Anthropic数学家凭Claude证明黎曼猜想
[AINews] Riemann Hypothesis proved...twice independently
AI辅助数学证明的重大突破极具范式意义,同时Inkling 2 Mini展示了独特的循环架构创新,值得从业者关注。
Within a few hours of each other (timeline), two mathematicians claimed independent proofs of the Riemann Hypothesis, the most famous open problem in mathematics (Scientific American).
在短短几小时内(按时间线),两位数学家声称独立证明了黎曼猜想,这是数学界最著名的未解之谜(《科学美国人》)。
- Levent Alpöge of Anthropic first made his Claude-driven proof in a note released with Lean code (1.3GB worth, ~250,000 page equivalent) rather than a full paper. Elliot Glazer confirms that the formalized statement is GRH and puts the chance of a Lean bug well under 1%. So if it finishes compiling with no extra axioms, it is probably correct, and no human will ever have read all of it…
- …prompting Alessandro Masullo to release his 940 page, 100% human, work in progress proof he had been working on for the last 12 years (and had been undergoing secret manual review and revision per Kevin Hartnett).
- Anthropic 的 Levent Alpöge 首先发布了一份附带 Lean 代码的证明笔记(价值 1.3GB,约相当于 25 万页),而非完整论文。Elliot Glazer 确认形式化陈述为广义黎曼猜想(GRH),并指出存在 Lean 错误的可能性远低于 1%。因此,如果编译完成且无需额外公理,该证明大概正确,但恐怕永远没有人能通读全部内容……
- ……这促使 Alessandro Masullo 发布了他的 940 页、完全由人类完成的、历时 12 年正在进行的证明工作(此前一直按照 Kevin Hartnett 的要求进行秘密的手工审查和修订)。
This is the third in the trilogy of frontier math breakthroughs, following from the Navier Stokes solutions and the Quasi-Riemann bounding, but it doesn’t feel over. OpenAI's September Navier-Stokes claim is still the cautionary tale: critics argued that its Lean proof did not match the natural-language result. Both AI-claimed Millennium Prize results will take years to verify.
这是前沿数学突破三部曲中的第三部,继纳维-斯托克斯方程解和准黎曼界限之后,但感觉尚未结束。OpenAI 九月关于纳维-斯托克斯方程的主张仍是一个警示案例:批评者认为其 Lean 证明与自然语言结果不符。这两项由 AI 声称的千禧年大奖难题结果都需要数年时间才能验证。
Apart from the incredible accomplishment of the most famous problem in mathematics, one cannot miss the contrast that the AI-written solution is probably correct but incomprehensible, and the other solution is probably technically incorrect but understandable.
除了数学界最著名问题的惊人成就外,人们无法忽视这种对比:AI 编写的解法可能正确但难以理解,而另一份解法可能在技术上不正确但可以理解。
Sponsored by Agentfield.AI
由 Agentfield.AI 赞助
Open models are ready. Most coding harnesses aren’t. On DeepSWE, CodeAF finished #1 of ten on the same open model, ahead of Claude Code, at the lowest cost per fix.
开源模型已就绪。大多数代码测试框架却不然。在 DeepSWE 上,CodeAF 使用相同的开源模型在十家机构中排名第一,领先于 Claude Code,且每次修复的成本最低。
Try CodeAF!
试试 CodeAF!
As Levent eloquently put it: “Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line.”
正如 Levent eloquently 所言:“在接近终点时,Claude 远远超越了我。不,Claude 是飞翔、起舞并闪耀着光芒冲过那条终点的。”
This also ties into other AI themes we’ve covered on the podcast and on this newsletter:
这也与我们在播客和本通讯中探讨的其他 AI 主题相关联:
- The nature of Slop: Levent says “I am presenting the world with a complete argument before having mastered a massive neighbourhood around its mathematics. It being obviously inappropriate to attempt to post some massive unreadable travesty of "slop", I decided to only present Lean.” One of the key tenets of the No Slop Manifesto is that you should understand what you ship (“Slop is when we put less effort into producing something than we ask the consumer to spend understanding it.”). And yet, this giant pile of Lean is verifiably correct and is incredibly significant in human history. Should we have delayed its release by a decade simply because we didn’t understand every single bit of it?
- Weak to Strong alignment: the typical experience of using agents is Strong to Weak - we are general intelligences guiding not-as-smart AI to do things we know how to do well. We also often expect AI to automate most work “within distribution”, and expect it to run into more trouble at the limits of capability where the “centaur” pattern of human-AI-collaboration is required. Levent’s lived experience is the opposite - collaborating with Claude only in the early stages and then leaving him far behind with the critical parts involving automorphic forms on higher-rank groups. We are starting to need Weak to Strong alignment — the open ended domain of controlling AI smarter than us.
- Slop 的本质:Levent 表示“我在尚未完全掌握其数学周边庞大领域之前,就向全世界呈现了一个完整的论证。显然,试图发布一堆庞大且难以阅读的‘Slop’(低质内容)是不合适的,因此我决定只展示 Lean 代码。”No Slop Manifesto 的核心原则之一是,你应该理解你所交付的内容(“当我们为生产某物所付出的努力,少于要求消费者去理解它时所花费的努力时,这就是 Slop”)。然而,这一大堆 Lean 代码在验证上是正确的,并且在人类历史上具有极其重要的意义。仅仅因为我们没有理解其中的每一个细节,我们就应该将其发布推迟十年吗?
- 从弱到强的对齐(Weak to Strong alignment):使用智能体的典型体验是从强到弱——我们是通用智能体,引导不如我们聪明的 AI 去做我们知道如何做得很好的事情。我们还经常期望 AI 自动化大部分“分布内”的工作,并预期它在能力极限处会遇到更多麻烦,而在那里需要人机协作的“半人马”模式。Levent 的实际经历则恰恰相反——仅在早期阶段与 Claude 合作,然后在涉及高阶群上自守形式的关键部分将其远远甩在身后。我们开始需要“从弱到强的对齐”——即控制比我们更聪明的 AI 的开放域问题。
Sponsored by Backblaze B2
由 Backblaze B2 赞助
If your data is growing faster than planned, you need storage that grows with your business, feeds flash and GPUs with terabit-scale throughput, and doesn’t charge egress.
如果你的数据增长速度超过计划,你需要能够随业务增长的存储,提供太比特级吞吐量以供给闪存和 GPU,并且不收取出口流量费用。
Backblaze B2 has delivered capacity tier storage for twenty years, manages more than six exabytes, and is trusted by leading neoclouds, labs, and gen AI companies.
Backblaze B2 已提供容量层存储二十年,管理着超过六艾字节的容量,并受到领先的新云服务商、实验室和生成式 AI 公司的信赖。
Growth shouldn’t be something you sweat. We’ve got room for whatever you’re building.
增长不应该是让你焦虑的事情。无论你在构建什么,我们都有足够的空间。
Come and try it out!
来试试吧!
AI Twitter Recap
AI Twitter 回顾
Claimed Proof of the Riemann Hypothesis with Claude
声称与 Claude 共同证明黎曼猜想
- The claim: Anthropic mathematician Levent Alpöge says that, "working with Claude," he has found a proof of the Riemann Hypothesis. He says the method also covers the Generalized Riemann Hypothesis (GRH) (Scientific American, statement summary, remark.pdf link).
- Artifact: Only Lean code was released. A written-out prose version reportedly ran past 250,000 pages (timeline).
- Timeline: The work began on Oct 6. Eight of the 20 steps landed on Nov 15, and the final entry on Nov 16 reads "GRH." Alpöge credits Claude for the parts involving automorphic forms on higher-rank groups, an area where he says he has little mastery (breakdown).
- Unknowns: The Claude model used and the degree of human steering have not been disclosed (open questions).
- Verification status: Elliot Glazer confirms that the formalized statement is GRH. The Lean code is "still compiling, will take a few hours." He puts the chance of a Lean bug at <<1% (analysis).
- Provisional status: Artificial Analysis notes that both AI-claimed Millennium results are unverified: this one and OpenAI's September Navier-Stokes claim, which used an unreleased model (tally).
- Competing human proof: Alessandro Masullo has announced a separate, purely human proof of RH. It is reported as a roughly 940-page GRH draft (announcement, page count).
- Interpretation debate: Some argue the result shows we have little idea how hard math problems objectively are, rather than showing broad superhuman ability (Theo Jaffee, Ege Erdil).
- Brute-force framing: Others compare the approach to olympiad-style "bashing" at a scale humans lack the willpower for (cloneofsimo).
- Jaggedness: Omar Khattab notes the contrast between solving Millennium problems and being "this dumb on 1000 other things" (post).
- Market reaction: ANT stock fell about 10% on the day (scaling01, Nabeel Qureshi).
- Unconfirmed rumor: A DeepMind "math horizon" repo appearing the same day prompted speculation about Gemini math results (unconfirmed).
- 声明:Anthropic 数学家 Levent Alpöge 表示,“在与 Claude 合作的过程中”,他找到了黎曼猜想的证明。他还称该方法也涵盖了广义黎曼猜想(GRH)(《科学美国人》报道,声明摘要,remark.pdf 链接)。
- 成果文件:仅发布了 Lean 代码。据报道,书面叙述版本超过了 250,000 页(时间线)。
- 时间线:工作于10月6日开始。20个步骤中有8个落在11月15日,而11月16日的最终条目写着“GRH”。Alpöge将涉及高阶群上自守形式的部分归功于Claude,他表示自己在该领域掌握甚少(分解)。
- 未知因素:所使用的Claude模型以及人类引导的程度尚未披露(待解决问题)。
- 验证状态:Elliot Glazer确认形式化后的陈述为GRH。Lean代码“仍在编译中,需要几个小时”。他将Lean存在bug的可能性估计为<1%(分析)。
- 临时状态:Artificial Analysis指出,AI声称的两个千禧年难题结果均未得到验证:一个是本文所述,另一个是OpenAI在9月提出的纳维-斯托克斯方程声明,后者使用了未发布的模型(统计)。
- 竞争性人类证明:Alessandro Masullo宣布了一项独立的纯人类RH证明。据报道,这是一份约940页的GRH草稿(公告,页数)。
- 解读争议:一些人认为,这一结果表明我们对数学问题客观难度知之甚少,而非表明具备广泛的人类超越能力(Theo Jaffee、Ege Erdil)。
- 暴力求解框架:另一些人将该方法比作奥林匹克竞赛风格的“暴力破解”,其规模超出了人类的意志力所能承受的范围(cloneofsimo)。
- 不均衡表现:Omar Khattab指出,解决千禧年难题与在其他1000件事上表现得“如此愚蠢”之间存在反差(帖子)。
- 市场反应:ANT股票当日下跌约10%(scaling01、Nabeel Qureshi)。
- 未经证实的传闻:同一天出现的DeepMind“math horizon”仓库引发了关于Gemini数学结果的猜测(未经证实)。
Thinking Machines Releases Inkling 2 Mini, an Open-Weights Looped Transformer
Thinking Machines发布Inkling 2 Mini,一款开源权重的循环Transformer
- Inkling 2 Mini: A 26.4B-parameter looped MoE under Apache 2.0, pitched as small enough for a laptop. It scores 20.1 on the AA Index (launch, Mira Murati).
- Architecture: There are 12 shared core layers run for 3 passes, with gated messages carried between passes (vLLM).
- Active parameters: 3.4B unique active parameters, or 5.4B effective counting the loops. It has a 1M context window and accepts text and image input (AA).
- Training recipe: Two-thirds of the experts were pruned during mid-training, followed by logit distillation from a larger Inkling (eliebakouch).
- Claimed results: It leads open models under 100B on DeepSWE v1.1 (48.2%) and Vals Index v2.2 (46.5%) (via teortaxes).
- Comparison to Inkling 1.0: It reportedly beats Inkling 1.0 at 2.7% of the size (Ziqiao Ma).
- Independent evals:
- Vals: It ties Claude Opus 5.5 on CyberBench at under a quarter of the cost and keeps 96–97% of Inkling 2 Small's Vibe Code and SAGE scores (Vals, details).
- Artificial Analysis: AA places it on the Pareto frontier for intelligence vs. active parameters (AA).
- ARC-AGI-2: It reportedly scores 49.3% at $0.14/task, well below Inkling 2 Small (teortaxes).
- Safety and interpretability rationale: Thinking Machines says the open release is meant to let researchers study how loops affect chain-of-thought monitorability (Alex Robey).
- Variable-loop model: A companion research model, TML-d24-varloop-research, shows long-context scores (GraphWalks, MRCR v2) improving with more passes. Knowledge benchmarks such as SimpleQA and AA-Omniscience do not improve (Ziqiao Ma).
- Goodfire analysis: Goodfire mapped information flow across layers, blocks and loops (Goodfire).
- NDIF hosting: NDIF hosts both models with tooling for activation patching across loops (NDIF).
- Serving and local availability: vLLM had day-0 support.
- GB200 decode: TP1 decode on GB200 drops from 10.29 to 3.84 ms at batch 1, with kernels per step falling from 4,278 to 927 (numbers).
- Inkling 2 Mini:一款拥有264亿参数的Apache 2.0许可循环MoE模型,号称足够小巧可运行于笔记本电脑上。它在AA Index上得分20.1(发布,Mira Murati)。
- 架构:包含12层共享核心层,运行3遍,各遍之间传递门控消息(vLLM)。
- 活跃参数:34亿唯一活跃参数,若计入循环则为54亿有效参数。它拥有100万上下文窗口,并接受文本和图像输入(AA)。
- 训练配方:在训练中后期剪枝了三分之二的专家,随后从更大的Inkling模型进行logit蒸馏(eliebakouch)。
- 宣称结果:在DeepSWE v1.1(48.2%)和Vals Index v2.2(46.5%)上,它领先于百B以下的开源模型(via teortaxes)。
- 与Inkling 1.0对比:据报道,它以仅2.7%的体积超越了Inkling 1.0(Ziqiao Ma)。
- 独立评估:
- Vals:在 CyberBench 上,其表现与 Claude Opus 5.5 持平,成本不到后者的四分之一,并保持了 Inkling 2 Small 96–97% 的 Vibe Code 和 SAGE 分数(Vals,详情)。
- Artificial Analysis:AA 将其置于智能度与激活参数数量的帕累托前沿上(AA)。
- ARC-AGI-2:据报道,其在每项任务 0.14 美元的成本下得分 49.3%,远低于 Inkling 2 Small(teortaxes)。
- 安全性和可解释性依据:Thinking Machines 表示,开源发布旨在让研究人员研究循环如何影响思维链的可监控性(Alex Robey)。
- 可变循环模型:配套研究模型 TML-d24-varloop-research 显示,随着迭代次数增加,长上下文分数(GraphWalks、MRCR v2)有所提升。知识类基准测试如 SimpleQA 和 AA-Omniscience 则未见改善(Ziqiao Ma)。
- Goodfire 分析:Goodfire 绘制了跨层、块和循环的信息流图(Goodfire)。
- NDIF 托管:NDIF 托管了这两个模型,并提供跨循环的激活修补工具(NDIF)。
- 部署和本地可用性:vLLM 提供了首日支持。
- GB200 解码:在 GB200 上,TP1 解码在 batch size 为 1 时从 10.29 毫秒降至 3.84 毫秒,每步内核数从 4,278 个降至 927 个(数据)。
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力