Even though LLMs are bullshit engines and none of us should waste a second talking of "AI" which only exists in the smokey heads of #TESCREAL cultists,
I have come up with a farily accurate notice:

chatGPT generates text that is syntactically adherent to your request. This text may or may not, at random, correspond to the meaning you attribute to the words you have used.
In the same way, the text produced by chatGPT may or may not, again at random, correspond to actual facts or events, while always being grammatically impeccable and worded in such a way as to appear authoritative.
ChatGPT does not possess, again by design, any concept of 'truth' or any knowledge model and is therefore structurally incapable of distinguishing true from false.
This is why we can say thatChatGPT does not literally lie, because in its code truth and falsehood are indistinguishable, but merely bullshits, in Professor Frankfurt's definition, i.e. makes entirely plausible statements without any constraint of reality.

Also, you may find the last two episodes refreshing :)

https://www.spreaker.com/show/dataknightmare-en

hello!
I hear you, but I have to firmly disagree. LLMs are by design exactly bullshit engines, in the definition of Frankfurt: they generate entirely plausible content in a sort of authoritative voice, with no relation whatsoever to facts. So they do not literally lie, but they neither tell the truth, since the truth value of everything that comes out of an LLM is not bound by facts.
Sorry if I seemed to suggest the term "bullshit" was just derogatory, it is actually an accepted term in academia when talking about bullshit.

Regarding the alleged disruptiveness, I would say so far it is an unsubstantiated marketing claim. Yes, LLMs can produce text. But unless that text is very thoroughly vetted by a competent human, I for one would not touch it with a stick. LLMs replacing journalists, lawyers and teachers? three answers:
1) it is a ridiculous claim
2) it is also very demeaning to socially important porofessions, and
3) with Weizenbaum, I say the point is not whether a machine can do a job, but if it should, and the answer I give myself is there are jobs which should not be automated, because we as humans would be degraded by that.
Thank you!