Charbel-Raphael (@CRSegerie) / X

Charbel-Raphael

2,796 posts

Charbel-Raphael

@CRSegerie

Director of CeSIA, OECD AI Expert Co-lead, Global Call for AI Red Lines, featured in 300 media mentions, including Le Monde, NYT, NBC Co-author, AI Safety Atlas

Ile-de-France, France

Joined November 2018

Pinned
Charbel-Raphael
@CRSegerie
Sep 22, 2025
The time for AI self-regulation is over. 200 Nobel laureates, former heads of state, and industry experts just signed a statement: "We urgently call for international red lines to prevent unacceptable AI risks" The call was presented at the UN General Assembly today by Maria
00:00
486K
Charbel-Raphael
@CRSegerie
Sep 1, 2024
The prompt used in the paper 'People Cannot Distinguish GPT-4 from a Human in a Turing Test' is quite revealing about humans. tldr: "Be dumb"
293K
Charbel-Raphael
@CRSegerie
May 8, 2023
What does superintelligence mean? Here are four properties: Speed - Humans communicate at a rate of two words per second, while GPT4 can process 32k words (50 pages) in an instant - superintelligences will be able to communicate gigabytes per second. Once GPTs can write "better"
196K
Charbel-Raphael
@CRSegerie
Oct 12, 2024
Hello à toutes et à tous, je suis Charbel-Raphael Segerie, directeur du Centre pour la Sécurité de l'IA. Très heureux d'avoir pu collaborer avec @ego_v1 sur le script de cette vidéo ! Retrouvez moi ce soir à 20h pour une FAQ ici, sur mon twitter !
74K
Charbel-Raphael
@CRSegerie
Aug 11, 2025
Je parle ici en tant que directeur du Centre pour la Sécurité de l’IA, et je confirme : cela fait des années que nous regardons Luc Julia raconter des fumisteries. Cela est d’autant plus préoccupant qu’il a une voix qui porte et qu’il a fait partie de la commission sur l’IA
MrPhi
@MonsieurPhi
Aug 11, 2025
NOUVELLE VIDEO ! Je décortique le cas Luc Julia, le réputé co-créateur de Siri et expert mondial de l'IA, encensé dans les médias et récemment auditionné au Sénat. youtu.be/e5kDHL-nnh4 youtu.be/e5kDHL-nnh4 youtu.be/e5kDHL-nnh4 Le résultat est salé mais nécessaire.
44K
Charbel-Raphael
@CRSegerie
Dec 12, 2024
My pDoom increased a bit during the last week. 🧵
57K
Charbel-Raphael
@CRSegerie
Jan 5, 2024
This post from @webmasterdave is still legendary Context: Eliezer sets out his theory of consciousness in a Facebook post, followed by a debate in the comments about animal consciousness. It's like the Super Bowl.
46K
Charbel-Raphael
@CRSegerie
Jun 21, 2022
My intellectual journey in AI safety. Have a good day
Charbel-Raphael
@CRSegerie
Oct 23, 2022
If it is really possible for AIs to surpass us as much in mathematics as AlphaGo surpasses us in go, what is the point of doing mathematics?
Charbel-Raphael
@CRSegerie
Apr 12, 2025
Boom! 4 million views on this 40-min video explaining the risks of superintelligence in French. Very proud to have contributed to the scientific advising for it. I think advising successful YouTubers is a fairly neglected strategy! 🧵
7.3K
Charbel-Raphael
@CRSegerie
Dec 7, 2023
I think I disagree, In a way that makes me think that Bayesian thinking will always be unsatisfactory. Imho, the hardest question for a bayesian is: "What's the space of different possibilities?". This is not the same question as "What's your prior?". Bayes' Rule Cannot Update
davidad 🎇
@davidad
Dec 7, 2023
The hardest question for a Bayesian is “where do your priors come from?” There are two pretty good answers: 1. Maximum entropy wrt Haar measure 2. Asymptotically doesn’t matter as long as it’s fully supported Sometimes, neither answer suffices. You need imprecise probability.
39K
Charbel-Raphael
@CRSegerie
Sep 18, 2024
A few tentative explanations for why France is so behind in AI Safety: 1. Yann LeCun 2. Cultural isolation: it's a large enough country to rely on itself. 3. Focus on autistic mathematical training: the selection process for the most important universities relies heavily on
25K
Charbel-Raphael
@CRSegerie
Apr 18, 2025
Props to @JonathanClaybr_ for having been the first to call out this worrying trend a few months ago.
Mechanize
@MechanizeWork
Apr 17, 2025
Today we’re announcing Mechanize, a startup focused on developing virtual work environments, benchmarks, and training data that will enable the full automation of the economy. We will achieve this by creating simulated environments and evaluations that capture the full scope of
5.2K
Charbel-Raphael
@CRSegerie
Mar 23, 2025
It's unclear if AI evaluations are really reducing AI risks. It might even be the case than evals are net negative Here's an extract of "Reasons to Doubt the Impact of AI Risk Evaluations" - from @gabemukobi
7.4K