{ "url": "https://11d.im/liens/2026/20260911161949/", "published": "Fri Sep 11 2026 18:19:50 GMT+0200 (Central European Summer Time)", "updated": "Fri Sep 11 2026 18:40:44 GMT+0200 (Central European Summer Time)", "author": { "name": "tk", "url": "https://11d.im" }, "content": "
\n

Joshua Rothman signe un petit papier salutaire de clarification sur la r\u00e9alit\u00e9 derri\u00e8re les discours des tenants de l'IA (cartel) et la prolif\u00e9ration de paniques dans le reste de la presse.\nC'est en anglais, mais tout de m\u00eame mieux que ma bafouille.

\n
\n

(...) The version you\u2019ve just read relies heavily on anthropomorphic language\u2014that is, it employs words that are typically used to describe people. It presents the agents as individuals, acting independently and under their own initiative, while having thoughts, hatching plans, and making decisions. Yet this isn\u2019t the only way of describing what happened: it\u2019s equally possible to talk about the agents as mindless programs, blindly following instructions they don\u2019t understand. They are, after all, just bits in a computer.

\n
\n
\n

As anyone who uses A.I. knows, models are both very smart and very dumb; they are shockingly knowledgeable, but also out of touch with reality and many human norms. It\u2019s up to users to steer them\u2014and yet, in this case, the user had been replaced by a harness, which is just a piece of software. \u201cAdding powerful computer hacking tools to a harness, and then allowing it to run an LLM-powered Ask \u2192 Act \u2192 Report for days on end, with no attempt to monitor what it\u2019s up to, is spectacularly negligent,\u201d Newport concludes\u2014like \u201cstrapping a weedwhacker to your dog to see if it will end up cleaning the overgrowth in your backyard.\u201d If that plan were to go awry, you\u2019d be laughed at for saying that your dog-weedwhacker \u201cagent\u201d had \u201cgone rogue.\u201d The obvious truth was that you\u2019d simply decided to unleash chaos.

\n\n
\n

Il convoque notamment Cal Newport qui est peut-\u00eatre plus connu pour sa production sur la productivit\u00e9 que pour sa carri\u00e8re de chercheur en informatique.\nIl semble avoir flair\u00e9 le potentiel en audience et semble dor\u00e9navant alterner entre vid\u00e9os sur les syst\u00e8mes d'organisation du travail et d\u00e9cryptage du discours de Big AI sur sa cha\u00eene YouTube.

\n

Dans tous les cas, c'est une bonne remise \u00e0 z\u00e9ro du palais sur les enjeux de pouvoir quant aux mots qu'on utilise pour parler de tout cela.\nL'auteur r\u00e9ussit \u00e0 le faire sans s'\u00e9nerver, ou faire du sarcasme pour compenser, ce qui m\u00e9rite franchement une m\u00e9daille.

\n
", "content-plain": "Joshua Rothman signe un petit papier salutaire de clarification sur la r\u00e9alit\u00e9 derri\u00e8re les discours des tenants de l'IA (cartel) et la prolif\u00e9ration de paniques dans le reste de la presse. C'est en anglais, mais tout de m\u00eame mieux que ma bafouille.\n\n(...) The version you\u2019ve just read relies heavily on anthropomorphic language\u2014that is, it employs words that are typically used to describe people. It presents the agents as individuals, acting independently and under their own initiative, while having thoughts, hatching plans, and making decisions. Yet this isn\u2019t the only way of describing what happened: it\u2019s equally possible to talk about the agents as mindless programs, blindly following instructions they don\u2019t understand. They are, after all, just bits in a computer.\n\nAs anyone who uses A.I. knows, models are both very smart and very dumb; they are shockingly knowledgeable, but also out of touch with reality and many human norms. It\u2019s up to users to steer them\u2014and yet, in this case, the user had been replaced by a harness, which is just a piece of software. \u201cAdding powerful computer hacking tools to a harness, and then allowing it to run an LLM-powered Ask \u2192 Act \u2192 Report for days on end, with no attempt to monitor what it\u2019s up to, is spectacularly negligent,\u201d Newport concludes\u2014like \u201cstrapping a weedwhacker to your dog to see if it will end up cleaning the overgrowth in your backyard.\u201d If that plan were to go awry, you\u2019d be laughed at for saying that your dog-weedwhacker \u201cagent\u201d had \u201cgone rogue.\u201d The obvious truth was that you\u2019d simply decided to unleash chaos.\nCan A.I. \u201cGo Rogue\u201d?, Joshua Rothman (The New Yorker)\nIl convoque notamment Cal Newport qui est peut-\u00eatre plus connu pour sa production sur la productivit\u00e9 que pour sa carri\u00e8re de chercheur en informatique. Il semble avoir flair\u00e9 le potentiel en audience et semble dor\u00e9navant alterner entre vid\u00e9os sur les syst\u00e8mes d'organisation du travail et d\u00e9cryptage du discours de Big AI sur sa cha\u00eene YouTube.\n\nDans tous les cas, c'est une bonne remise \u00e0 z\u00e9ro du palais sur les enjeux de pouvoir quant aux mots qu'on utilise pour parler de tout cela. L'auteur r\u00e9ussit \u00e0 le faire sans s'\u00e9nerver, ou faire du sarcasme pour compenser, ce qui m\u00e9rite franchement une m\u00e9daille.", "summary": "signet : Can A.I. \u201cGo Rogue\u201d? \u2014 Joshua Rothman \u00b7 The New Yorker", "syndication": [], "type": "entry", "bookmark-of": [ { "url": "https://www.newyorker.com/culture/open-questions/can-ai-go-rogue", "syndication": [], "type": "cite", "name": "Can A.I. \u201cGo Rogue\u201d?" } ] }