Yann LeCun, one of the most respected minds in artificial intelligence, is making headlines for charting a new course away from Meta. For over four decades, LeCun has been at the center of major shifts in the field, pioneering techniques that power today’s AI. He now criticizes the industry’s obsession with large language models (LLMs), arguing they’re useful but fundamentally limited. In his view, these models, including ChatGPT or Meta’s Llama, lack real understanding of the world, persistent memory, reasoning, and the ability to plan complex actions core ingredients of genuine intelligence. LeCun believes the next leap will come from “world models”.
AI systems that learn from visual data, like watching videos, and focus on how things physically evolve. These systems aim to predict causal relationships, not just generate text. After 12 years at Meta, LeCun is starting a new venture dedicated to pushing this line of research, hoping to spark a revolution in machine intelligence. His new company will develop technology that understands the physical world, potentially setting a new paradigm for how AI is applied across industries.
While Meta continues investing billions in LLMs, LeCun urges a rethink: building smarter, context-aware AI could drive the future far beyond today’s generative tools.
