Timeline for Policy: Generative AI (e.g., ChatGPT) is banned
Current License: CC BY-SA 4.0
26 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 9, 2024 at 23:14 | comment | added | FNia | If you are going to use AI to detect AI-generated answers, why not instead use it to flag potentially incorrect answers and actually help the "swamped volunteer-based quality curation infrastructure"? | |
| Dec 31, 2022 at 9:48 | comment | added | Akshay Sehgal | Who said anything about "Einstein of programming"; just because you know/mastered a specific concept in programming doesn't make you an Einstein. That's a childish analogy. The whole point of SO is that it's a community-curated platform. If something is not answered by someone correctly, the community corrects it, and improves/maintains it over time. If you were following anything around ChatGPT, its clearly being recognized that it is poor in its accuracy and ends up just generating "confident" sounding answers. The issue is the misuse of ChatGPT by folks that have 0 knowledge of that concept | |
| Dec 31, 2022 at 9:17 | comment | added | The Muffin Man | You act like all questions on Stackflow need the Einstein of programming to answer them. Many of the problems don't require that. I'd rather see AI generated answers that can be downvoted and a feature to request that a human answer because the question is involved further than generating one function. | |
| Dec 6, 2022 at 17:49 | comment | added | Akshay Sehgal | Solutions will come sooner or later. Academia uses methods to detect AI-generated plagiarism (models that modify input text while not changing context), Media/Content platforms are using Deepfake detection models, and there are a lot of AI-generated text detectors out there, that can and will be finetuned for ChatGPT once the model is made available officially. SO is not the only platform that will be negatively impacted by the misuse and exploitation of this model. For now, the temporary ban is a great strategy to assess the impact vs hype situation around the model. | |
| Dec 6, 2022 at 17:21 | comment | added | user400654 | A public temporary ban, such as this one, and mods enforcing it, is enough to deter well-meaning users from abusing it. Nothing short of a ban will stop the others. That's better than doing nothing at all until a better solution exists. | |
| Dec 6, 2022 at 17:18 | comment | added | CrandellWS | @AkshaySehgal I do love SO and its community. AI can be useful or abused. ChatGPT would help me better state my questions which means I may get help I would not have otherwise gotten. ChatGPT is here and is not going away. We have to adapt, we simple do. I like the idea of a "temporary ban" but people are still posting with it now. I think we need extreme restrictive actions on all SO user as part of the "Temporary ban". I wish I had that magic bullet suggestion to fix all this but it is above my pay grade. I just worry cause I rely on SO to help me. | |
| Dec 6, 2022 at 17:08 | comment | added | user400654 | If only the curation/moderation tools were flexible enough to create new flag categories on the fly | |
| Dec 6, 2022 at 16:54 | comment | added | Akshay Sehgal | @CrandellWS, This is what other folks have previously mentioned. How do you know that an "expert" has verified it before posting it? Regarding the "lack of desire to write", SO is a community. You will always find people active here. Whether for the act of solving problems, gaining a reputation, building their portfolio or just boredom, there will be people who will answer your question, if it's constructed and written well. And, just to further motivate the "big league", there is a bounty feature. | |
| Dec 6, 2022 at 16:48 | comment | added | CrandellWS | @AkshaySehgal One posting to SO seeking an answer may not have the expertise to validate the AI answer is correct and therefore needs human expertise help. The expert may be able to verify an answer as being correct fairly quickly may just be using AI help for speed writing. And just cause they know a correct answer does not mean they could write it or are even willing to try and explain it. AI help could Bring new experts who were previous shy to answer because of a lack of desire to write. | |
| Dec 6, 2022 at 16:36 | comment | added | Akshay Sehgal |
@CrandellWS, quoting your previous comment - he can simply verify the AI is correct and copy and paste that to me ... Just one question about your hypothetical scenario; why can you do the same directly? Why post on SO for 'expert assistance' when you would be happy to rely on an answer by an AI model? You could ask the AI generator, test the answer, and then if it doesn't work, ask the question on SO. Most of our guidelines here encourage self-attempt to solve the problem first, before asking a question. Why can't the AI-generated answer be a part of your own attempt at the answer?
|
|
| Dec 6, 2022 at 10:15 | comment | added | CrandellWS | @BDL I agree but we in a flood and I am just saying a spillway might be a good idea | |
| Dec 6, 2022 at 10:14 | comment | added | CrandellWS | @bdl I get that I support extreme temporary measures till a solution is found... no matter how restrictive or nonrestrictive the extreme is | |
| Dec 6, 2022 at 10:14 | comment | added | BDL | @CrandellWS: Checking those answers takes considerably more time than what is acceptable. You can't expect volunteers on SO to serve as a review service for AI generated content. Either the author of the answer checks for correctness before posting, or AI answers aren't possible. | |
| Dec 6, 2022 at 10:12 | comment | added | BDL | @CrandellWS: In a perfect world, people would use AI to generate an answer, then test it, verify that the AI answer is correct, and only then post it on SO. Unfortunately, this is not what's happening right now. We have tons of users copy-pasting AI answers to Stack Overflow without checking at all. Some of the answers are correct. Some are total nonsense. Only today (in the last three hours), I saw at least 5 users just copy-pasting without any checking. They posted answers in wrong programming language, self contradicting answers, and so on. | |
| Dec 6, 2022 at 10:11 | comment | added | CrandellWS | and I am not suggesting that experts be correcting AI writings that is a waste of time. Rather if they can just check it and then send me a correct ai written answer I fully welcome it and if they check the AI answer and deem it incorrect and dont share I like that to | |
| Dec 6, 2022 at 10:07 | comment | added | CrandellWS | My main thing is that maybe the expert I need can answer the question but hates writing and may only answer my question if he can simply verify the AI is correct and copy and paste that to me... I would appreciate that because the fact AI wrote it matters not to me... it being an accurate or correct answer does though, I mean I can not afford the people I tend to need help from and if AI makes it faster for them to correctly answer my question and that is my only option I dont want to eliminate it because they may be to busy otherwise to answer at all | |
| Dec 6, 2022 at 5:59 | comment | added | Akshay Sehgal | @Makoto - the community helps detect a lot more than just quality, right now as well. Moreover, just adding this flag is not a misuse of communities time, because that is a function of how much AI generated content actually gets posted on SO. As the mods comment on another answer mentions, users still would be flagging it for mod intervention. Infact it will help reduce the burden on the mods who will have to review each answer and will help cleanup SO faster. Finally, a flag like this will help lead research on detecting AI generated content, if the data is made available by SO. | |
| Dec 6, 2022 at 5:51 | history | rollback | Akshay Sehgal |
Rollback to Revision 4
|
|
| Dec 6, 2022 at 5:50 | history | rollback | Akshay Sehgal |
Rollback to Revision 3
|
|
| Dec 6, 2022 at 0:50 | comment | added | Peter Cordes | @Makoto: Having a way to get tons of bad answers that look good on the surface deleted is a new problem that might warrant new tools. If a comment that it looks like AI-generated garbage is enough to swing the tide of voting towards downvoting when deserved, then maybe not. But the community helping mods identify users who post this kind of AI crap might be helpful to get them suspended if normal rate-limits don't block them. (Either because they fooled enough folks into upvoting garbage, or they had some rep to start with but still failed to grok how terrible this is.) | |
| Dec 5, 2022 at 22:32 | comment | added | Mark Harrison | Professor Turing, please call your testing office... | |
| Dec 5, 2022 at 22:15 | comment | added | Makoto | This sounds like...not a good use of time or resources. The community should be evaluating the quality of the response, not necessarily if it came from an AI or not. But your other points I can agree with. | |
| Dec 5, 2022 at 21:59 | history | edited | cottontail | CC BY-SA 4.0 |
added 5 characters in body
|
| Dec 5, 2022 at 21:57 | history | edited | Heretic Monkey | CC BY-SA 4.0 |
Changed use of code format for emphasis to bold.
|
| Dec 5, 2022 at 18:41 | history | edited | Akshay Sehgal | CC BY-SA 4.0 |
deleted 1 character in body
|
| Dec 5, 2022 at 18:33 | history | answered | Akshay Sehgal | CC BY-SA 4.0 |