People asking questions are frequently terrible at choosing a title for their question. I think it might be an improvement to have members below a certain reputation required to use a ChatGPT-generated title for their question.
-
2That would be AI generated content and thus, by definition, not permitted.Paulie_D– Paulie_D2024-05-15 13:00:48 +00:00Commented May 15, 2024 at 13:00
-
1@Paulie_D User-submitted AI-generated content is not permitted, and that rule is specifically targeting user-submitted answers because they're often wrong. The suggestion was to have Stack Overflow generate the question titles, like Loom does for video titles when you record a video using their app. SO can do whatever it wants, and the problem the ban discusses with LLMs often generating incorrect answers does not necessarily apply to their ability to generate summaries of a question.Nathan Wailes– Nathan Wailes2024-05-15 17:37:19 +00:00Commented May 15, 2024 at 17:37
Add a comment
|
1 Answer
Let's not.
Aside from the fact that AI-generated content is flat-out banned on SO, if people have trouble coming up with a good question title, they're likely to have just as much difficulty telling a language model what the title should be.
LLMs are often (mistakenly) seen as a magic box that fixes all your problems.
Providing good input to them is at least as difficult as writing a decent title is, so I don't see this fixing any problem.
-
15Also, been there, done that, and it was horrible2024-05-15 13:05:50 +00:00Commented May 15, 2024 at 13:05
-
1@Cerbrus the idea would be to have the LLM generate a summary of the body of the question, like how Loom auto-generates a summary title for a video you record. So, no input from the user.Nathan Wailes– Nathan Wailes2024-05-15 17:03:14 +00:00Commented May 15, 2024 at 17:03
-
1@Cerbrus Also your reference to the fact that AI-generated content is banned seems to miss the point of the ban. It's a ban on user-submitted AI-generated content (not a ban on SO using AI-generated content), and it's specifically about answers, because the answers are often wrong. LLMs are quite good at generating summaries and I would suspect the average quality of their question titles would be better than the average quality of titles written by users with reputation less than 200 (for example).Nathan Wailes– Nathan Wailes2024-05-15 17:33:02 +00:00Commented May 15, 2024 at 17:33
-
@Zoe Thank you for the link.Nathan Wailes– Nathan Wailes2024-05-15 17:40:02 +00:00Commented May 15, 2024 at 17:40
-
Maybe we can revisit this when AI can consistently draw better feet than Rob Liefeld.user4581301– user45813012024-05-15 18:45:29 +00:00Commented May 15, 2024 at 18:45
-
1That would require users to write good questions.Cerbrus– Cerbrus2024-05-15 23:06:12 +00:00Commented May 15, 2024 at 23:06
-
That does happen from time-to-time.user4581301– user45813012024-05-16 00:04:10 +00:00Commented May 16, 2024 at 0:04
-
Maybe you're sure. I'm not. I've seen the output of auto-summarizing tools. If we're lucky, the asker gets a decent starting point for a decent title..user4581301– user45813012024-05-16 17:46:31 +00:00Commented May 16, 2024 at 17:46