"I bet if I ask a human to generate a bunch of sentences no matter how unrelated, they could do it infinitely ... I don’t think an LLM can “undo” it’s probabilities already generated like this (maybe it can?) and then generate infinitely"

Unless it zooms in on "Unrelated" and ignores the sentences it produces. reinforced by unrelated sentences.

I got to think thinking about "people with no filters" and LLMs. LLMs don't really have an inner dialog.

I also think we shouldn't compare ChatGPT with LLMs 100%. While I'm sure ChatGPT is mostly an LLM at the very least it has a programmatic veneer for moderation. How know what is behind the curtain as far as every thing else.

I curious about layers in LLMs. so there's a concept of a frame in AI where a specific event has slots that get filled. so like a story has a beginning middle and end. so it learns that patterns at a higher level. maybe the concept of a story triggers a very high level concept and then it knows to keep things in that order.

Yet another post about LLMs (but give it a chance!) English is not my first language, I know it is a little rough in places, but I did my best.