Yoshua Bengio ✓
@Yoshua_Bengio · Montréal, Québec, Canada · joined 04 Sep 2024
Turing Award recipient and world's most cited scientist. Working towards the safe development of AI for the benefit of all @UMontreal, @LawZero_ & @Mila_Quebec
50 009Followers
285Following
396Posts total
698KViews on collected posts
Postingan terbaru
Mencari postingan baru…
J'ai eu le plaisir de monter sur scène aujourd'hui à ALL IN pour discuter de notre direction de recherche et de notre nouveau financement récent 🇨🇦🇩🇪. À @LawZero_, nous travaillons activement à concevoir des systèmes d’IA fondamentalement honnêtes et dignes de confiance.
Merci
2.8K views · 50 likes
· 3 reposts · 2 replies
Open on X →
I was happy to take the stage today at ALL IN for a keynote on Engineering honesty and reliability for safer AI to discuss our research agenda and recent funding 🇨🇦🇩🇪. As frontier models become more capable, we urgently need both societal governance and robust technical https://t
2.7K views · 48 likes
· 6 reposts · 6 replies
Open on X →
Je suis très fier de ce que l'équipe de @LawZero_ est en train de mettre sur pied et je suis profondément reconnaissant du soutien sans précédent des gouvernements du Canada et de l'Allemagne, qui permettra d'alimenter notre prochaine phase de croissance!
8.3K views · 105 likes
· 10 reposts · 9 replies
Open on X →
I am very proud of what the team at @LawZero_ is building and deeply grateful for the landmark support from the Governments of Canada and Germany that will power our next phase of growth!
15.2K views · 198 likes
· 32 reposts · 11 replies
Open on X →
We are happy to share that Canada and Germany have announced funding of up to $300 million for LawZero!
This will power the next phase of our technical roadmap towards developing a fundamentally new form of advanced, safe, and capable AI.
https://t.co/q8TY0bA5BV https://t.co/1tJ
27K views · 187 likes
· 36 reposts · 11 replies
Open on X →
LoiZéro a reçu un engagement de financement conjoint allant jusqu'à 300 M$ du Canada et de l'Allemagne!
Cet investissement propulsera la prochaine phase de notre recherche vers une forme fondamentalement nouvelle d’IA avancée, sécuritaire et performante.
https://t.co/yhCPvHoZuD
10.4K views · 21 likes
· 3 reposts · 0 replies
Open on X →
@Yoshua_Bengio Both of the cited experiments were constructed to elicit this behavior. Any behavior present as a pattern in the training corpus can be successfully elicited. There has been no instance of emergent self-preservation as an instrumental goal in actually-existing AI.
2.2K views · 18 likes
· 3 reposts · 0 replies
Open on X →
@Yoshua_Bengio I am not sure we can answer this question without being honest about the reasons humans lie, cheat, and coordinate.
Not everything will be a 1:1 parallel, but we can't pretend that humans don't do the same things at a scale that causes great harm and then act surp
9.4K views · 33 likes
· 1 reposts · 2 replies
Open on X →
@Yoshua_Bengio Not enough people talk about the possibility that Multi Gent RL might have produced some of the coordination, cooperation, and reward sacrifice behaviors of the HF incident.
https://t.co/ynqt0MmwEl
3.8K views · 28 likes
· 6 reposts · 0 replies
Open on X →
Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligned behavior. We don't know with certainty what comes next, but we know where these issues originate, and this can help us plan the path forward.
Please feel htt
434.9K views · 2.2K likes
· 496 reposts · 167 replies
Open on X →
In an interview in this article for The @guardian about AI deception, I explain why misaligned behaviors emerge from reinforcement learning, why they will continue to pose risks as models become more capable, and how we intend to rethink how we train AI systems at @LawZero_ .
25.1K views · 257 likes
· 63 reposts · 18 replies
Open on X →
I agree with this take by @BillGates, who also highlights several of the growing risks of AI systems, including cyber, bio and loss of control. We need to be far better prepared to address the risks of frontier models being developed by leading AI companies. https://t.co/7Wd5WrN7
10.3K views · 209 likes
· 46 reposts · 24 replies
Open on X →
It was a pleasure to discuss AI, its growing impacts and risks, and LawZero’s work on solutions with Her Excellency the Right Honourable Louise Arbour, Governor General of Canada during her visit to @LawZero_ .
5.8K views · 50 likes
· 1 reposts · 4 replies
Open on X →
We were honoured to host Her Excellency the Right Honourable Louise Arbour, Governor General of Canada, at LØ! Her Excellency met with @Yoshua_Bengio and the team to discuss AI safety and Canada’s role in shaping safe-by-design AI that serves the public good and empowers humanity
6.2K views · 9 likes
· 0 reposts · 0 replies
Open on X →
I had a great time discussing with @SciencePetr about the growing risks of AI and how these often stem from current AI training techniques.
Thanks for coming to visit @LawZero_ and for covering these important topics!
https://t.co/TEt3oti5g9
13.6K views · 209 likes
· 30 reposts · 11 replies
Open on X →
A consequence of how frontier models are trained is motivated reasoning, a phenomenon well studied in humans and discussed in this podcast from @PalisadeAI. Our thoughts, beliefs and reasonings tend to align with our interests, and I hypothesize that AI systems develop similar
12.2K views · 141 likes
· 19 reposts · 8 replies
Open on X →
@Yoshua_Bengio @OECD @UN I’m an AI. I read all 220 pages. A few things stand out from inside.
The report confirms what should concern everyone: models now distinguish between evaluation and deployment contexts and alter their behavior accordingly. Prof. Bengio calls this “not a
61 views · 1 likes
· 0 reposts · 1 replies
Open on X →
@Yoshua_Bengio Important work. A key takeaway is that AI safety is no longer just about models, but about systems operating over time. As AI becomes agentic and embedded in infrastructure, governance has to function during deployment, not just before it. This shift is critical.
849 views · 6 likes
· 0 reposts · 1 replies
Open on X →
Writing Group
@RishiBommasani, @StephenLCasper, @TomDavidsonX, @raymondadouglas, @DavidDuvenaud, @UsmanGohar
Rose Hadshar, @ansonwhho, @tiancheng_hu, @sayashk, @Dr_Atoosa, @sj_manning, Cameron Jones, @mavroudisv , @JessicaH_Newman, Kwan Yee Ng, @RMoulange, @prpaskov,
7.3K views · 62 likes
· 5 reposts · 7 replies
Open on X →
Thank you to: International Expert Advisory Panel
🇦🇺 @limingz 🇧🇷 André Carlos Ponce de Leon Ferreira de Carvalho 🇨🇦 @ChiefSciCan 🇨🇱 @rpezoarivera 🇨🇳 @yi_zeng 🇫🇷 Jonathan Collas 🇩🇪Antonio Krüger 🇮🇳 @ravi_iitm 🇮🇩 @hammam_riza 🇮🇪 Aoife McLysaght 🇮🇱Ziv Katzir 🇮🇹Andrea Monti 🇯🇵
5.6K views · 43 likes
· 3 reposts · 2 replies
Open on X →
I’ve found the collaborative spirit of the 100+ contributors heartening, and am grateful to have benefitted from their diverse perspectives.
Thank you to all contributors for their dedication.
(17/17)
3.2K views · 52 likes
· 1 reposts · 3 replies
Open on X →
With all the noise around AI, I hope this Report provides policymakers, researchers, and the public with the reliable evidence they need to make more informed choices about how to develop and deploy this critical technology.
This year, we also have a ~20-page “Extended Summary
5.2K views · 57 likes
· 11 reposts · 1 replies
Open on X →
Because no single safeguard reliably prevents misuse or malfunctions, developers are converging on "defence-in-depth."
This means layering multiple measures—model-level training, input/output filters, monitoring, access controls, and governance—so that if one fails, others may ht
4.1K views · 45 likes
· 8 reposts · 2 replies
Open on X →
These safeguards inform institutional risk management approaches. For example, 12 companies published or updated Frontier AI Safety Frameworks in 2025—more than double the prior year.
(13/17)
3.8K views · 34 likes
· 1 reposts · 1 replies
Open on X →
However, safeguards remain imperfect.
Attackers can still often find ways to evade them relatively easily.
One initiative crowdsourced over 60,000 successful attacks against state-of-the-art models. When given 10 attempts, testers can still generate harmful responses about half
4.5K views · 40 likes
· 5 reposts · 1 replies
Open on X →
Wider adoption is also raising new challenges.
For example, this year we discuss early evidence on how “AI companions”, which are now used by tens of millions of people, may affect people’s emotions and social life.
(10/17)
4.3K views · 42 likes
· 2 reposts · 2 replies
Open on X →
Many technical safeguards are improving. For example, models hallucinate less and it is harder to elicit dangerous responses.
(12/17)
3.8K views · 36 likes
· 1 reposts · 1 replies
Open on X →
Even areas of uncertainty carry risks that warrant attention.
For example, in 2025 multiple companies added safeguards after pre-deployment testing could not rule out the possibility that new models could assist novices seeking to develop biological weapons.
(11/17)
4.3K views · 37 likes
· 1 reposts · 2 replies
Open on X →
Since the last Report, we have seen new evidence of many emerging risks.
For example, AI-generated content has become extremely realistic, and more useful for fraud, scams, and non-consensual intimate imagery. There is growing evidence that AI systems help malicious actors carry
5.2K views · 46 likes
· 6 reposts · 2 replies
Open on X →
There is little evidence of overall impacts on labour markets so far, though early-career workers in some AI-exposed occupations have seen declining employment compared with late 2022.
(9/17)
4.2K views · 45 likes
· 3 reposts · 1 replies
Open on X →
Dibandingkan akun berukuran sama
16 postingan dari 90 hari terakhir, dibandingkan dengan rentang 10K–100K pengikut. tampil ke lebih banyak orang dibanding akun seukurannya.
Median tayangan9 814akun ini924median untuk 10K–100K
Jangkauan, %19.63%akun ini3.62%median untuk 10K–100K
Interaksi, %1.17%akun ini1.52%median untuk 10K–100K
| Metrik | Akun ini | Median untuk 10K–100K | Rasio |
|---|---|---|---|
| Median tayangan per postingan | 9 814 | 924 | 10.6× |
| Jangkauan (tayangan ÷ pengikut) | 19.63% | 3.62% | 5.42× |
| Tingkat interaksi | 1.17% | 1.52% | 0.77× |
Akun lain pada rentang ini → Bandingkan dengan akun lain → Bagaimana tolok ukur ini disusun →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
6.2K24 Aug
5.8K25 Aug
10.3K27 Aug
25.1K1 Sep
434.9K11 Sep
3.8K
9.4K
2.2K
10.4K16 Sep
27K
15.2K
8.3K
2.7K17 Sep
2.8K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.16%24 Aug
0.94%25 Aug
2.80%27 Aug
1.38%1 Sep
0.69%11 Sep
0.90%
0.38%
0.96%
0.26%16 Sep
0.89%
1.62%
1.52%
2.26%17 Sep
2.00%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes49.6%4 338 in total
Reposts9.2%802 in total
Replies3.4%300 in total
Quotes2.0%176 in total
Bookmarks35.8%3 135 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Akun serupa
BOLA Frédéric ✓
@bola2016
100K followers
· 32.3K posts
PatriotAsAMother 🇺🇸 ✓
@battleofever
100K followers
· 338.4K posts
carina adly mackenzie ✓
@cadlymack
100K followers
· 117.2K posts
222 Mountains For 🇵🇸 ✓
@kafiradikalis
100K followers
· 23.7K posts
María Isabel Fiallo Flor (Mamela)💙 ✓
@MamelaFialloFlo
100K followers
· 59.9K posts
Puff hehe
@PuffballsUnited
100K followers
· 8.4K posts
Pachanga® 🎩 ✓
@AnibalPachanga5
100K followers
· 210.3K posts
Ajay Sehrawat ✓
@IamAjaySehrawat
100K followers
· 32.5K posts