Re: ChatGPT ???
"Przemysław Kamiński (as cgenie at pm dot me)" <[email protected]>
| Newsgroups | gmane.lisp.lispworks.general |
|---|---|
| Message-ID | <[email protected]> |
https://www.pcgamer.com/software/ai/microsoft-launches-copilot-ai-function-in-excel-but-warns-not-to-use-it-in-any-task-requiring-accuracy-or-reproducibility/ Recently I asked Claude Sonnet to help me identify a mushroom that I found. It responded with a common name wildly different from the latin name in the same sentence -- the mistake is easily verifiable by one wikipedia search. Still, I think even a Markov chain wouldn't give me such random names -- in most text, common names of species and their latin names are very close together. I do use LLM sometimes, they are good for repetitive task, but you still have to check everything carefully. I think people concentrate on the "intelligence" part (which isn't that good), but forget the "language" part -- I find LLMs very nice for having conversation in a spoken language that I want to learn. It can correct my errors quite well. P. W dniu nie, 28 wrz 2025 o 21:44:03 -07:00:00 użytkownik David McClain (as dbm at refined-audiometrics dot com) <[email protected]> napisał: > Yes, I think I spotted a mathematical equivalent to ChatGPT drawing > an image of a human hand with thumbs on both sides. It invented a > division by 10 where no such division should have been present. > > It was clearly lifting the equation from something it had seen in the > ISO-226 model. But for some reason it felt like adding some creative > factors to the equation. > > You should not use ChatGPT for anything where you don’t have > superior knowledge of the subject, or at least a way to double check > its answers. > > 30 years ago we had college students that we called “Calculator > Cripples”. I hate to see how many new construction projects, like > bridges, will start falling down in the coming years… > > > >> On Sep 28, 2025, at 21:02, [email protected] wrote: >> >> On 2025-09-28 22:41, David McClain (as dbm at refined-audiometrics >> dot com) wrote: >>> … ChatGPT … willfully misleads … >> >> No, it has no will nor any notion of correctness nor incorrectness >> because it has no notions either. We attribute intent and thought >> like we see a face in an electric socket. Hallucinations Я us. >> >> Peace >> --Devon > > > _______________________________________________ > Lisp Hug - the mailing list for LispWorks users > [email protected] > http://www.lispworks.com/support/lisp-hug.html _______________________________________________ Lisp Hug - the mailing list for LispWorks users [email protected] http://www.lispworks.com/support/lisp-hug.html