Yann LeCun has long been the loudest critical voice against the large language model approach to artificial general intelligence. Now, backed by over a billion dollars, he has the resources to prove his alternative vision right.
His new venture, Advanced Machine Intelligence (AMI) Labs, has closed a $1.03 billion seed round with Nvidia and Bezos Expeditions among the backers — one of the largest seed rounds in AI history.
The World Models Thesis
AMI Labs is built around a fundamentally different architectural philosophy from the transformer-based models dominating the industry.
Rather than predicting the next token in a sequence of text, world models learn by developing an internal representation of physical reality — understanding cause and effect, spatial relationships, and the laws that govern how objects behave in the real world.
LeCun has argued for years that this grounded, embodied approach is necessary for AI to become truly intelligent rather than merely fluent. The application targets are concrete:
- Robotics — manipulation, navigation, and interaction in unstructured physical environments
- Manufacturing — autonomous quality control, assembly, and process optimization
- Physical simulation — generating accurate models of systems too complex or expensive to run in the real world
Why This Funding Round Matters
A $1.03B seed — even in the inflated AI investment environment of 2026 — signals serious conviction from sophisticated investors. Nvidia’s participation is particularly notable: the chip giant is betting on AMI’s architecture requiring the same compute-intensive training runs that have made Nvidia indispensable to the LLM era.
If LeCun’s world-model approach gains traction, it could redirect a meaningful portion of AI capital away from pure language modeling and toward physically grounded intelligence — a shift with enormous implications for both software and hardware investment.
Source: devflokers.com, buildez.ai
