Yann LeCun says LLMs can’t truly reason.

World models train systems to simulate physical environments rather than predict the next word.

World Labs and LeCun’s new startup each raised roughly $1 billion, but today’s models remain narrow.

Sources: Ars Technica