Yann LeCun to depart Meta and launch AI startup focused on 'world models' (www.nasdaq.com)

Yann LeCun, Meta’s chief AI scientist and founder of FAIR, is leaving the company to launch a startup focused on "world models" — next‑generation systems that learn from visual and spatial data to build an internal, action‑oriented representation of the physical world. LeCun, a Turing Award winner who has publicly warned that large language models (LLMs) are “useful but fundamentally limited,” has begun early fundraising. His proposed research aims to tackle robust reasoning, planning and embodied understanding — capabilities he says could take a decade to mature — by moving beyond text‑only training toward models grounded in perception and dynamics. The break underscores a major strategic split at Meta as CEO Mark Zuckerberg pivots FAIR toward productized LLM work, installs Alexandr Wang to lead a new Superintelligence division after a $14.3B stake buy in Scale AI, and recruits high‑paid talent for a TBD Lab. The shakeup follows disappointing reception for Llama 4, recent AI layoffs, and executive departures, and signals both a philosophical rift about the path to AGI and the emergence of a potential new competitor. Technically, LeCun’s departure highlights two competing research threads — grounded, multimodal world models vs. scale‑driven LLMs — and could reshape funding, talent flows, and the broader timeline for more human‑like machine reasoning.