Ted Chiang wrote an article on large language models a couple of years ago that was published in the New Yorker. It is a really excellent conceptual introduction to what passes as “Artificial Intelligence” these days. He is an outstanding writer, whether it’s technical writing or his world-class short stories. This time, it’s technical. Chiang clearly presents several fundamental concepts clarifying large language models in about as straightforward and clear a manner as I have yet seen. I highly recommend it.
(Note that the article applies to all large language models, not just to ChatGPT)