Timeline for Policy: Generative AI (e.g., ChatGPT) is banned
Current License: CC BY-SA 4.0
23 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 9, 2024 at 23:18 | comment | added | FNia | You probably meant "unverified" instead of "unmodified". There is no way to "ban" AI. This platform itself will have no option but to use AI in order to flag potentially incorrect answers and help its "swamped volunteer-based quality curation infrastructure". | |
| Apr 5, 2023 at 10:25 | comment | added | Christof Kälin | That's exactly the sane middle-ground (not a compromise to be clear) barely anyone is taking these days. Although difficult to manage as DubDub said. | |
| Apr 3, 2023 at 16:03 | comment | added | DubDub | "I think unmodified answers from ChatGPT should be banned" how do you prove this, and there is no point banning something that can't be proven. | |
| Feb 7, 2023 at 22:49 | comment | added | user400654 | Nothing we can do here would inform on how correct it is in a way that would be valuable to it's improvement. @Rahul | |
| Feb 7, 2023 at 22:43 | comment | added | Rahul | I don't think acting all purest and banning suggestions from ChatGTP is a good idea. In fact, letting it know how incorrect it is, is way more constructive in building better models for the future. There should be a clear demarcation if the suggestion is from ChatGTP. | |
| Feb 2, 2023 at 11:38 | comment | added | Paul Masri-Stone | I agree with this answer and the comment by @V2Blast. If someone has used ChatGPT to get an answer, taken the time to verify its correctness — the essential human curation step — and then posts it on SO, it could be useful to SO users. I would propose a citation along the lines of "Solution by ChatGPT, verified by me". Ultimately the important thing is answer quality, not where the answer came from. | |
| Jan 17, 2023 at 4:38 | comment | added | Dawood ibn Kareem | @iBug - "... takes more time and effort to verify a machine-generated answer than to write one yourself ..." is probably true if your first language is English. However, for non first language English speakers, writing your own answer could be far more time-consuming than verifying and tweaking an AI-generated one. | |
| Jan 1, 2023 at 6:58 | comment | added | The Muffin Man | I love how existing scaffolding tools in Visual Studio/code aren't under fire, but using GPT is. This is the whole guns kill people instead of people (with guns) kill people argument. There's a lot of immaturity in this entire post. Furthermore, the arguments being made here act like GPT is spitting out 1000 line methods that no one understands. If you copy paste an answer from anywhere on SO without vetting it that is the source of the problem, not whether it came from GPT, a VS extension or MSDN forums. | |
| Dec 28, 2022 at 13:44 | comment | added | Machavity Mod | @V2Blast More relevant than that: using ChatGPT could prove self-referential (which is not a good thing at all) | |
| Dec 19, 2022 at 0:28 | comment | added | symbiont | if you can verify it, then you should just write the answer yourself. "correct it to the best of your knowledge as needed" and what if you don't have enough knowledge, you will let ChatGPT's mistakes slip in? this is a bad idea | |
| Dec 13, 2022 at 7:27 | comment | added | iBug | @NikS As others have pointed out multiple times: It's much faster at producing plausible answers that poses a greater problem to Stack Overflow than it seems. | |
| Dec 12, 2022 at 22:21 | comment | added | JonathanReez | @iBug I use ChatGPT every day for my work as a software developer and I disagree. It's a much faster version of StackOverflow basically. | |
| Dec 5, 2022 at 16:43 | history | edited | Peter Mortensen | CC BY-SA 4.0 |
Active reading - but more could be done.
|
| Dec 5, 2022 at 15:51 | comment | added | V2Blast StaffMod | Relevant XKCD: xkcd.com/810 | |
| Dec 5, 2022 at 15:51 | history | edited | The Thonnu | CC BY-SA 4.0 |
Grammar
|
| Dec 5, 2022 at 15:47 | comment | added | V2Blast StaffMod | If users are taking the time to go through an auto-generated answer and verify its correctness before posting, credit its sources (even if this is actually possible for an auto-generated answer) and avoid plagiarism, etc. – and posting them at a reasonable rate (rather than flooding the site with bad answers), and improving them based on feedback in comments (rather than just dumping them on the site and then abandoning them) – then I would imagine it's harder to tell that they're even auto-generated. But at that point, there's little differentiating it from an answer fully written by a human. | |
| Dec 5, 2022 at 15:00 | comment | added | Gimby | It is a useful tool to help yourself, not others. I see chatgpt as an alternative to posting a question on Stack Overflow. | |
| Dec 5, 2022 at 14:05 | comment | added | user492203 | you can't necessarily tell for sure (well, you can try asking it the same question and see if you get a similar answer), but if it's an obviously low-effort/incorrect answer then it should be deleted. if there are no issues with the answer then it shouldn't really matter if chatgpt was involved in it. | |
| Dec 5, 2022 at 12:55 | comment | added | NoDataDumpNoContribution | "unmodified answers from chatgpt should be banned" But how does one detect if the output from chatgpt was taken unmodified or not? | |
| Dec 5, 2022 at 12:23 | comment | added | user492203 | yeah i guess you're right, it probably has more value for asking about specific details than answering the entire question. | |
| Dec 5, 2022 at 12:16 | history | edited | user492203 | CC BY-SA 4.0 |
deleted 2 characters in body
|
| Dec 5, 2022 at 12:14 | comment | added | iBug | I don't think this needs to be explicitly exempted. It usually takes more time and effort to verify a machine-generated answer than to write one yourself, with less garbage and verbosity included. | |
| Dec 5, 2022 at 12:12 | history | answered | user492203 | CC BY-SA 4.0 |