With the rise of AI, we see more and more platforms, including forums, get flooded with AI-generated messages. This has pretty obvious downsides, reducing the quality of answers, making them more verbose as AI tend to be less concise, etc.
I would propose banning the use of AI in the forum for user messages. Concretely, this would forbid pasting AI-generated messages inside user comments, even if some human text surrounds it, unless it is for the sake of demonstrating what AI provided as an answer to a specific prompt.
I also believe AI-generated images should be banned, but this is a broader topic as there can be valid use cases for allowing such content.
Obviously, an AI-generated answer that is entirely rewritten by a human from scratch with their own words would be authorized. The goal of this proposal is to ban AI “slop”.
I strongly believe that, on top of AI answers being harder to read, allowing AI content would lead to progressively getting human content get more and more replaced with generated ones, which would be a big problem.
Whoever writes comment on the forum becomes the owner of what they write and is thus responsible. If the content is good (whether it was written by them, by an LLM, by a friend or other means) then it doesn’t matter.
I find a lot of value in AI-generated information (much better than what most humans do in general - better explained, more details, etc).
Also, it’s not because the LLM write something that it’s not been discussed/agreed with the person posting it. So I really see no reason to forbid this, it would just lower the quality.
One idea: we could be to have an AI-attribution clause (same as with git commits), but I don’t really see the advantage of doing that. It doesn’t matter with whom you worked to produce the content in the end. What matters is the quality of the content. I wouldn’t oppose an attribution rule (I just find it painful and nowadays there’s almost always an LLM involved in some way).
So if we want to fight something, I’d suggest we fight about the content quality and not discriminate about who wrote it. And I’d prefer to base any rule that we set on a precedent. Have you noticed some forum post where you saw some content generated by a LLM that wasn’t useful?
I have the opposite view Human-written content are much harder to read because they’re so sloppy, they forget context and details, only answer to 1 question when 2 are asked, etc.
So I’m rather -1 on your proposal. Let’s see what others think.
I think it’s important to link or quote the sources you based your decisions / arguments on. E.g. if I ask you a question and your answer influences or confirms my decisions then I prefer to link (when possible) or quote your answer when presenting my decisions. I apply the same rule for answers given by AI. Thus, I don’t think we should prevent forum users from pasting AI-generated messages inside their comments.
Then there is also the language barrier. We use English in this forum, but most of our users are not native English speakers. I don’t see a problem if they use AI to better formulate their questions and asnwers in English.
I agree with Vincent that each user is responsible for what they write (push) on the forum. E.g. if they use AI and write long overly verbose questions then I might not have the patience to read and provide an answer. But as long as their AI generated message is on the topic, related to XWiki, I don’t see why we would prevent them from posting.
Finally, even if we were to decide to forbid AI-generated messages, I’m not sure how you would be able to detect this automatically without false positives (that would annoy our users very much).
Human-written content are much harder to read because they’re so sloppy, they forget context and details, only answer to 1 question when 2 are asked, etc.