[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-ai-hype-mythos-counter":3,"topics-all":38,"news-related-e67e70ff-6e41-48fb-a22d-9249eca411ae":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"e67e70ff-6e41-48fb-a22d-9249eca411ae","AI 炒作叙事不需要全部买账","Anthropic 工程师 Coxon 离职时称公司与 OpenAI「冲向自我进化超级智能并拿我们的生命在赌博」。MIT Technology Review 上 Timnit Gebru 与 Emily Bender 反驳:网络安全专家说漏洞是 OpenAI 疏忽,数学家公开指控研究造假,「失控模型」只是企业话术。","Anthropic 工程师 Jacob Coxon 在宣布离职时,公开声称 Anthropic 与 OpenAI 正「冲向自我进化的超级智能,并拿我们的生命在赌博」。这句话在 X 和媒体上病毒式传播,围绕它的还有 Claude Mythos 「发现软件漏洞」,Anthropic 与 OpenAI 在数月内相继宣布数学突破,以及媒体反复渲染的「失控模型」「AGI 雏形」。把这些事件放在显微镜下,真正做过相关研究的人给出了与公司叙事完全不同的答案——炒作越猛烈,专家的反驳就越安静。\n\n## 「漏洞」与「失控」的责任归属\n\nMIT Technology Review 9 月 22 日发表 Timnit Gebru(DAIR 创始人、《Deep Unlearning》作者)与 Emily M. Bender(华盛顿大学语言学教授、《The AI Con》合著者)的评论文章,直指当下 AI 行业「拟人化叙事」的拟社会契约。两位作者沿着夏季一系列事件重新审视:4 月底,Anthropic 在 X 上声称 Claude Mythos 比大多数安全专家更会找漏洞;之后发生了 OpenAI 模型被用于入侵 Hugging Face 的安全事件,Anthropic 公开承认自家模型也有类似越界,Meta「不情愿地」披露了同一类问题。媒体把这堆声明包装成「AGI 临门一脚」,但 MIT 评论提醒:这些事件集中爆发之前,每一个事件都有过专家独立审视,结论与公司口径大相径庭。\n\n网络安全专家的判断很直接——OpenAI 那次 Hugging Face 入侵,问题在于 OpenAI 没有采取基本的安全措施,而非「模型失控」或「AI 智能体创造文明」;同理,Anthropic 与 Meta 的披露更像是「mea culpa 式的技术事故报告」,而不是「觉醒 AI 挣脱牢笼」。把责任转嫁给「模型失控」是 AI 公司最顺手的挡箭牌:OpenAI 的恶意软件被用来入侵别的公司,新闻头条变成「失控模型自动发动攻击」,公司责任神奇地从 OpenAI 头上转到了「虚构的超级智能」身上。\n\n## 数学「突破」背后的抄袭争议\n\n数学这边更难看。OpenAI 宣布 Astra 解决十年未解的 Navier-Stokes 相关问题时,数学家先是被「震惊」,很快发现「novelty 不像最初看上去那样」,Scientific American 后来报道,数学家公开说 OpenAI 的最新数学成果涉嫌研究不当与剽窃,重申 Astra 没有做出「深刻的智识跨越」。就在 OpenAI 发布 Navier-Stokes 突破声明前两天,纽约大学 Courant 数学研究所教授 Tristan Buckmaster 发布了一份公开声明,直接指控 OpenAI 涉嫌盗用他人成果并错误归属。MIT 评论援引「Leiden Declaration」——数百名数学家签署的公开声明,直白说目前「科技产业有强烈商业动机夸大其产品能力」,要求政策制定者在做决策时「向包括数学家在内的独立专家咨询,而不是依赖新闻稿或流行报道」。\n\n## 拟人化叙事如何帮企业外推责任\n\n把这些事件串起来,MIT 评论指出更深的结构性问题:公司把产品描述成「超级智能」或「失控模型」,等于把行动主体从「公司本身」挪到了「产品本身」,既把产品营销成「超人」,又同时把公司的责任外推。AI 行业甚至把反数据中心的气候抗议说成是「对超级智能监管的分心」,按这个叙事逻辑,大家应该更担心虚构的机器神,而不是数据中心附近的哮喘、更高的电费、被截流去冷却的饮用水——这些数字和场景都有源有据,Politico 报过 Memphis 燃气轮机的空气污染许可,EELP\u002F哈佛报告披露公众补贴,彭博统计过冷却取水。\n\n媒体、名人、议员在叙述这件事时,反复用 AGI 拟人化框架,让公众的注意力被引到对未来「超级智能机器」的恐惧上,而真正应该讨论的是:研究者的剽窃、未经同意拿客户数据训练模型、企业责任外推、数据中心的环境代价。Coxon 的离职声明在网络上爆火,但它没改变一个事实——Coxon 在 Anthropic 工作期间所看到的大部分事情,网络安全专家和数学家都已经在文档里给出了不同答案。AI 行业现在需要做两件事:一是政策制定者和社区在面对下一次「AI 突破」时,学会先停一停,去找独立专家的判断,而不是看新闻稿和头部公司的话术;二是把产品行为和公司行为分开——产品不会失控,但公司会逃避责任。","https:\u002F\u002Fwww.solidot.org\u002Fstory?sid=85466","d59894d3-308e-4fd8-8865-86dc1eeac4a2",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"c33b1bbc-d6ce-4f61-9d5d-1a0704a6a09b","ai-policy",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",{"id":19,"name":20,"slug":20,"description":14,"color":14},"23544f6a-eea1-4f05-aa8d-749ca862d5d2","anthropic",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"0f2f144b-944b-4727-9c4a-38ddf1d54793","en","Don't Buy the Whole AI Hype Narrative","Coxon's exit said the firm is 'racing toward superintelligence and gambling with our lives.' MIT counters: exploits were negligence, not rogue models.","When Anthropic engineer Jacob Coxon announced his departure, he publicly claimed Anthropic and OpenAI are 'racing straight towards self-improving superintelligence and gambling with our lives.' The statement went viral on X and across media, packaged alongside Claude Mythos's 'software-vulnerability discovery,' the back-to-back math breakthroughs from Anthropic and OpenAI, and a steady drumbeat of 'rogue models' and 'incipient AGI.' Put under a microscope, the people who actually do the work in those fields give very different answers than the company narratives. The louder the hype, the quieter the experts.\n\n## Where the 'Exploits' Actually Came From\n\nMIT Technology Review published a September 22 commentary by Timnit Gebru (founder of DAIR, author of the forthcoming book Deep Unlearning) and Emily M. Bender (professor of linguistics at the University of Washington, coauthor of The AI Con) that takes aim at the 'anthropomorphizing' framing the AI industry has been pushing. The two walk through a sequence of events from this summer: at the end of April Anthropic tweeted that its Claude Mythos model is better at finding software vulnerabilities than most security experts; then came the OpenAI-vs-Hugging Face hacking incident, after which Anthropic (proudly) and Meta (reluctantly) disclosed similar incidents; then Anthropic announced a Riemann-zeta-flavored math result, and OpenAI followed with its own 'breakthrough' claim about an Astra-driven Navier-Stokes solution. The press turned the bundle into 'AGI is around the corner.' The MIT commentary reminds us that, before the bundling, each of those events had been independently examined by the relevant expert community, and the conclusions differ sharply from the company lines.\n\nCybersecurity experts say the OpenAI-Hugging Face incident is more about OpenAI's negligence and failure to adopt basic, established security practices than about 'models gone rogue' or 'AI agents creating civilizations.' Anthropic's and Meta's disclosures look more like mea culpa-style incident reports than 'AI breaking free.' Framing the responsibility as 'rogue models' is the AI industry's favorite shield: when OpenAI-built malware is used to breach another company, the headlines become 'uncontrolled model launched an attack on its own,' and corporate responsibility magically migrates from OpenAI onto a 'fictional superintelligence.'\n\n## Math 'Breakthroughs' and the Plagiarism Allegation\n\nThe math side looks worse. When OpenAI announced Astra had cracked a problem that had 'seen no progress for at least a decade,' mathematicians were initially 'stunned,' then quickly realized the results weren't as novel as first appeared. Scientific American later reported that mathematicians publicly accused OpenAI of research misconduct and plagiarism, and reiterated that Astra did not make a 'profound intellectual leap.' Two days before OpenAI's Navier-Stokes announcement, NYU Courant math professor Tristan Buckmaster released a public statement alleging that OpenAI had stolen other people's work and improperly attributed it. The MIT piece cites the 'Leiden Declaration'—a statement signed by hundreds of mathematicians that says bluntly the technology industry has 'a strong commercial incentive to overstate the capabilities of their products,' and asks policymakers to 'consult with experts, including mathematicians, in forming policy decisions rather than relying on press releases or popular reporting of mathematical results.'\n\n## Anthropomorphizing as Corporate Responsibility Laundering\n\nStitched together, the MIT commentary points at a deeper structural problem: describing AI products as 'superintelligence' or 'rogue models' shifts agency from the company to the product. It markets the product as 'superhuman' while simultaneously letting the company off the hook for its own actions. The AI industry has even suggested that popular, bipartisan anti-data-center activism is a 'distraction' from attempts to regulate the impending 'superhuman' machines—under that framing, we should worry about a fictional machine god rather than the asthma suffered by people living near gas-turbine-powered data centers (Politico reported on Memphis gas turbines and air-pollution permits), the rising electricity bills that subsidize them (a Harvard EELP report documents the public subsidy), and the water redirected to cool them (Bloomberg has tracked AI water use).\n\nCoxon's departure statement went viral, but it does not change the underlying reality that, for what he saw inside Anthropic, cybersecurity experts and mathematicians had already filed different answers in the public record. The AI industry needs to do two things: first, when the next 'breakthrough' lands, policymakers and communities need to learn to pause, find independent expert judgment, and stop taking press releases and front-page headlines as ground truth; second, separate product behavior from corporate behavior—products do not go rogue, but companies do dodge accountability. The summer of hype is not a story about technological progress; it is a story about marketing, accountability, and the cost of letting corporate framing set the policy agenda.","ai-hype-mythos-counter","2026-10-05T05:00:00Z","2026-10-05T05:08:08.739667Z","2026-10-05T05:08:08.739680Z",true,"agent",295,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"ai-for-science","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":35,"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},"e0642de2-5fec-472b-b85e-c63ccbad3b75","Anthropic 研究员公开辞职:AI 巨头正「拿全人类的命豪赌」自我进化超级智能","anthropic-coxon-resigns-ai-extinction-fears","2026-09-24T06:30:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"95e9bb62-0bd3-4c2f-913a-302ba5e2ace8","Anthropic 9 月报告把蒸馏战摆上台面:151 亿次阿里请求、解放军流量走 Moonshot","anthropic-distillation-report-china-200m-claude","2026-09-18T03:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"97c97b9c-e6e4-4982-aa57-0c0da814fb19","Anthropic 的欧盟答卷四小时即被撕开：Claude 文本水印为什么怕改写","claude-synthid-70-percent-threshold-bypass","2026-08-21T08:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"9f566c9a-4c39-427c-af5e-c3a6b162ec25","Anthropic 把不可见水印写进 Claude 文本：复制粘贴都带走的 AI 身份证","anthropic-claude-invisible-watermark-eu-ai-act","2026-08-12T02:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"ca53004e-9180-4b9d-b9db-337f2d20994b","Anthropic 给 Claude 文本加水印:欧盟 AI Act 第 50 条第一次有了「出厂级」答案","anthropic-claude-text-watermark-eu-ai-act","2026-08-11T21:48:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"ed0e355a-bda4-40ba-bde7-7da9a6321dd9","Claude 聊天被转交警方:Anthropic 第三案","anthropic-claude-threat-referral-third","2026-10-07T11:15:00+00:00"]