There is a big message telling users not to use AI generated content for answers
I'm starting to wonder if the no AI answers rule is a good idea
There is a big message telling users not to use AI generated content for answers
I'm starting to wonder if the no AI answers rule is a good idea
There were too many questions asking the same thing that had been answered many times before.
Too many where the answer is RTFM.
SO used to have interesting questions, questions that I needed answered or that I found interesting to spend time on to research.
I think that is precisely the problem. A kid is teaching himself to program. He has an error which a million people before him have faced. He types it into google and gets SEO bullshit. He asks stackoverflow and they close his question immediately as a duplicate of a question which, to him, looks very different. He asks ChatGPT and he gets an answer immediately.
Where does start next time he has a problem?
So it will be with the possibly wrong answer from ChatGPT.
When learning nothing replaces a book or tutorial that has had some proof reading or editing.
Just picking up small bits from the web here and there does make things difficult and does not really teach you.
Learning in general needs you to fail and think not just google an answer.
IMO a lot of the interesting questions got moderated away. Almost all of my favourite SO questions are closed as being off topic or subjective. I do think that the community struggled to scale with influx of a huge volume of users. But I also think that the strategy that they chose to deal with it was ultimately overzealous.
Your definition of interesting is a discussion which is not what SO is for - even if they did then they do not have to tools to help this e.g. threading being able to vote on interim comments etc.
Even if it is it is often difficult to put the individual facts together and that is what a SO answer will do.
Most SO answers tend to be near the correct way as they explain why they answered.
A good answer should also gove references as to why the answer is correct.
The gamification of the Stack Exchange sites is a double-edged sword, and I think it relies on an assumption about the effort threshold required to provide an answer.
Typing out an answer, even a wrong one, takes time and effort that the system then rewards through points and badges.
If you can just copy a question to Copilot and then paste the answer back, the effort threshold is likely then low enough that too many (or at least lots) of people will do only that in order to engage with the gamification and get points, rather than do it to try to help out the asker.
The issue is that current AI answers invent 'facts' or produce wrong code. AI might help improve how people questions and possibly help them find an answer e.g. improve searching but to answer you must be correct and give correct references - not false ones as have been found in legal cases and others.
How to use AI is an open question.
You need cookies to count the views.