Timeline for Policy: Generative AI (e.g., ChatGPT) is banned
Current License: CC BY-SA 4.0
17 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 31, 2023 at 22:23 | comment | added | uhoh | @Makyen slightly related to above comments space.meta.stackexchange.com/a/3057/12102 | |
| Mar 26, 2023 at 11:55 | comment | added | Makyen Mod | Hopefully, having a duplicate-target question available will result in fewer people being mislead by the "eloquent bullshit" which these things produce. Unfortunately, the number of people helped in a more general way regarding trust in current AI generated content is probably just a drop in the bucket, because most people care a lot more about getting an easy and quick answer, rather than a correct answer. | |
| Mar 26, 2023 at 11:54 | comment | added | Makyen Mod | @uhoh Yeah. Being able to close such questions as a duplicate of a general one that discusses the current level of reliability and accuracy for answers from AI generative technologies is about the only way to corral such questions, unless the site wants to go the further step of declaring them off-topic and have a site-specific close reason for them. | |
| Mar 26, 2023 at 11:46 | comment | added | uhoh | @Makyen yes agreed, like a canonical How do we know the Apollo Moon landings are real? target for closing as duplicate? | |
| Mar 26, 2023 at 11:33 | comment | added | Makyen Mod | @uhoh A) SO's policy (above) is that AI generated content is only in users' "about me". On another site, I'd either 1) ban such questions, or 2) A single canonical question something like "For [questions about site's topic], are AI generated responses reliable and accurate" and then close every question like what you describe above as a duplicate of that question. It's just a waste of effort to explain, over, and over, and over again that the current level of large language model AI generated content is "eloquent bullshit" and has absolutely nothing built in about being correct. | |
| Mar 26, 2023 at 11:18 | comment | added | uhoh |
@Makyen Okay got it. There are of course some questions in SO that are about coding but don't ask for code as answers, just for example those from the family of *theory* tags but I think those will be edge cases and handled case-by-case. My interest in this is in part how it might translate to or inform individual SE site policies, cf. space.meta.stackexchange.com/q/2999/12102 where I took "content" to include question premises) (since SO often takes the lead on things. There, "Chatterot says X, is it true?" questions might become an irritant if not a problem. Anyway, thanks!
|
|
| Mar 26, 2023 at 11:07 | comment | added | Makyen Mod | @uhoh The answer to that is a bit nuanced. It's covered in answer to: "Should we flag human-written questions that use code generated by ChatGPT?". My personal opinion is we should be more restrictive about asking about how to fix AI generated code, because such questions are effectively useless to future readers. In my opinion, it would be much better for us to have a "How do I do X" question, rather than a question that is "Please fix this broken code that I got from this tool, which often produces complete garbage code, 'how do I do X'." | |
| Mar 26, 2023 at 9:45 | comment | added | uhoh | @Makyen apologies if the answer can be found here somewhere; the statement mentions "...generated text for content..." on SO but most of the discussion here is about answers. Does SO have a policy on material cited as for example the premises of questions, say of the form "Chatterot says X, is it true?" or more elaborate questions that can be reduced to that? Would they be covered implicitly by the word "content" in the statement, or explicitly but I'm missing it, or for the purposes here is the material in questions not considered content, or none of the above? | |
| Feb 19, 2023 at 22:47 | comment | added | einpoklum | @Makyen Please consider moving your comments to my reply on the linked question (after which I can remove my comment here). | |
| Feb 19, 2023 at 22:32 | comment | added | Makyen Mod | I'd also note that asking in a comment like that is contrary to our normal advice to users about how to handle suspected violations of the rules by other users. The advice to users is almost always "raise a flag; explain the issue you think might be there and why you think that; disengage". This type of thing is what moderators are supposed to handle, because it involves dealing with the user, not just the content of the post. Moderators can see the user's full history, which often allows them to see patterns which regular users don't have the information available to be able to see. | |
| Feb 19, 2023 at 22:32 | comment | added | Makyen Mod | @einpoklum I'm not going to do so. You're welcome to spend that time if you want to. Unfortunately, we're talking about 10's of thousands of posts, so asking regular users (or moderators) to expend the substantial extra time to post a comment and then return to the post some time later to see if the OP actually responds doesn't scale. It just doesn't. We (people using the site, doing curation, moderators, etc.) just don't have the vast amount of additional time that would be needed. We, the site, don't have the curation resources available to handle each of these posts individually. | |
| Feb 19, 2023 at 21:51 | comment | added | einpoklum | @Makyen: Perhaps be more sparing with this advice. I'd first ask the answerer: "Did you use ChatGPT to figure this out?" - of course they might deny, but they are quite likely to either provide an alternative source of knowledge, or simply not to answer, and in both cases one has more information to base a decision whether or not to raise a flag. | |
| Feb 3, 2023 at 22:09 | history | rollback | Cody GrayMod |
Rollback to Revision 1
|
|
| Feb 3, 2023 at 21:47 | history | edited | starballMod | CC BY-SA 4.0 |
inline comment
|
| Dec 8, 2022 at 18:48 | comment | added | Silvio Mayolo | It doesn't look like anyone else has done it yet, so I've made the dedicated question post. I'll let an official moderator post an answer, rather than writing one up myself. | |
| Dec 5, 2022 at 23:25 | comment | added | Makyen Mod | Please raise an "in need of moderator intervention" flag on the answer and explain what the issue is (i.e. explain that you believe the answer was generated by ChatGPT). | |
| Dec 5, 2022 at 23:20 | history | answered | Hoppeduppeanut | CC BY-SA 4.0 |