Yann LeCun, Meta’s former chief AI scientist and Turing Award winner, released a new architecture called VL-JEPA (Vision-Language JEPA) that doesn’t think in words at all.
Instead of generating text token by token like ChatGPT, VL-JEPA predicts meaning directly building internal world models from vision and time.
Why this matters:
• AI no longer needs language to think
• Meaning-based reasoning scales better than text generation
• Robots and autonomous agents gain real world understanding
• The path to AGI may not be language-first
VL-JEPA uses a continuous meaning space, learns faster with fewer parameters, and shows how intelligence can emerge without narrating reality frame by frame.
This isn’t a feature upgrade.
It’s a fork in the road for the entire AI industry.


Who Do You Want To Be Today?
Incognito
Seller
Showoff
Batman
Superman
Spider Man
James Bond
Einstein
Donald Trump
Elon Musk
Jeff Bezos
Plaza Exotica
Pizzeria
Sushi Resto
Slagerij
Log In