I think Mikhail’s example shows why it’s so bad when AI generates “original content” (as opposed to translation/voice transcription for those who need it). It’s very long, it’s unclear what it’s based on, and it’s just badly written, at least on the first attempt.
My only concern with total ban of AI-generated posts is, how can it be moderated efficiently? If the poster doesn’t want to put a label on it, it’s going to be hard to prove.
May I suggest to replace “AI generated content” to “any auto generated content” in the discussion? And to express the reason more clearly and directly. Something like we don’t want to see junk here.
Translation or any other help from any AI (including but not limited to GO related) is not auto generated content. So the proposed term solve this issue of clarification.
I think generated content could be allowed only in the context of discussion AI or such content.
I think your concern about moderating effectively is a valid concern whether there is a total ban or a partial ban.
That being said, I imagine you could level the same criticism against attempts to moderate the use of AI in games. There will be cases where it’s unclear, and if someone is careful enough they could get away with it. However, egregious cases will be clear and there is value in tackling them.
Likewise, for the use of AI in the forums, even if some people are able to get away with prohibited AI use I think there is still value in tackling the obvious cases. Even if it is explicitly written in a sticky thread that copy & pasting an AI responses is prohibited, someone will inevitably make a post that begins with “I asked chatgpt and here is what it said…”. Even if we miss some more subtle cases of prohibited AI use, we’d still be able to catch these obvious ones and I think there would be value in doing so.
Why do should we care of prohibited AI use in communication? I think we should care only about worthless posts that disturb other users. In most cases AI generated content is such thing. So there is a reason to prohibit it. But should we analyze every post with the sole purpose of determining whether it is generated or not?
Sorry if my last post wasn’t clear. I was using the term “prohibited AI use” as a catch all for whatever is ultimately decided to be prohibited, whether that be a blanket ban on all AI use, just generated images, wholly generated posts, etc.
Given that no final decision has been made, I left open the possibility that there may be a ban on some form of AI use that may be hard to detect. My point was that even though some potentially prohibited AI use may be difficult to moderate, that does not detract from the fact that there is value in banning and moderating the obvious cases, such as when someone explicitly states that they asked chatgpt for a response and then paste the contents.
To be clear, I’m not for a total ban on all AI use in the forums. I am perfectly okay with it for, say, translation purposes and I don’t think this even needs to be labled as AI.
I’m strongly against proliferation of people spamming forums with AI generated content that they just copypaste like parrots. But using AI for translation is acceptable: not everyone is fluent in English, and I prefer communication with AI interference than no communication at all.
On the other hand if a forum user’s English skills are so limited that they are totally unable to check the AI’s translation, then some disclaimer would be appreciated (like “AI generated translation, apologies for any mistakes”).
Perhaps of interest, this has been topic(s) of discussion over at Discourse Meta
It seems like they havent had an outright ban on AI (makes sense), but have banned users for excessive AI slop/spam
Also makes me wonder what AI tools we might be using to improve the forums experience 
I think any that do not impose thoughts. Definitely not the ones that create content out of nowhere.
For example, AI-powered search that warns you about similar topics when you create a topic is fine. There is such feature here. I guess it’s based on the full-text search algorithm. I don’t think it works well enough. AI could perform better and help a lot.
Yes, good point!
In a similar vein, I’ve seen AI summaries of long threads work ok (though I wouldn’t trust them to capture all the important nuance).
I chose the option:
But I think that even if this is what the forum decides on, there will still be posts of AI content which people will reasonably find unhelpful. I hope any decision will be regarded as a first step, with an expectation of further refinements, and not a statement of an all encompassing community decision.
Yes - I think that like any other “inappropriate/unhelpful” content, this sort of thing can be flagged as such and assessed on its merits.
The goal of this thread was to undestand whether or not simply because it is AI generated would mean that we would rule it out.
Recent reactions to some AI content was leading to discussions among moderators that maybe this should be our policy,
However, the broad sentiment in this thread seems to point towards the fact that “just because it is AI generated does not mean we have to rule it out”.
Rather, AI content should be assessed on its merits like any other.
This means that flags like “hey this is AI generated, therefore remove it please” would mostly be rejected.
In order for a post to be removed, it would need to be more like “This post is purely AI generated content that is demonstrably wrong, and (or) completely irrelevant and derailing the discussion”. Or whatever specific objection there is.
Do you mean you will not remove non AI generated content that is demonstrably wrong, and (or) completely irrelevant and derailing the discussion?
No, the opposite, I was saying that a flag raised on a post for these reasons would be a “good flag”.
I assume most of these reports are not simply about AI, but that they are low effort/low quality uses. In other words, might still qualify as spam. s this not the case?
I guess I’m interested to know what kinds of posts might “make it through” depending on my vote 
I think these are probably good examples of what we’re talking about:
I assumed only the latter might have garnered reports, because it was an trojan horse of sorts (not clearly labeled) i can’t imagine people are reporting the former since it’s obvious that the content is AI-generated. But maybe I am wrong, and i think that’d be a useful insight to have.
But it is exactly the spam. Why do we need subdivide it into AI or non AI generated?
I’m not sure about the double negative 
But, the main underlying question in this thread is “should AI content be disallowed because it is AI generated”.
We were receiving flags stating this: “Remove this post, it is AI generated”.
This thread appears to have established that AI content can be allowed.
It means that there needs to be a valid reason to reject it that is not simply “it is AI generated”.
Content would not be rejected “because it is AI” but rather for some other pre-existing reason (such as spam, abuse, off topic etc).
I don’t want to go off topic, but that’s just a single negative!
I’m actually surprised about this. Or rather, I’m not surprised the reports came in like this, but surprised that there isn’t also subtext “and it’s low effort/spam/unlabeled slop”
I guess maybe the outcome in this thread is that the reports need to be more specific 
Do you suggest add an option “AI generated spam”?
Nope.
Basically I would say examples like
Do we want people to be helpful by saying
hey I’ve pasted the contents of this thread into ChatGPT, Gemini, Copilot etc, and now I’m pasting the output of that back here with very little additional comments of my own
in many threads or not?
That’s the thing I think we should really be voting on. It’s the most practical example of AI content we will see and have been seeing.
Personally I don’t mind if we have a thread for stuff like testing AI image generation, or people experimenting in their own thread with an AI tutor etc. You can always mute threads and even categories you no longer want to see.
“Summary” posts within threads from AI either just summarising for the sake of it, or using it as a sort of “I want to make a point so here’s the AI’s take” are things I believe are controversial and likely to cause issues.
It’s going to be more work than even just rereading a thread, to verify if the AI summary or AI’s take on things are accurate or hallucinating. Not to mention how the AI is prompted will make a difference.