ARTICLE · 1062018

OpenAI新模型24天攻破100+世界级数学难题:该欢呼还是保持审慎?

发布时间：2026-09-23 10:21:38 最近访问：2026-09-23 10:21:28

近期OpenAI对外公布一则备受行业关注的消息：一款8月28日启动训练的内部原型模型，在短短24天时间内，完成了100多道长期悬而未决的数学开放问题的推导工作，覆盖数学多个细分领域，其中包含知名的纳维‑斯托克斯千禧年大奖难题相关课题。这一推进速度，就连OpenAI内部负责数学方向的研究人员也表示超出预期。

OpenAI recently released an industry‑watched announcement. An internal prototype model, whose training started on August 28, produced derivations for more than 100 long‑standing open mathematical problems across multiple sub‑fields of mathematics within just 24 days, including research related to the well‑known Navier‑Stokes Millennium Prize Problem. Even OpenAI’s in‑house mathematics researchers noted that this pace of progress exceeded their expectations.

很多普通读者会好奇，千禧年难题究竟意味着什么。简单来说，这是学界筛选出的七大高难度数学问题，每一道都困扰全球数学家数十年，解开任意一题都将获得百万美元奖励，迄今为止仅有一道难题被人类完整证明。纳维‑斯托克斯方程主要用来描述流体运动，小到水杯里的水流，大到大气与洋流，都受这套方程支配。

Many general readers may wonder what Millennium Prize Problems stand for. Simply put, they are seven extremely difficult mathematical challenges selected by academia. Each has puzzled mathematicians worldwide for decades, and solving any one of them carries a one‑million‑dollar reward. So far, only one has been fully proven by human researchers. The Navier‑Stokes equations describe fluid motion, governing phenomena ranging from water flow in a cup to atmospheric patterns and ocean currents.

面对AI快速产出大量数学推导结果，OpenAI同步宣布组建独立的数学与人工智能顾问组，驻地设在普林斯顿高等研究院，集合9位全球顶尖数学家，小组独立运营，不从企业领取酬劳。该小组主要承担成果评估、学术标准校验、向学界传递意见等工作，用来平衡AI快速产出带来的各类问题。有相关研究显示，近年多家AI机构都陆续在数学猜想、理论推导上取得阶段性进展，同时也不断收到来自数学界的顾虑声音，部分学者担忧把破解难题当作AI性能跑分，会打乱基础科研的固有节奏。

In response to the large volume of mathematical outputs generated rapidly by AI, OpenAI also announced an independent Advisory Group on Mathematics and Artificial Intelligence hosted at the Institute for Advanced Study in Princeton. It brings together nine top‑tier global mathematicians, operates independently and receives no payment from the company. The group will mainly assess outputs, verify academic standards and relay feedback from the academic community to balance challenges brought by rapid AI outputs. Related studies show that multiple AI institutions have achieved phased progress on mathematical conjectures and theoretical derivations in recent years. Meanwhile, concerns keep emerging from the mathematics community. Some scholars worry that treating puzzle‑solving as an AI performance benchmark may disrupt the established rhythm of fundamental scientific research.

需要客观看待的是，AI给出推导不等于已经被整个学术界接纳。数学成果需要经过严谨的同行评审、反复核验逻辑漏洞，才能够真正被写入知识体系。AI擅长快速生成猜想与证明思路，但完整的逻辑校验、意义解读，依旧离不开人类数学家的判断。

English： Objectively speaking, derivations produced by AI do not equal full acceptance by the academic community. Mathematical findings need rigorous peer review and repeated checks for logical gaps before being formally integrated into established knowledge systems. AI excels at quickly generating conjectures and proof sketches, yet thorough logical verification and interpretation of significance still rely heavily on human mathematicians’ judgment.

这场事件更像是一个时代信号：AI不再只做聊天、绘图这类应用型任务，开始深度踏入基础科学的前沿阵地。未来它究竟会成为科研人员高效的协作工具，还是带来更多未知的争议，目前行业还没有统一答案。无论如何，人类与AI共同探索科学的新阶段，已经缓缓开启。

English： This development serves more as a signal of the times. AI is no longer limited to applied tasks such as chatting and image generation; it is stepping deep into the frontiers of fundamental science. There is no industry‑wide consensus yet on whether it will become a highly efficient collaborative tool for researchers or bring more unforeseen debates. Regardless, a new era of scientific exploration shared by humans and AI has quietly begun.

内容整理自网络公开信息，仅供参考交流，不代表本平台立场