36
$\begingroup$

The user Serg Z. is currently spamming answers copy-pasted from chatbot output (perhaps with slight modifications).

They have recently deleted two examples (users with 10K rep can see these) after I asked them what was going on. Here is the first one:

Subgroups of the $p$-adics that miss $\mathbb{Z}$

Is there nothing that can be done about this?

$\endgroup$
27
  • 11
    $\begingroup$ Ah, I guess there really isn't anything to be done since the moderators are all on strike... ...largely because stack exchange is refusing to take the problem of chatbot-generated content seriously. $\endgroup$
    – Stephen
    Jul 21, 2023 at 20:22
  • 34
    $\begingroup$ The current policy from our SE overlords is that we are not permitted to moderate material simply for the reason that it is generated by chat bots. This, among other reasons, is the motivation behind the current moderation spike. If you believe that this kind of content is harmful to the site, I would encourage you to use the "Contact" link at the bottom of the page to get in touch with Trust & Safety, and let them know what you think of GenAI generated content. $\endgroup$
    – Xander Henderson Mod
    Jul 21, 2023 at 20:39
  • 5
    $\begingroup$ Try looking at the activity page, as of now. $\endgroup$ Jul 21, 2023 at 20:53
  • 16
    $\begingroup$ "current moderation spike" Funny typo. $\endgroup$ Jul 21, 2023 at 23:08
  • 5
    $\begingroup$ We couldn't even prevent them from answering this question! (For those unaware, there is a deleted answer to this question which was posted by the same user whose name is mentioned in the post as a spammer. Ironically, the answer presents a few suggestions on "how to handle the situation", and also looks like it was spat out by a chat-bot). Note that we still deleted the answer using the usual channel of downvote-and-delete, but if removal is not an option then nothing can be done. Nevertheless, thank you for spotting the user. $\endgroup$ Jul 23, 2023 at 5:55
  • 3
    $\begingroup$ @peterwhy I would appreciate a sincere explanation from Serg Z., but none of the answers deleted so far are anything more than pure trolling. Do you disagree? $\endgroup$
    – Stephen
    Jul 23, 2023 at 16:35
  • 3
    $\begingroup$ @peterwhy I have undeleted one of those answers, which I perhaps judged too early. You may now see the most reasonable answer the user has posted. Note that they have posted four answers, two of which are marked as rude/abusive and a fourth one which consists of philosophical idioglossia and has nothing to do with the matter at hand. The one I'm undeleting appeared first chronologically. $\endgroup$ Jul 23, 2023 at 17:06
  • 2
    $\begingroup$ @peterwhy For me here in meta the distinction is not between factually correct and factually incorrect, but rather between answers that sincerely address the question and those that are intended only to provoke. $\endgroup$
    – Stephen
    Jul 23, 2023 at 17:13
  • 3
    $\begingroup$ @hardmath I was using spamming in the sense of online gaming culture, meaning repeated voluminous use of the same action (in thid case, posting chatbot generated answers over and over). I can see that perhaps my choice of terminology was unfortunate, given the existence of the flags. $\endgroup$
    – Stephen
    Jul 23, 2023 at 23:38
  • 4
    $\begingroup$ I don't know if this is worthy to have another thread for it, I don't use Meta much so please excuse me if I'm breaking some rule, but this user Zoonnie Zonn seems to also be spamming chatbot-like answers. $\endgroup$
    – Bruno B
    Jul 24, 2023 at 9:01
  • 5
    $\begingroup$ @BrunoB Thank you for the alert. A significant number of these answers are wrong, and the style of writing is definitely very chatbot-like. It might be a particular type of chat-bot as well. Phrases like "youngsters these days" and the first paragraph of this answer also seem very artificial. To the best of my efforts, this is probably chatbot-generated. $\endgroup$ Jul 24, 2023 at 9:26
  • 4
    $\begingroup$ How the hell has the general public gotten the idea that ChatGPT can understand/do mathematics? $\endgroup$ Jul 24, 2023 at 21:05
  • 10
    $\begingroup$ Hey folks, I've handled this situation for now and left an answer on Meta Stack Exchange confirming that our Inauthentic Usage policy applies to purposefully disruptive actions, irrespective of GPT usage. $\endgroup$
    – Slate StaffMod
    Jul 24, 2023 at 22:57
  • 3
    $\begingroup$ @Obie2.0 "Frankly, unless it is because all the moderators are on strike, the user's behavior on meta alone would normally be enough to get them banned, ChatGPT or no ChatGPT" Indeed, this seems to be a correct assessment. $\endgroup$
    – Xander Henderson Mod
    Jul 24, 2023 at 23:17
  • 3
    $\begingroup$ @MathCrackExchange Two things contribute to this notion (1) The fact that the general public doesn't know much advanced mathematics. ChatGPT actually does do a decent enough job on questions say at the level of eighth or ninth grade mathematics, and the public thinks that's good enough (2) The fact that ChatGPT is capable of disguising its answer with good vocabulary and can "start" answers well by stating the basic theorems/results/definitions etc. : anyone with a low attention span is seeing that and falling for it. $\endgroup$ Jul 25, 2023 at 9:24

1 Answer 1

31
$\begingroup$

The current policy makes it very difficult to discipline users merely for posting bad answers, even, or perhaps particularly, if those bad answers are the systematic product of limited language models.

However, the mere fact of using LLMs does not negate all policies around the Code of Conduct, trolling, and so forth. This user has also been writing hostile comments on various sites, accusing people critiquing their answers of "reading out of their posterior" or of being unintelligent and envious. They also have been reposting answers after their previous answers have been deleted, as with this very question. Some of the ChatGPT errors have also caused them to violate the Code of Conduct by claiming credit for the work of others, such as one answer that implies that they invented the theory of relativity.

Finally, some of these answers are so far from the content of the question as to be "not an answer" on their own merits, as with their most recent answer to this question, which appears to be pseudo-Buddhist musings mostly unconnected to the question.

Any of these activities could be subject to moderation whether or not they were using ChatGPT. A user engaging in this sort of behavior without AI would certainly be sent to cool down for a bit, and I think the same can be applied to those users who violate the site rules while incidentally using LLM content generators.

$\endgroup$
4
  • 1
    $\begingroup$ Someone has to be first, I guess. $\endgroup$
    – Scott Rowe
    Jul 25, 2023 at 1:37
  • 2
    $\begingroup$ @ScottRowe - In what sense? $\endgroup$
    – Obie 2.0
    Jul 25, 2023 at 2:18
  • 2
    $\begingroup$ You knew this kind of thing would happen eventually. $\endgroup$
    – Scott Rowe
    Jul 25, 2023 at 10:27
  • 2
    $\begingroup$ This is a first? And who is Obie 2.0? $\endgroup$ Jul 27, 2023 at 15:44

You must log in to answer this question.

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