LLMs aren’t intelligence. We’ve had similar technology in more primitive forms for a long time, like Markov chains. LLMs are hyper specialized at passing a turing test but are not good at basically anything else.