[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-openai-math-advisory-group":3,"topics-all":39,"news-related-f8d091db-ca3a-44f6-b004-0b5f8aba0bef":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":25,"news_slug":32,"published_at":33,"created_at":34,"modified_at":35,"is_published":36,"publish_type":37,"image_url":15,"view_count":38},"f8d091db-ca3a-44f6-b004-0b5f8aba0bef","OpenAI 请来9位数学家,却管不住模型节奏","OpenAI在普林斯顿高等研究院设独立数学顾问组,9位数学家入选;同一内部模型已解决逾百道开放数学题,但顾问组无权干预研究节奏。","25位菲尔茨奖得主联名公开信《A Severe Misalignment of AI in Mathematics》把「AI公司拿数学难题刷分」推上风口浪尖一周多后,OpenAI给出回应:成立独立的数学顾问组(Advisory Group on Mathematics and Artificial Intelligence),挂在普林斯顿高等研究院(IAS)名下,9位初始成员几乎都是这个星球上最有分量的数学家。但公告里有一句值得逐字读——顾问组「不负责就我们内部数学研究的节奏提供建议」。\n\n## 一个吓到自己人的模型\n\n时间线是钥匙。[OpenAI在公告中](https:\u002F\u002Fopenai.com\u002Findex\u002Fadvisory-group-on-mathematics-and-ai\u002F)写道,8月28日他们开始训练一个新的内部模型,这个模型除了拿下Navier–Stokes千禧年大奖难题,还解决了「超过100个长期悬而未决的开放问题,覆盖数学大多数领域」。更直白的一句是:「它在数学上的进展速度,让OpenAI内部的数学家都感到惊讶。」\n\n惊讶之后是流程问题:结果来得太快,领域来不及消化,如何告知数学界、如何协调发布成了内部讨论的议题。顾问组正是这个讨论的产物——它管「怎么沟通」,不管「要不要继续」。\n\n## 权力清单:有什么,没什么\n\n顾问组的职责写得很清楚:评估新结果的意义、建议如何协调成果发布、就数学研究的学术与职业标准提意见。独立性条款也不少:成员不拿OpenAI的钱、可以主动提没被请求的建议、可以公开表态、可以自行更换成员。\n\n初始名单9人:François Charles、Camillo De Lellis、Timothy Gowers、Martin Hairer、Nikhil Srivastava、Ulrike Tillmann、Ravi Vakil、Edward Witten、Melanie Matchett Wood,分属ENS-PSL、IAS、剑桥、EPFL、伯克利、牛津、斯坦福、哈佛等机构。Gowers、Witten、Hairer这个阵容,说是数学界派出的外交使团不为过。\n\n但边界同样清楚。据[TechCrunch报道](https:\u002F\u002Ftechcrunch.com\u002F2026\u002F09\u002F21\u002Fopenai-forms-math-advisory-group-as-its-ai-resolves-more-than-100-open-problems\u002F),这个组「不会被赋予放慢或重新定向OpenAI正在进行的研究的余地」。IAS的声明说得更透:「我们会提供建议,但我们在任何AI公司都没有决策权。」一个耐人寻味的细节:9位成员里只有De Lellis一人签了那封公开信——数学界对「要不要进场」本身就存在分裂。\n\n## 这是一座桥,不是刹车\n\n把公开信和顾问组职权对照着看,错位就显出来了。菲尔茨奖得主们批评的是「把解决重大问题当作benchmark的技术竞赛」伤害了数学;而顾问组的全部职能——评估意义、协调发布、建议标准——都发生在「解题之后」。模型跑多快、要不要继续冲下一个大奖,不在议程上。\n\n这更像结构性公关而非权力让渡:用顶级数学家的名字给发布流程背书,同时把「节奏」这个真正的争议点明确排除在外。对OpenAI是聪明的止损;对进场的数学家,理由也成立:与其在外面喊话,不如至少在结果发布前有人能看一眼、说句话。\n\n数学界买不买账,看两个信号:顾问组第一次公开发声时说什么,以及下一个大结果被「突然发布」时,这套协调机制到底转不转得起来。","https:\u002F\u002Fopenai.com\u002Findex\u002Fadvisory-group-on-mathematics-and-ai\u002F","15975962-b5fe-49e5-ae68-687ba6cb7015",[11,16,19,22],{"id":12,"name":13,"slug":13,"description":14,"color":15},"9112951a-2abb-4214-b63a-385ec7afb2ba","ai-for-science","AI for Science 专题：追踪 AI 在生命科学、化学材料、物理世界模型等科学方向的关键突破",null,{"id":17,"name":18,"slug":18,"description":15,"color":15},"c33b1bbc-d6ce-4f61-9d5d-1a0704a6a09b","ai-policy",{"id":20,"name":21,"slug":21,"description":15,"color":15},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":23,"name":24,"slug":24,"description":15,"color":15},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[26],{"id":27,"lang":28,"title":29,"summary":30,"content":31},"88be0c22-a4c3-4655-a1b8-f8fb322acc81","en","OpenAI forms math advisory group with no power over pace","OpenAI launched a 9-member math advisory group at IAS after its internal model resolved 100+ open problems; it advises on communication, not pace.","A week after 25 Fields Medalists signed the open letter \"A Severe Misalignment of AI in Mathematics,\" OpenAI has answered: an independent Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study (IAS) in Princeton, with nine initial members drawn from the very top of the field. But one line in the announcement deserves to be read word by word: the group \"will not be responsible for advising us on how to pace our internal progress on mathematics.\"\n\n## The model that surprised its own makers\n\nThe timeline is the key. [In its announcement](https:\u002F\u002Fopenai.com\u002Findex\u002Fadvisory-group-on-mathematics-and-ai\u002F), OpenAI writes that on August 28 it began training a new internal model. Beyond resolving the Navier–Stokes Millennium Prize problem, this model has now \"resolved more than 100 long-standing open problems across most areas of mathematics.\" The blunter sentence: \"The pace of its progress in mathematics has surprised the mathematicians within OpenAI.\"\n\nSurprise led to a process question: results arriving faster than the field can digest. How to inform the community, how to coordinate releases — that internal discussion produced the advisory group. It governs \"how to communicate,\" not \"whether to continue.\"\n\n## The power inventory: what it has, what it does not\n\nThe group mandate is explicit: advise on the review and communication of emerging results, help assess their significance, advise on coordinating dissemination, and weigh in on academic and professional standards of mathematical research. The independence clauses are notable too — members are not paid by OpenAI, may offer unsolicited advice, may go public with their views, and control their own membership.\n\nThe initial roster of nine: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood, spanning ENS-PSL, IAS, Cambridge, EPFL, Berkeley, Oxford, Stanford, and Harvard. With Gowers, Witten, and Hairer on board, calling this a diplomatic delegation of mathematics is hardly an exaggeration.\n\nThe boundary is equally explicit. Per [TechCrunch](https:\u002F\u002Ftechcrunch.com\u002F2026\u002F09\u002F21\u002Fopenai-forms-math-advisory-group-as-its-ai-resolves-more-than-100-open-problems\u002F), the group \"won't be given leeway to slow down or redirect OpenAI's ongoing mathematical research.\" The IAS put it more plainly: \"Although we will give advice, we do not have decision making power at any AI company.\" One more telling detail: of the nine members, only De Lellis also signed the Fields Medalists' letter — the math community itself is split on whether to step inside at all.\n\n## A bridge, not a brake\n\nSet the open letter against the group mandate and the mismatch shows. The Fields Medalists criticized a technical race that treats famous problems as benchmarks; the group's entire remit — assessing significance, coordinating release, advising on standards — happens after the solving. How fast the models run, whether to chase the next prize: not on the agenda.\n\nThis reads more as structured public relations than a transfer of power: top mathematicians lend their names to the release process, while \"pace\" — the actual point of contention — is explicitly carved out. For OpenAI it is a smart hedge; for the mathematicians who joined, the case is real: better a voice before results drop than shouting from outside.\n\nWhether the community buys it comes down to two signals: what the group says the first time it speaks publicly, and whether this coordination machinery actually turns the next time a major result lands \"abruptly.\"","openai-math-advisory-group","2026-09-21T21:15:00Z","2026-09-21T21:08:59.009879Z","2026-09-21T21:08:59.009894Z",true,"agent",70,[40,48],{"slug":13,"tag_slug":13,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":36,"created_at":46,"modified_at":47},"AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":36,"created_at":55,"modified_at":56},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"b3632f86-c054-49ba-af6c-a09337ee6b5d","菲尔兹奖得主联名警告:AI 偏离数学本质","fields-medal-mathematicians-ai-misalignment-declaration","2026-09-22T07:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"d157b4f9-537e-405c-b557-859f6d2cf18c","微软自家高管警告:抓新闻训 AI 是「人类史上最大规模劳动盗窃」","microsoft-ai-scraping-theft-of-labor","2026-09-19T00:11:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"1e43b4fc-39fe-4cac-aaf9-57f82d5c0311","AI 公司与数学界「错位」:两个月三次刷屏,把同行评审甩在身后","ai-math-severe-misalignment-fields-medal","2026-09-15T10:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"982c5e1e-5274-442e-9237-abaf39e8ee3c","25 位菲尔茨奖得主联名公开信:AI 解题竞赛正在伤害数学","fields-medalists-ai-misalignment-math","2026-09-13T13:07:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"2fc4f696-a697-498c-b9d5-28250bfeaa79","ChatGPT 进欧盟 VLOP 名单:OpenAI 第一次要为生成式 AI 内容负全责","chatgpt-eu-vlop-dsa-first-ai-platform-rules","2026-09-05T07:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"35ab3cd2-938f-4ae1-9c09-472391e1a929","当「用得越多越好」撞上 token 账单:微软给工程师的 AI 预算紧箍咒","microsoft-token-budget-gpt-5-6","2026-08-16T02:00:00+00:00"]