Meta’s star AI scientist Yann LeCun plans to leave for own startup (arstechnica.com)

Yann LeCun, Meta’s chief AI scientist and Turing Award winner, plans to leave the company in the coming months to launch a startup focused on “world models,” the Financial Times reports. LeCun is already in early fundraising talks. His exit comes amid a major leadership shake-up at Meta after CEO Mark Zuckerberg acknowledged the company had fallen behind rivals such as OpenAI and Google following setbacks including the disappointing Llama 4 launch and a struggling Meta AI chatbot. World models aim to build internal, simulation-style understandings of the physical world by learning from video and spatial data rather than relying primarily on text prediction like today’s large language models (LLMs). Proponents expect these systems to simulate cause-and-effect, reason about physics, and support planning and embodied behavior — capabilities that could complement or diverge from Transformer-based approaches. LeCun has cautioned the work could take a decade to mature, and the community remains split: some argue Transformers show emergent physical reasoning from large-scale training, while evidence so far suggests sophisticated pattern-matching more than true causal models. LeCun’s move could redraw research priorities and talent flows, signaling renewed interest in non-LLM architectures for more general, world-grounded AI.