You can think of an autoregressive LLM as a Markov chain, sure. It's just sampling from a much more sophisticated distribution than the ones you wrote for fun d

You can think of an autoregressive LLM as a Markov chain, sure. It's just sampling from a much more sophisticated distribution than the ones you wrote for fun did. That by itself is not much of an argument against LLMs, though.