The Godfather of Neural Networks Just Bet His Career That ChatGPT Is a Dead End

The man who invented the tech behind your phone's face recognition just quit the most powerful AI company in the world to prove everyone else is doing it wrong.

Yann LeCun left Meta in December after 12 years. Not to retire. Not to consult. To launch AMI Labs — a startup building AI that understands the physical world instead of predicting the next word in a sentence.

He's raising €500 million at a €3.5 billion valuation. Before the company has shipped a single product.

That's not hype. That's conviction.

The $3.5 Billion Disagreement

LeCun won the Turing Award in 2018 for inventing convolutional neural networks — the tech that lets computers see. Every face unlock, every self-driving car, every medical imaging AI owes something to his work in the 1980s.

He's spent years saying large language models like ChatGPT are useful but limited. That they seem smart in conversation but can't reason about physics, can't plan multiple steps ahead, can't learn how objects actually work in space.

At MIT last year he said it plainly: "Within three to five years, world models, not language models, will be the dominant AI architecture. Nobody in their right mind would use large language models of the type that we have today."

Meta kept doubling down on LLMs. Llama, Llama 2, Llama 3. Two new models codenamed "Mango" and "Avocado" shipping in 2026.

LeCun had almost nothing to do with any of them after the first Llama.

So he left.

What's a World Model?

LLMs learn from text. World models learn from video, spatial data, sensor inputs — everything that shows how things actually work.

LLMs can write you a poem about gravity. World models understand that if you drop a ball, it falls.

AMI Labs is building "systems that understand the physical world, have persistent memory, can reason, and plan complex action sequences," according to LeCun's LinkedIn announcement.

Translation: AI that doesn't just talk like it's smart — AI that can actually do things in the real world without breaking everything.

The first customer? Nabla, a health tech company. They'll get early access to AMI's world model tech to build FDA-certifiable medical AI.

Investors backing the bet include Cathay Innovation, Hiro Capital, HV Capital. LeCun himself was an early investor in Nabla.

The Industry Is "LLM-Pilled"

LeCun's departure wasn't quiet. At AI House Davos 2026, he called out the industry for becoming "LLM-pilled" — so convinced that scaling up text-based models will solve everything that alternative approaches get ignored.

Reddit users who've followed his work say the technical disagreement was simple: Meta bet hard on LLMs. LeCun's world model approach kept losing internal battles.

"You certainly don't tell a researcher like me what to do," LeCun said in a recent interview.

So he's building his own lab.

The Fork in the Road

Here's what's at stake.

If LeCun is right, the entire LLM industry — OpenAI, Anthropic, Google, Meta — is building on sand. Chatbots will plateau. The real breakthroughs will come from systems that learn like humans do: by interacting with the world, not by reading the internet.

If he's wrong, he just left one of the best-funded AI labs on Earth to chase an idea that the industry already decided doesn't scale.

Either way, the most credible dissenter in AI just went all-in.

The world models race is real. DeepMind has Genie. Nvidia has Cosmos. Meta has V-JEPA. OpenAI has Sora. Wayve raised $1.2 billion building self-driving systems without maps.

But none of them are betting their entire career on it.

LeCun is.

---

FAQ: Why did LeCun leave Meta?

He's spent years arguing LLMs can't reason about physics or plan complex actions. Meta kept investing in LLMs. He wanted to prove his alternative approach works.

What's the difference between LLMs and world models?

LLMs learn from text (predicting the next word). World models learn from video, spatial data, and sensors (predicting what happens next in the physical world).

Is this just hype?

LeCun won the Turing Award for foundational AI work. Investors are backing him at a $3.5B valuation before any product launch. The tech community is split — some think he's right, others think LLMs will scale to solve these problems.