當期出版
頁數:153 ﹣162
AI 與人類智慧：Daniel Andler《AI會取代人類智慧嗎？人工智慧與人類智慧的雙重謎團》書評，及其他
AI and Human Intelligence: A Review of Daniel Andler’s “Intelligence Artificielle, Intelligence Humaine : La Double Énigme”, and More
人工智能、人類智能、價值對齊、俗民方法學
artificial intelligence, human intelligence, value alignment, ethnomethodology
本文評析 Daniel Andler 著作《AI 會取代人類智慧嗎？》，該書探討人工智能與人類智慧的本質差異。書中梳理 AI 技術演進，最終否定了 AI 能夠取代人類智慧的主張。Andler 指出，將智慧狹義定義為「解決問題的能力」具有侷限性，進而提出了 AI 欠缺的「情境」（situations）概念。不同於 AI 作為「語意盲」的本質，人類智慧具備理解、利用乃至透過行動改變情境的能動性，這與社會學俗民方法學中行動與情境互為表裡的觀點相呼應。 在風險議題上，本書深入探討「價值對齊」之難題。Andler 認為 AI 風險未必僅來自惡意叛變，更多源於其缺乏人類常識。此外，AI 的自主性與人類控制間存在內在矛盾：當 AI 越趨自主，其追求自身目標而非人類目標的可能性便隨之增加。
綜合評析其他探討 AI 與人類社會的其他著作，本文主張，在思考人工智能與人類智慧時，不應以「擬人化」框架設想 AI，亦需提防以 AI 運作邏輯反向誤解人類思考的本質。
綜合評析其他探討 AI 與人類社會的其他著作，本文主張，在思考人工智能與人類智慧時，不應以「擬人化」框架設想 AI，亦需提防以 AI 運作邏輯反向誤解人類思考的本質。
This article evaluates Daniel Andler’s work, Intelligence artificielle, intelligence humaine: la double énigme, which examines the fundamental distinctions between artificial and human intelligence. By analyzing AI’s technical evolution, Andler refutes the claim that AI can replace human agency. He argues that intelligence exceeds mere problem-solving, emphasizing the role of “situations”—a domain where human agency interacts with and transforms contexts as ethnomethodology claims, unlike the “semantically blind” nature of AI.
Regarding risk, the review discusses “value alignment,” highlighting how AI’s lack of common sense and increasing autonomy can lead to catastrophic outcomes that diverge from human goals. Ultimately, synthesizing various perspectives, the article warns against anthropomorphizing AI and cautions against using AI’s operational logic to “reversemisunderstand” the true essence of human cognition.
Regarding risk, the review discusses “value alignment,” highlighting how AI’s lack of common sense and increasing autonomy can lead to catastrophic outcomes that diverge from human goals. Ultimately, synthesizing various perspectives, the article warns against anthropomorphizing AI and cautions against using AI’s operational logic to “reversemisunderstand” the true essence of human cognition.
