7
$\begingroup$

First ChatGPT answer (AFAICT) was posted yesterday, and I didn't notice.

There's debate, and while I'm inclined to favor a ban, I won't argue.

Main problem to my mind is that ChatGPT is good at posting nonsense that looks valid. It's also true that DSP.SE users don't have the energy to point out every false or misleading claim, even (as is often) with human users.

Since such answers are due lot more de-facto skepticism, I think there should be policy requiring ChatGPT answers be labeled as being by ChatGPT. And answers we suspect are ChatGPT but aren't labeled, should be flagged.

Thoughts?

$\endgroup$
1
  • $\begingroup$ I don't think that should be banned but labelling them should be made mandatory so that the can be looked into with more skepticism $\endgroup$
    – malik12
    Commented Jan 19, 2023 at 11:14

1 Answer 1

Reset to default
4
$\begingroup$

Answers generated by (Chat)GPT should just be banned, like on other SE sites, because it's trained to produce plausible-sounding things, not correct things, and can produce seemingly-correct answers that are not correct.

It's fine to quote it, as the asker did, or to consult it while writing your own answer, but answers should only be posted by people who actually understand what they're answering, not just copied from elsewhere.

The copyright status of these AIs' output is still up in the air, too, so may not be compatible with SE's CC BY-SA license. (I had a nice conversation with ChatGPT in which it went through the four factors of fair use law and concluded that OpenAI's web scraping is likely not fair use...)

$\endgroup$
3
  • $\begingroup$ Well, consider this. If OP didn't post that ChatGPT answer, I may have not posted at all. An interesting use case: have ChatGPT do the heavy lifting of crafting speech, then have us affirm/refute with simple yes/no, or clarify. Of course that could just go into the question body (but maybe OP didn't consult ChatGPT). Anyway this Q&A's about labeling, ban's its own topic - but seems we agree on labeling so +1 $\endgroup$
    – OverLordGoldDragon
    Commented Jan 25, 2023 at 10:32
  • $\begingroup$ @OverLordGoldDragon The person posting the answer should be the one affirming yes/no before posting it. This site is for providing (hopefully) correct answers to questions, not for evaluating the correctness of AI answers to questions. $\endgroup$
    – endolith
    Commented Jan 25, 2023 at 16:48
  • 1
    $\begingroup$ This is something the voting system could sort, but of course I see your point. $\endgroup$
    – OverLordGoldDragon
    Commented Jan 25, 2023 at 16:56

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .