[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-why-jensen-huang-is-wrong-about-agi-zh":3,"tags-why-jensen-huang-is-wrong-about-agi-zh":23,"related-lang-why-jensen-huang-is-wrong-about-agi-zh":24,"related-posts-why-jensen-huang-is-wrong-about-agi-zh":28,"series-industry-a02e5dd9-b4a1-4add-89bc-2dcff8214b38":64},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":10,"language":12,"translated_content":10,"views":13,"is_premium":14,"created_at":15,"updated_at":15,"cover_image":11,"published_at":16,"rewrite_status":17,"rewrite_error":10,"rewritten_from_id":18,"slug":19,"category":20,"related_article_id":21,"status":22,"google_indexed_at":10,"x_posted_at":10},"a02e5dd9-b4a1-4add-89bc-2dcff8214b38","Why Jensen Huang Is Wrong Abo…","\u003Cp>Jensen Huang is wrong: today’s AI systems are not AGI.\u003C\u002Fp>\u003Cp>把一個能寫程式、能摘要文件、能在某些基準上拿高分的系統，直接稱作通用人工智慧，是把商業里程碑誤當成科學結論。現在的模型確實很強，但它們仍會在長對話中丟失上下文、在壓力下幻覺、在多步推理中失穩，也會在一個場景裡表現驚人、換個場景就崩掉。這些不是小瑕疵，而是「還沒到 AGI」的證據。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>能力清單不等於整體智能。現在的 found\u003Ca href=\"\u002Fnews\u002Fopenai-chatgpt-images-2-0-launch-zh\">at\u003C\u002Fa>ion models 可以寫 code、回答專業問題、甚至幫人做產品原型，但這些勝利多半是局部的。它們證明模型在某些輸入分佈下很強，不證明它有一個穩定、統一的世界模型。OpenAI、\u003Ca href=\"\u002Fnews\u002Fanthropic-amazon-5gw-compute-claude-zh\">Anth\u003C\u002Fa>ropic、Google 的公開測試都反覆顯示同一件事：模型可能在數學題或 coding benchmark 上表現亮眼，卻在一致性檢查、跨輪次追蹤、或需要自我修正的任務裡失手。這不是「差一點」而已，而是能力沒有被整合成一個整體。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777077220933-v0sn.png\" alt=\"Why Jensen Huang Is Wrong Abo…\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>人類不會因為某人很會做一類題目，就說他已經具備一般智能。一般智能的標準，是能在陌生情境中轉移能力，而且不會一換規則就散架。2023 到 2024 年間，多個模型在 MMLU、HumanEval 這類測試上大幅進步，但同時也在長上下文、工具調用、和多步規劃上暴露脆弱性。這種「單點很強、整體不穩」的特徵，正是專才系統，不是通才系統。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>第二個問題在於，scale 是策略，不是心智理論。更大的模型、更多參數、更多算力，確實帶來更好的輸出品質。\u003Ca href=\"\u002Fnews\u002Fwhy-gpt-image-2-matters-more-than-another-ai-image-launch-zh\">GPT\u003C\u002Fa>-4、Claude 3、Gemini 1.5 這些模型都比前代更強，這點沒有人否認。但提升不等於質變。scale 可以放大模式匹配、記憶片段與語言流暢度，卻不會自動長出穩定抽象、因果控制、以及在不確定環境中的目標導向調適能力。把更多計算等同於更高層次的認知，是產業最常見、也最危險的偷換。\u003C\u002Fp>\u003Cp>現實中的失敗模式已經很清楚。即使是最強模型，幻覺仍然存在，長文檔推理仍然不穩，工具使用仍需要精心設計的外部框架。以 2024 年多家團隊的評測來看，模型在長上下文任務中的可靠性會隨著序列拉長而下降，並不會因為上下文窗口變大就自動解決。若真是 AGI，這些應該只是邊角問題；但現在它們是結構性問題。模型更像是在更會產生答案，而不是更會理解問題。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>最強的反對意見是：AGI 應該按結果定義，而不是按哲學定義。只要一個系統能跨足夠多的經濟任務工作，能從回饋中改進，能在高層工作中持續產生價值，那它就已經足夠「通用」了。從這個角度看，AGI 不是神秘門檻，而是移動目標；今天的 frontier models 已經夠接近，所以 Huang 的說法只是提前命名。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777077220858-xuyj.png\" alt=\"Why Jensen Huang Is Wrong Abo…\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這個說法不是空話。科技產品本來就是先被採用，再被定義。沒有人等到哲學共識形成才承認平台改變了世界。若 AI 已經能寫作、編碼、搜尋、協作、規劃，從產品和市場角度看，它確實像一個通用系統。Huang 講的是部署語言，不是神經科學語言。\u003C\u002Fp>\u003Cp>但這仍然不足以把它叫做 AGI，因為有用不等於一般智能。系統可以商業價值極高，卻仍缺乏真正通用認知所需的整合式、自我穩定結構。這不是文字遊戲，而是功能差異。現在的模型不會可靠地維持跨時間的意圖，不會穩健地修補自己的錯誤，也沒有展示出足以支撐 AGI 標籤的統一控制能力。把它們現在就叫作 AGI，只會把科學名詞降格成行銷獎章。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程師，別再用五分鐘 demo 當作能力證明，改去測長鏈路一致性、未知任務轉移、和壓力下的失敗模式。如果你是 PM，別因為模型在簡報裡看起來很廣，就把 AGI 當成對外敘事。如果你是創辦人，產品要建立在現有模型真正擅長的事上，而不是建立在「scale 已經解決認知」的幻想上。下一次突破，會來自更好的記憶、更好的控制、更好的整合，而不是單純更大的模型。在那之前，請把它們叫作它們真正的名字：強大，但還不是 AGI。","Jensen Huang is wrong: today’s AI systems are not AGI, and calling them that confuses benchmark wins, business value, and genuine general intelligence.","qubic.org","https:\u002F\u002Fqubic.org\u002Fblog-detail\u002Fjensen-huang-agi-achieved-artificial-general-intelligence-analysis",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777077220933-v0sn.png","zh",0,false,"2026-04-25T00:33:24.610192+00:00","2026-04-25T00:33:24.414+00:00","done","c40b5c1f-94ec-4558-8908-57f58f6baf7e","why-jensen-huang-is-wrong-about-agi-zh","industry","0a8dd623-df76-495d-bb24-d96240c8e0a5","published",[],{"id":21,"slug":25,"title":26,"language":27},"why-jensen-huang-is-wrong-about-agi-en","Why Jensen Huang Is Wrong About AGI","en",[29,35,40,46,52,58],{"id":30,"slug":31,"title":32,"cover_image":33,"image_url":33,"created_at":34,"category":20},"2584ed18-c8c7-4739-85fd-f7807b99c505","jensen-huang-ai-warning-coworker-productivity-zh","黃仁勳的 AI 警告，其實在講同事","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777080116173-8gfg.png","2026-04-25T01:21:35.402019+00:00",{"id":36,"slug":37,"title":5,"cover_image":38,"image_url":38,"created_at":39,"category":20},"53fcc0da-62a4-4863-a169-d4ba7cb49db2","why-jensen-huang-is-wrong-about-agi-being-achieved-zh","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777079923343-wyz8.png","2026-04-25T01:18:21.319991+00:00",{"id":41,"slug":42,"title":43,"cover_image":44,"image_url":44,"created_at":45,"category":20},"60108039-d66c-4796-a8e3-2a0534daef09","why-gpt-image-2-matters-more-than-another-ai-image-launch-zh","為什麼 GPT Image 2 比另一個 AI 圖像發布更…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777032939651-hlt2.png","2026-04-24T12:15:23.628035+00:00",{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":20},"259a408a-3045-47ef-ae32-1a2d7b76233e","anthropic-amazon-5gw-compute-claude-zh","Anthropic 與 Amazon 鎖定 5GW 算力","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777032245147-w0wg.png","2026-04-24T12:03:39.154665+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":20},"80bc00cb-e02a-44c5-a364-b5aaf013adaf","why-enterprises-should-stop-treating-codex-like-a-pilot-proj-zh","為什麼企業應該停止把 Codex 當成試點專案","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776989560691-7bfn.png","2026-04-24T00:12:24.005304+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":20},"061c7d61-98fd-4192-b1ac-37f0f7b412df","why-the-mythos-rollout-is-a-mistake-zh","Why the Mythos rollout is a m…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776989193541-gbq8.png","2026-04-24T00:06:18.457286+00:00",[65,70,75,80,85,90,95,100,105,110],{"id":66,"slug":67,"title":68,"created_at":69},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":71,"slug":72,"title":73,"created_at":74},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":76,"slug":77,"title":78,"created_at":79},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]