Imagine a cube

Based on this article by Mike Pearl posted on Gizmodo.

Legendary AI scientist Yann LeCun is reportedly parting ways with Meta, convinced that the industry's current obsession with scaling Large Language Models (LLMs) is a "dead end" for achieving true human intelligence. Instead, he champions "world models," a new paradigm designed to understand the physical world by processing sensory data—something he argues even a housecat can do better than the most advanced text-trained LLMs.

His core belief is that AI must be able to internally model and reason about the world, like a human can effortlessly "imagine a cube floating in the air and rotate it." This high-profile split, reportedly fueled by organizational shifts at Meta, signals a profound philosophical divergence at the top of the AI world about the foundational path to future superintelligence.

Check out this article.

Related Stories