19:07
I'm just gonna leave some of my thoughts on AI generated answers here
I think there's several issues at play with AI answers. First of all is whether or not it generates a verifiably correct answer. Secondly, if it can indeed create a correct answer indistinguishably from a human author, then there may or may not be also an ethical(?) obligation to disclose that it was AI generated.
My opinion is that if you use an AI to assist in writing an answer, and it is correct, and the community is unable to discern if it came from a human author, or an AI author, or a human author with some AI contribution, then it is empirically as functional as an actual answer and should be treated as such.
The problem isn't as severe if the AI generates a verifiably wrong answer; we already have the downvoting system for that to filter out wrong answers. It doesn't matter if it is wrong and written by a human, or wrong and written by an AI. It will be treated as a low quality contribution.
Now I think the next issue is the volume of posts that can be generated with the help of AI. If previously we were getting 10 wrong answers, we don't want to be in a situation where we are getting 1000 wrong answers in the same timeframe, and the downvoting or flagging system becomes unable to cope with the sheer amount of nonsense being filtered through.
For that reason, I am particularly intolerant of using our site as a testing ground for AI generated content, and I do not want to encourage a environment where we have to sift through mostly garbage in order to find a good coherent post.
