处理他们能找到的最难的问题”，以观察模型的表现。

Altman 还估计，OpenAI 在攻克数学难题上已投入了数亿美元资金，其中包括在其纳维-斯托克斯（Navier-Stokes）方程解法上投入的约 1000 万美元——“甚至可能更多”，该问题是数学界赫赫有名的千禧年大奖难题之一。这笔巨额开支展现了前沿 AI 实验室在解决科学难题上所调动的庞大资源，同时也引发了研究人员的担忧：他们可能会因成本过高而被挤出自己的研究领域。

看来，期盼能喘口气的数学家和其他科研人员，短期内恐怕无法如愿了。

After months of bombarding mathematicians with a relentless stream of AI-generated results, OpenAI CEO Sam Altman says he sympathizes with researchers struggling to make sense of it all. But he has no intention of stopping — and hints there could be much more to come from the company’s stockpile of discoveries.

“I think, on net, the world deserves to understand how the universe works,” Altman said during an episode of the Cheeky Pint podcast released Thursday. “But I do have empathy for the method.”

His comments mark one of the first times the OpenAI chief has directly acknowledged the company’s impact on mathematicians, following a string of genuinely impressive AI-generated results whose chaotic, poorly communicated rollouts have left researchers bewildered and struggling to keep up, even amid widespread excitement about AI’s potential as a tool for discovery.

Altman argued that the public’s right to new scientific knowledge “outweighs” the disruption caused by OpenAI’s releases, though he acknowledged the company needs to improve how it communicates its findings. “I feel unequivocal on this point: Humanity deserves fundamental knowledge, and we should not be sitting on a major fundamental science discovery,” he said. “The world deserves that.”

Mathematicians speaking to The Verge have repeatedly criticized OpenAI for how it announces its results, citing inadequate attribution, little context, and insufficient verification. Researchers say this has left them doing the painstaking work of checking and making sense of their findings even as the company presses ahead with more. This month alone, OpenAI has released hundreds of mathematical results, often with little more than brief writeups that researchers have described as poorly written and hard to follow.

Yet OpenAI is now facing criticism for doing precisely what Altman says it should not be: sitting on potentially significant findings. Mathematicians and scientists speaking to The Verge this week said they are frustrated by what they see as OpenAI’s gatekeeping of research, following the announcement of three major breakthroughs in mathematics, physics, and computer science. Details for one of the claimed results — involving Yang-Mills theory, the mathematical foundations of modern particle physics — have not been released publicly. OpenAI says it is due to the recommendations of a hand-picked group of experts it consulted on how best to release it. The practice runs directly counter to the advice of a previous group of mathematicians the company consulted, but forms part of what the company says is a new approach to communicating its findings.

On the podcast, Altman hints that OpenAI is sitting on many more mathematical discoveries than it has made public, too. Asked whether the company has more than the hundreds of proofs they have already released, Altman replied: “We do. Of course.”

Altman also acknowledged that OpenAI’s pursuit of mathematical problems did “get out of control” for a short period of time, but said that was deliberate and necessary to probe the limits of its AI models. He described an internal Slack channel called “Model Name-Hold My Beer,” where employees “were just throwing that at the hardest problems they could” to see how the models would perform.

Altman also estimated that OpenAI has spent in the order of hundreds of millions of dollars tackling mathematical problems, including around $10 million — “maybe a little more” — on its Navier-Stokes solution, one of mathematics’s legendary Millennium Prize problems. The spending offers a glimpse of the vast resources AI labs are bringing to bear on scientific problems, raising concerns among researchers that they may be priced out of their own fields.

It looks like mathematicians and other researchers hoping for a breather may be out of luck.