On Tue, 2023-05-30 at 15:06 +0200, David Deutsch wrote: > I have checked the competence of GPT4 for make many times before I > decided to test it on mailing list questions. I have always found it > to be surprisingly good.
I've seen people post answers to GNU Make questions on StackOverflow that were generated by ChatGPT (probably ChatGPT3) and I was not impressed overall. The response mainly starts out OK but by the end it's clear that the response was generated and it has misleading or even incorrect content. I'm sure, like anything, it will improve. Maybe, with ChatGPT4, it already has. Mainly, though, I just don't think it's useful for these mailing lists to become a way for people to ask ChatGPT for answers, without typing their question into ChatGPT themselves. If they wanted to do that, they could do it directly: we don't need to become a group of ChatGPT stenographers!