Models & Researchyann lecunami labsworld modelslarge language models

Yann LeCun develops more flexible AI at AMI Labs

|July 2, 2026|By LDS Team

6.8

Relevance Score

Yann LeCun develops more flexible AI at AMI Labs — Photo: ichef.bbci.co.uk · rights & takedowns

Yann LeCun, the Turing Award-winning researcher who left Meta in 2025, has built AMI Labs into a Paris-based startup with over $1 billion in seed funding, and now argues today's leading AI chatbots cannot reach human-level intelligence. Speaking at VivaTech in Paris, LeCun told the BBC that ChatGPT, Claude, and Gemini "are not a path towards human level or human-like intelligence," and that AI still doesn't understand the physical world as well as a rat. AMI Labs, which raised its $1.03 billion round in March 2026 at a $3.5 billion valuation with backers including Nvidia and Jeff Bezos's Bezos Expeditions, is building "world models" using LeCun's Joint Embedding Predictive Architecture (JEPA) to give AI systems the physical and causal reasoning pure language models lack.

For AI practitioners, LeCun's world-model bet is a live test of whether embodied physical reasoning, not language scale, is the harder unsolved problem in AI. AMI Labs' Joint Embedding Predictive Architecture (JEPA) approach explicitly targets the causal, multi-outcome reasoning that trips up next-token predictors, and the scale of investor backing means the industry will get an unusually well-resourced answer within a few years.

What happened

The BBC reports that Yann LeCun, who left Meta in 2025 to found Advanced Machine Intelligence Labs (AMI Labs) in Paris, said current large language models cannot reach human-level intelligence. Speaking at VivaTech in Paris, LeCun told reporters that ChatGPT, Claude, and Gemini "are not a path towards human level or human-like intelligence" and that AI systems still don't understand the physical world "as well as a rat." AMI Labs previously announced a $1.03 billion seed round in March 2026, the largest seed round in European history, at a $3.5 billion pre-money valuation, with backers including Nvidia and the fund managing Jeff Bezos's private wealth, Bezos Expeditions.

Technical context

AMI Labs is built around JEPA, an architecture LeCun proposed in 2022 that learns abstract representations of the world and predicts how they evolve, rather than generating every pixel or token. LeCun argues models need this kind of anticipatory, causal reasoning to plan reliable actions - his recurring example of the uncertain way a dropped pen falls illustrates why statistical text prediction differs from the multi-outcome reasoning embodied agents need.

For practitioners

AMI Labs says it will publish research and open-source code as it develops JEPA-based systems. Its first disclosed commercial partner is Nabla, a digital health startup chaired by AMI Labs CEO Alexandre LeBrun. Teams working on robotics, simulation, or agentic planning should watch for published world-model benchmarks and training curricula, which could offer reusable tooling well before commercial products ship.

What to watch

LeBrun has said world models could take years to move from research to deployment, unlike a typical applied AI product. Given the funding involved and Nvidia's participation, expect competitive pressure from other world-model efforts, including Fei-Fei Li's World Labs, which has raised roughly $1 billion of its own.

Key Points

1AMI Labs, founded by Yann LeCun after he left Meta, has raised over $1 billion to build AI world models instead of large language models.
2LeCun argues LLMs lack physical and causal reasoning, telling the BBC that ChatGPT, Claude, and Gemini are not a path to human-level intelligence.
3Backing from Nvidia and Bezos Expeditions signals investor appetite for JEPA-based world models as an alternative bet on general AI.

Scoring Rationale

A prominent former Meta chief AI scientist publicly restating his case against LLMs at a major conference, backed by AMI Labs' already-reported $1 billion-plus round, is notable for research and infrastructure teams. It is not a fresh funding or product event since the round was first reported in March 2026, so it stays in the notable rather than major tier.

MoreAI Research news