IT之家 9 月 22 日消息,OpenAI 宣布成立“数学与 AI 顾问组”(Advisory Group on Mathematics and Artificial Intelligence)。据介绍,该顾问组设在普林斯顿高等研究院,由九位数学家组成,将独立于 OpenAI 运作。
相较于此,更重磅的是 OpenAI 在公告中披露 —— 公司于 8 月 28 日开始训练的一个神秘模型已成功应对了多个数学领域中 100 多个长期未应对的开放难题(包含此前引发争议的纳维-斯托克斯千年大奖难题)。
OpenAI 称,该模型在数学领域的进展速度令公司内部的数学家感到惊讶,这引发了关于如何以最佳方式告知学界的内部讨论。
OpenAI 认为,数学属于基础科学领域,AI 取得新的数学发现后可能产生广泛应用,因而相关能力的开发和部署需要采取更稳重的方式。公司当下正在研究如何更广泛地部署与数学相关的 AI 能力。
此前,陶哲轩、邓煜、彼得 · 舒尔茨等数学家联名发表公开信,对将应对数学开放难题作为新型 AI 平台基准可能产生的负面影响表达担忧。OpenAI 表示,这些意见表明 AI 公司有必要与数学界开展更充分的交流。
OpenAI 明确,该顾问组独立于 OpenAI 运作,有权提出公司未请求的建议、公开评论 OpenAI 对数学的影响,并可自行决定调整成员构成,成员不从 OpenAI 领取报酬。不过,OpenAI 同时说明,该顾问组不负责就公司内部数学研究的推进节奏给出建议。
顾问组首批成员涵盖:
其中 Gowers、Hairer 和 Witten 均为菲尔兹奖得主(IT之家注:九位成员中仅 Camillo de Lellis 一人同时是前述公开信的签署者)。
该顾问组将就新兴成果的评审与传播给出建议,协助 OpenAI 评估成果的关键性,并就数学研究的学术标准给出意见。小组还将就 OpenAI 的工具如何兼容数学研究与学习给出建议。
相关阅读:
本文来源:https://www.ithome.com/1/005/500.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Home > IT News > Industry
On September 22, OpenAI announced the formation of the "Advisory Group on Mathematics and Artificial Intelligence." According to the report, this advisory group is based at the Princeton Institute for Advanced Study and consists of nine mathematicians, operating independently of OpenAI.
More significantly, OpenAI disclosed in its announcement that a mysterious model, beginning training on August 28, has successfully tackled over 100 long-standing open problems across multiple mathematical fields, including the previously controversial Navier-Stokes Millennium Prize problem.
OpenAI stated that the pace of this model’s progress in mathematics surprised even its own mathematicians, leading to internal discussions about the best way to inform the academic community.
The company believes that since mathematics is a fundamental science and new AI discoveries in this field could have wide-ranging applications, the development and deployment of such capabilities require a more measured approach. OpenAI is currently researching how to deploy mathematics-related AI capabilities more broadly.
Previously, mathematicians like Terence Tao, Yu Deng, and Peter Schultz signed an open letter expressing concern that using open mathematical problems as benchmarks for new AI platforms could have negative consequences. OpenAI responded that these opinions indicate a need for greater communication between AI companies and the mathematical community.
OpenAI clarified that the advisory group operates independently, with the authority to make unrequested recommendations, publicly comment on OpenAI’s impact on mathematics, and adjust its membership at its discretion. Members do not receive compensation from OpenAI, though the group does not advise on the pace of internal mathematical research.
The initial members include:
Among them, Gowers, Hairer, and Witten are all Fields Medal winners (note that only Camillo de Lellis is also a signer of the previous open letter).
The group will advise on the review and dissemination of emerging results, help evaluate their significance, and provide guidance on academic standards in mathematical research. It will also recommend how OpenAI’s tools can integrate with mathematics education and research.
Further reading:
