Suivre
- Site webdavidscottkrueger.com
- Wikipédiaen.wikipedia.org
David Krueger
Professeur adjoint en IA à l’Université de Montréal, membre académique principal de Mila et fondateur de l’organisme sans but lucratif Evitable, à l’origine de la Statement on AI Risk de 2023, qui affirme qu’à défaut d’un changement de cap il est plus probable qu’improbable que l’humanité soit définitivement dépossédée de son pouvoir par une IA échappant à tout contrôle d’ici cinq ans, après quoi l’extinction ne sera plus qu’une question de temps.
Fonctions
- 2021 – 2024Assistant Professor in Machine Learning, University of CambridgeSource →
- 2023 – 2023Research director on the founding team, UK AI Security InstituteHis own page dates this only to the year — "In 2023, he was a research director on the founding team" — and prints the institute under its present name; it was founded as the UK AI Safety Institute and renamed in 2025.Source →
- 2024 – aujourd’huiAssistant Professor in Robust, Reasoning, and Responsible AI, University of MontrealAlso a Core Academic Member at Mila, the Quebec Artificial Intelligence Institute, holder of a CIFAR AI Chair and of the IVADO Professorship in Responsible AI. He trained in deep learning under Yoshua Bengio from 2013 to 2021 and is on leave from the faculty post for 2026. Both sources give the year only, not a month.Source →
- 2025 – aujourd’huiChief Executive Officer, EvitableThe nonprofit he founded to organise the public against societal-scale risks of AI. He initiated the 2023 Statement on AI Risk signed by Dario Amodei, Sam Altman and Geoffrey Hinton, all three of whom are on this register.Source →
Au registre
14 sept. 2026 · Entretien · CTV News
We’d have to try and shut down all of them, and that becomes very, very difficult … So, we have a very limited time window in which to act to prevent those kinds of things.
Chief Executive Officer, Evitable à l’époque
Said of a rogue system that has copied itself across many machines. The sentences either side of these are the reporter's paraphrase carrying one quoted fragment — that “we wouldn’t even know which computers to shut down” — so they are not stitched in here.
Human extinction ‘just a matter of time’ if AI slowdown doesn’t go far enough, researcher warns →
14 sept. 2026 · Entretien · CTV News
Imagine a world where all companies are run by AI, where there are robots all over the place, where basically humans don’t have any work except for nominal work, and even politicians and everyone is deferring to AI because the AI just seems to have such good judgment … That’s the kind of future that we’re headed towards.
Chief Executive Officer, Evitable à l’époque
His description of what permanent disempowerment would look like from inside. The ellipsis stands where the page prints an attribution tag (“,” he added. “).
Human extinction ‘just a matter of time’ if AI slowdown doesn’t go far enough, researcher warns →
14 sept. 2026 · Entretien · CTV News
The tests basically don’t work, and they’re going to get less and less reliable as the AI gets smarter, because the AI can basically pretend … It can play nice, it can play dumb, and it can fool our testing.
Chief Executive Officer, Evitable à l’époque
On the safety evaluations that Dario Amodei's proposal to pace the frontier would rely on, which Krueger judges inadequate. The ellipsis stands where the page prints an attribution tag.
Human extinction ‘just a matter of time’ if AI slowdown doesn’t go far enough, researcher warns →
14 sept. 2026 · Entretien · CTV News
If we don’t course correct, it’s more likely than not that humanity will be permanently disempowered from rogue AI within the next five years … And then human extinction at that point is just a matter of time.
Chief Executive Officer, Evitable à l’époque
Said after calling the more-than-one-in-ten odds of human extinction given the week before by an Anthropic safety researcher “way too low”. The ellipsis stands where the page prints an attribution tag (“,” he said. “), so the two halves are consecutive blocks of the same remark rather than one sentence. By disempowered, the page explains, he means a world in which AI is effectively running society.
Human extinction ‘just a matter of time’ if AI slowdown doesn’t go far enough, researcher warns →