Yann LeCun is at it again 'If future AI is still based on the current autoregressive LLMs, they’ll be very knowledgeable but still very dumb.' Hallucinations, hard to control, just parroting—this is an architecture issue, not a scale issue. LeCun is betting on 'World Models,' saying the next AI breakthrough might come from Meta rather than OpenAI. So, what are world models?