It’s a story that’s been unfolding in the fast-paced world of artificial intelligence, a tale of a seasoned pioneer, a massive funding round, and a fundamental disagreement about the very path AI should take. Yann LeCun, a name synonymous with AI research, has just secured a staggering $1.03 billion in seed funding for his new venture, AMI Labs. This isn't just another funding announcement; it's a bold declaration that the current dominant approach to AI – the one powering chatbots like ChatGPT – might be fundamentally flawed.
LeCun, a Turing Award laureate and former chief AI scientist at Meta, spent over a decade at the social media giant, where he helped build its influential AI research lab, FAIR. He was instrumental in developing key technologies like convolutional neural networks (CNNs). However, as the world became captivated by Large Language Models (LLMs) like OpenAI's ChatGPT, Meta, under CEO Mark Zuckerberg, pivoted heavily towards this LLM-centric strategy. This included the release of their own open-source LLMs, Llama and its successors.
But LeCun saw things differently. He openly voiced his concerns, stating that LLMs, while impressive at language tasks, lack a true understanding of the physical world. "They can describe a chair, but they can't understand what it means to sit on a chair, to balance on a chair, or to catch a falling chair," he explained. To him, LLMs are essentially "statistical illusions," generating fluent text by predicting the next word, a far cry from genuine intelligence. He believes true AI needs to learn like humans and animals do – through perception and experience, grasping the causal relationships, spatial logic, and object permanence of the real world.
This divergence in vision led to LeCun's departure from Meta in late 2025. He felt he could achieve his goals more effectively outside the company, and interestingly, Meta has indicated a willingness to collaborate, with AMI Labs' technology potentially finding its way into products like Ray-Ban Meta smart glasses.
AMI Labs, whose name cleverly echoes the French word for "friend," is built on LeCun's concept of "World Models," specifically leveraging his Joint Embedding Predictive Architecture (JEPA). Unlike LLMs that focus on predicting the next word or pixel, JEPA learns abstract representations of the world, filtering out noise and predicting within this abstract space. Crucially, by incorporating 'actions' as input, these models can simulate outcomes and plan sequences of actions, offering a level of predictability and reliability that LLMs currently lack. This is particularly vital for high-stakes applications like medical diagnostics, industrial control, and autonomous driving, where the "hallucinations" of LLMs could have severe consequences.
The founding team at AMI Labs reads like an AI all-star roster, including former Meta colleagues and top researchers like Saining Xie, whose work on Diffusion Transformers (DiT) was foundational for OpenAI's Sora video generation model. The company's ambitious vision has attracted significant backing, with investors including Amazon founder Jeff Bezos's family office, Nvidia, Toyota Ventures, and even Tim Berners-Lee, the inventor of the World Wide Web.
LeCun isn't alone in championing this "world model" approach. Just last month, Fei-Fei Li, a leading figure in computer vision and founder of Stanford's HAI, announced $1 billion in funding for her venture, World Labs. World Labs is also focused on building AI that understands spatial intelligence and the physical world, aiming to generate editable 3D environments. This growing interest from prominent researchers and substantial investment signals a potential paradigm shift in AI development, moving beyond the "scaling laws" that have driven LLM progress towards architectures that truly grasp and interact with reality.
The AI landscape is clearly at a crossroads. While companies like OpenAI continue to push the boundaries of LLMs, a new wave of "rebel" startups, armed with deep scientific expertise and a focus on understanding the world, are emerging. This "research gene + industry landing" trend highlights the increasing complexity of AI, where breakthroughs require not just massive data and compute, but a fundamental rethinking of how AI learns and perceives. For Europe, AMI Labs' success is a significant moment, showcasing its potential in foundational AI research and open-source ecosystems, even as the global race for AI's future intensifies.
