Dovid > same applies to human beings.
ChatGPT is different from more professional AI apps as it learns from a large corpus of human text and speaks like your average person, just with less safeguards. Some theoreticians posit that the next generation of GPTs will perform worse than the current (contrary to the usual tech improvement trends) – as current human generations learns from GPT and next generation of GPT will learn from current writers …
A traditional AI poses a question and then measures how well it can answer it. It can be wrong but in a controlled way. There is now work to combine GPTs with responsible algorithms, will see how that goes.