25
$\begingroup$

For details, see Sites can now request to enable a banner to warn about their policy on AI-generated content

Edit note: I've added in three answers for upvoting. This should cover all options without needing to downvote, but users are free to vote as they choose. As of this edit no moderator has commented, so there is no guarantee that there will be any outcome from this vote, but if there is a clear majority one way or another we can ask the moderators to consider the vote and take action.

The summary is that it is possible to have (currently) one of two banners, or no banner at all. The two banner options are:

  1. a reminder that AI-generated[1] content is not permitted on our site
  2. a reminder that AI-generated content must be cited as such on our site

The advantage that I see of having a banner is that the policy is incontrovertibly visible to all would-be users of the site. Not having a banner doesn't mean we can't have a policy, but arguably having a policy that is less easy to find could lead to more frequent violations of it.

If there is a disadvantage to having a banner, it is likely that it might bring up more complaints that MSE is somehow 'hostile to newbies' which we seem to get every six months or so.

Thoughts?

[1] Despite months of discussion just about everywhere I still see no evidence that a large language model is meaningfully AI and not just an expert-system with a large database and better hardware.

$\endgroup$
10
  • 1
    $\begingroup$ What about having two answers (one for each option) that can be voted on separately? I have seen this on Code Review: codereview.meta.stackexchange.com/q/10917/35991. $\endgroup$
    – Martin R
    Jan 6 at 9:25
  • 4
    $\begingroup$ @MartinR If no-one else does, then when this has been up for 24 hours (to give people time to comment and make suggestions) I'll post three answers: the two banners options and a no-banner option. $\endgroup$
    – postmortes
    Jan 6 at 9:46
  • 4
    $\begingroup$ I think on previous such polls on Mathematics Stack Exchange, the convention has been to upvote the answer that you agree with, but not downvote any answer. So I would suggest that we do this here as well. $\endgroup$
    – Joe
    Jan 6 at 22:06
  • 2
    $\begingroup$ Also, I would remove the text saying "Note that this is tacitly a vote for permitting AI-generated content" on the two answers. I think this makes the text a lot less neutral than it would otherwise be, and it is best not to make assumptions about the reasons people have for voting a certain way. $\endgroup$
    – Joe
    Jan 6 at 22:16
  • $\begingroup$ Note that the banners only cover AI generated answers. They do not cover such content in questions. $\endgroup$
    – PM 2Ring
    Jan 7 at 4:58
  • $\begingroup$ BTW, LLMs aren't like expert systems (eg, Wolfram|Alpha). As I said a year ago meta.stackoverflow.com/a/422397/4014959 they're good at certain language structure transformations, but they really aren't designed to be expert in anything beyond making plausible sequences of words. $\endgroup$
    – PM 2Ring
    Jan 7 at 5:15
  • 2
    $\begingroup$ @Joe I see your point about removing the line about 'a vote for AI-generated content' but I think it clarifies, for non-native English speakers, that voting for this isn't just a vote for some text on the site, but has a deeper meaning. So I think it needs to stay. $\endgroup$
    – postmortes
    Jan 7 at 8:04
  • 1
    $\begingroup$ @Joe : People are downvoting all the same, so I think that it would be good if somebody (you, I, and postmortes have enough reputation to see it, but not everyone does) would report the upvote totals in a comment or edit after some time has passed. $\endgroup$ Jan 11 at 5:09
  • 5
    $\begingroup$ FYI (responding to a couple of flags and comments): the tag status-review is typically added to a question on meta once the community has come to some kind of consensus. It seems that there is consensus that Math should have a "No LLM" banner, so I have bumped this question up the line with the review tag. $\endgroup$
    – Xander Henderson Mod
    Jan 12 at 12:57
  • 3
    $\begingroup$ Near one week update : Banner : not permitted has $56$ upvotes, $5$ downvotes for a total score of $51$. Banner : citation please has $9$ upvotes, $10$ downvotes for a total score of $-1$. No banner has $5$ upvotes, $14$ downvotes for a total score of $-9$. If anyone wishes they may take only the upvotes into account or the entire score into account. These numbers haven't changed in the last $16$ hours, going by my last visit to this page. $\endgroup$ Jan 13 at 12:09

3 Answers 3

54
$\begingroup$

Banner: not permitted

Upvote this answer if you are in favour of having a banner reminding users that AI-generated content is not permitted on math.stackexchange.com

$\endgroup$
-3
$\begingroup$

Banner: citation please

Upvote this answer if you are in favour of having a banner reminding users that any AI-generated content must be declared as such when posting on math.stackexchange.com (Note that this is tacitly a vote for permitting AI-generated content).

$\endgroup$
2
  • $\begingroup$ Is "permitting explicitly specified AI-generated content, and disallowing uncited content which was secretly AI-generated/paraphrased etc" implied by your last sentence in parenthesis? $\endgroup$
    – Snared
    Jan 8 at 23:32
  • 4
    $\begingroup$ @Snared I think that's exactly the meaning of the banner -- no implication required. The three answers here are mutually exclusive (or are intended to be, at least). $\endgroup$
    – postmortes
    Jan 9 at 7:47
-11
$\begingroup$

No banner

Upvote this answer if you are in favour of having no banner regarding AI-generated content visible on the site. (This is tacitly a vote for permitting AI-generated content).

$\endgroup$

You must log in to answer this question.

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