Chief AI Scientist. No change since 2018.
@ylecun
-
Superintelligence vs AGI: Redefining Long-Term AI Goals
By
–
Artificial Superintelligence has always made sense as an aspiration and long-term goal.
It has always been FAIR's long-term goal (as well as mine). It still is, now more than ever. Artificial General Intelligence never made sense: it's meant to designate "human-level AI", but -
AI Impact: Six Key Lessons from Newsweek Coverage
By
–
AI on the cover of Newsweek https://
newsweek.com/2025/07/04/ai-
impact-six-lessons-2088669.html
… -
JEPA: Latent-Space Reasoning and Planning Architecture
By
–
JEPA is designed for latent-space reasoning and planning.
-
SandboxAQ Releases Awesome New Dataset
By
–
Awesome new dataset from @SandboxAQ https://t.co/PtmSmM5W6H
— Yann LeCun (@ylecun) 22 juin 2025Awesome new dataset from @SandboxAQ
-
RL Limitations and AGI Pursuit: DeepMind Strategic Challenges
By
–
No. I told them two things.
1. The RL-based pursuit of AGI will not work. The sample complexity of RL is just too high (that was the motivation for "the cake analogy").
2. Pursuing AGI is cool, but at some point, Google will ask DeppMind to work on topics that could have a -
SandboxAQ Releases Awesome New Dataset for AI Research
By
–
Awesome new dataset from @SandboxAQ https://t.co/PtmSmM5W6H
— Yann LeCun (@ylecun) 20 juin 2025Awesome new dataset from @SandboxAQ
-
Fine-tuning data scarcity limits model improvement potential
By
–
It would if we had infinite fine-tuning data.
But we most definitely don't. -
Human Reasoning Beyond Token Prediction: Mental Models
By
–
Human reasoning is not based on auto-regressive discrete symbol (token) prediction.
It is based on the manipulation of mental models in continuous representations spaces.
It is based on *searching* for a set of manipulations of this model to arrive at a particular result.