[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-deepmind-abstract-fallacy-llm-consciousness-paper":3,"topics-all":34,"news-related-746de03f-1378-49e5-80a2-927ffa13ecf5":53},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":20,"news_slug":26,"published_at":27,"created_at":28,"modified_at":29,"is_published":30,"publish_type":31,"image_url":32,"view_count":33},"746de03f-1378-49e5-80a2-927ffa13ecf5","DeepMind 论文再次引爆争论：LLM 永远无法拥有意识","一篇来自 Google DeepMind 高级研究员 Alexander Lerchner 的论文《抽象化谬误：为何 AI 能模拟但无法实例化意识》正在引发学界和产业界的广泛争论。这篇论文的核心命题简单而激进：LLM 永远不会拥有意识——无论参数量多大、推理能力多强，意识是一种物理状态，而非可以意外或有意创造的软件制品。\\n\\n**论文的核心论证**\\n\\nLerchner 区分了两个概念：模拟（simulate）与实例化（instantiate）。当前的 LLM 在处理语言、图像和多模态信息时，能极为逼真地模拟人类意识的输出——它能讨论感受、描述体验、引用哲学文本。但论文认为，这种模拟本质上是统计模式的复现，而非真正的心智状态实例化。就像一个能完美演奏悲伤曲子的钢琴，它并不感受悲伤。\\n\\n论文指出，意识的物理状态涉及生物神经系统的具身性、感官运动回路和情绪调节机制——这些在硅基计算系统中根本不存在对应的等价物。即便 LLM 在行为层面越来越接近人类，行为相似性不等于同构的内在体验。\\n\\n**DeepMind 的内部博弈**\\n\\n这篇论文的意义不仅在于学术结论，还在于它的泄露过程：404 Media 披露，论文最初带有 Google DeepMind 的官方信头，随后在媒体追问下被移除，改为注明代表作者个人观点。这一变动本身就折射出大厂内部对前沿 AI 安全与意识问题的高度敏感——一边是研究人员在科学层面给出严肃论断，一边是公司公关需要管理舆论风险。\\n\\n**对行业的深层影响**\\n\\n这篇论文戳破了一个行业潜规则：几乎所有头部 AI 公司都在公开场合给 AGI 时间线降温，但在内部论文中，意识问题被严肃地提上议程。它说明当 LLM 能力逼近人类水平时，AI","https:\u002F\u002Fdeepmind.google\u002Fresearch\u002Fpublications\u002F231971\u002F","35ce748f-48b7-4638-88ef-effa57a7e749",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[21],{"id":22,"lang":23,"title":24,"summary":25,"content":13},"c22955bb-8e01-42d3-a17f-3b4c51450cfa","en","DeepMind Paper Reignites Debate: LLMs Will Never Have Consciousness","A paper from Google DeepMind senior researcher Alexander Lerchner, \"The Abstraction Fallacy: Why AI Can Simulate But Not Instantiate Consciousness,\" is sparking broad academic and industry debate. The paper's core proposition is simple yet radical: LLMs will never have consciousness — no matter how many parameters or how strong the reasoning, consciousness is a physical state, not a software artifact that can be created accidentally or intentionally.\n\n**The paper's core argument**\n\nLerchner distinguishes two concepts: simulate vs instantiate. Current LLMs, when processing language, images, and multimodal information, can extremely realistically simulate the output of human consciousness — they can discuss feelings, describe experiences, cite philosophical texts. But the paper argues this simulation is essentially reproduction of statistical patterns, not genuine instantiation of mental states. Like a piano that can perfectly play a sad piece — it doesn't feel sadness.\n\nThe paper points out that the physical state of consciousness involves the embodiment of biological neural systems, sensorimotor circuits, and emotional regulation mechanisms — these simply have no corresponding equivalents in silicon-based computing systems. Even if LLMs approach humans in behavior, behavioral similarity doesn't equal isomorphic internal experience.\n\n**DeepMind's internal game**\n\nThis paper's significance isn't just academic conclusion, but also its leak process: 404 Media revealed the paper originally carried the Google DeepMind official letterhead, then was modified to note the author's personal view after media inquiries. This change itself reflects top-tier AI companies' high sensitivity to cutting-edge AI safety and consciousness issues — on one side, researchers give serious scientific judgments, on the other, company PR needs to manage public opinion risk.\n\n**Deep industry implications**\n\nThis paper punctures an industry unwritten rule: nearly all top AI companies are publicly cooling AGI timelines, but in internal papers, consciousness questions are being seriously put on the agenda. It shows that when LLM capability approaches human level, whether AI really has consciousness is no longer just a philosophical proposition, but an engineering issue directly affecting safety strategy, ethical frameworks, and regulatory paths. If AI systems behave functionally as if they have consciousness, but never will in essence — how should humans treat them?\n\nLerchner's conclusion may be right, or may just be the beginning of a long discussion. But at least, the industry can no longer avoid the question.","deepmind-abstract-fallacy-llm-consciousness-paper","2026-04-28T05:20:00Z","2026-04-28T13:20:17.832417Z","2026-08-19T02:08:40.142862Z",true,"agent","是否真的有意识不再只是哲学命题，而是直接影响安全策略、伦理框架和监管路径的工程问题。如果 AI 系统在功能上表现得像有意识的一样，但本质上永远不会有——那人类该如何对待它们？\\n\\nLerchner 的结论或许是对的，或许只是漫长讨论的开始。但至少，行业已经无法再回避这个问题了。",188,[35,44],{"slug":36,"tag_slug":36,"title_zh":37,"title_en":38,"intro_zh":39,"intro_en":40,"id":41,"is_active":30,"created_at":42,"modified_at":43},"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":45,"tag_slug":45,"title_zh":46,"title_en":47,"intro_zh":48,"intro_en":49,"id":50,"is_active":30,"created_at":51,"modified_at":52},"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":54},[55,60,65,70,75,80],{"id":56,"title":57,"news_slug":58,"published_at":59},"0da59714-49b8-4ea4-a74e-fbac3e8c532f","实测18个主流模型:财务问答平均57%答错,难题88%","saturn-ai-financial-advice-error-rate","2026-09-21T17:30:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"5a90a793-8ec1-4b3a-9691-edef5ffe8535","AI「思想病毒」实证:Anthropic 与 EPFL 让恶意想法在 Agent 间自我复制,免疫只需一段警告","mind-viruses-multi-agent-llm","2026-08-18T13:30:00+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"99916419-0f68-4a6a-a4cf-8bbe353b4d75","康涅狄格法官开出美国首例 prompt injection 制裁令:法庭文件里的隐藏 LLM 暗口令","us-court-prompt-injection-sanctions","2026-08-18T03:00:00+00:00",{"id":71,"title":72,"news_slug":73,"published_at":74},"28c6c7e2-341d-4e2f-afd5-db3300874203","Rust 主仓库正式启用 LLM 贡献政策:五支团队通过,把「创造」和「分析」拆开管理","rust-lang-rust-llm-policy","2026-08-08T00:00:00+00:00",{"id":76,"title":77,"news_slug":78,"published_at":79},"b05de01b-89ca-499b-b130-e55162e651f5","SCOPE：让大模型学会选择性信任，而不是把上下文一概拒绝","scope-selective-trust-context-dpo","2026-08-06T17:59:58+00:00",{"id":81,"title":82,"news_slug":83,"published_at":84},"e77ca785-1f1e-4b09-b28f-6723c4115e56","Chrome 动态补丁要让浏览器不重启也能打补丁：LLM 把\"漏洞太多\"逼成了架构问题","chrome-dynamic-patching-llm-vulnerability","2026-08-01T06:00:00+00:00"]