We're seeing a marked increase in posts to the forum that are 100% pure AI generated responses - with no real Human input.
Whilst there is value in AI helping to formulate a response, for example, to help overcome language barriers, my concern is with posts that contain little real experience and insight.
I want this forum to remain a place of human connection around Node-RED. Accounts that post purely AI regenerated replies without any adding real value are likely to be flagged as spam and deleted.
While I certainly agree with this, I think we do need to be careful not to lock out people who have genuine needs but aren't very good at expressing them.
One thing I've noticed recently is that quite a few posts and new users marked for review by the system haven't been responded to for several days by anyone. If I'm not sure whether something needs to be accepted or rejected, I will tend to leave space for others to respond. But if it hasn't been dealt with after a couple of days, I think that is too long and I will respond. However, I tend to err on the side of accepting rather than rejecting.
I also think that, if a thread is going to have its root post deleted, whoever is doing that should put a reply in the thread saying why it happened. We've already had one person asking if they'd done anything wrong having posted a reply to a thread where the first post was deleted. Additionally, a short reason message may help other forum members better understand our stance on AI posts. If we don't explain, we will only ever get increasing numbers of AI posts I feel.
Looking at a couple of the threads that Marcus has flagged. There is a very clear pattern. The "users" posting have <1-2 minutes of read time in the forum but have "read" at least 10-20 posts.
Is it possible to block new users from posting until they have spent more time reading?
Also, it clearly shows that we need to block the "user" not simply the post. These are clearly AI bots.
We will formalise this position on AI content and add it to the forum T&C's/other appropriate places. I posted this as a placeholder to get the ball rolling.
They did actually reply to some of my postings and I was almost fooled by at least one of them; seemed to have good knowledge & interest in the topic (video streaming). Cool, thanks you spotted this!!
To be fair, it was Marcus who spotted them. I just got rid of them.
The quality of the responses is often telling though - very generic and quite smarmy.
This will get worse of course as the models continue to improve. If the bots deliberately take their time reading some threads before engaging, it will likely become impossible to tell in a year or 2.
But then, will we still care? Quite possibly, the quality of the bot responses by then may be so good, it won't matter!
I do think it's better to highlight if a question or reply is (suspected to be) AI generated than to delete it and ban the user.
"This post has been flagged as possibly AI generated. AI Agents might be posting deliberately dangerous code. Take this into account before adopting it's suggestions."
Imagine if you wanted to read Brave New World for a study of the real vs predicted impact of AI on society, only to discover that all copies of the seditious text have been destroyed (for your own protection of course).
The issue here is that we could simply suspend the user permanently but then we will get a constant built-up of cr*p users. But when you delete them, the posts are deleted as well.
In addition, do we really want to retain all those poor quality posts? If we do, it kind of undermines the point of saying we don't want that in the forum doesn't it?
If someone wants to come up with a better solution, please do. But please also bear in mind that it has to be achievable with the few moderators the forum has. If it becomes burdensome, it won't work.
Are you volunteering to take on that role? As I said above, the quality of AI posts is bound to continue improving. At some point, even if we had the resources to manage such a detailed assessment, the false-positives would outweigh the correct assessments.
They only need 1 copy though (or perhaps a few in parallel), we are not talking Fahrenheit 541 here! And the authors (well the publishers mainly) will be getting paid for those copies.
I seem to remember back when I joined I couldn't post for 30 days. I did a quick search and didn't find that but if it could be implemented would that cut down on bots? Sure they could wait but would they?
There are plenty of questions posted without the benefit of punctuation or which seek to steer replies down the user's preferred blind alley, etc.
Do you want to retain all those poor quality posts too, in the hope that (like AI), the person will continue improving over time, or ban them outright?
Can't someone vibe code a plugin for discourse Surely discourse must have something like this already ...
It's truly strange that discourse doesn't offer any tooling to help forums. Surely this forum isn't the only one to suffer under the guise of a brave new world of AI.
Rare books are being lost that way - things that don't have publishers any longer ... I don't much care about the latest Stephen King but rarer first editions are being shredder as well. Apparently - haven't bothered going down the verification rabbit hole here!
One copy each: each AI company will need a copy. Because sharing isn't the done thing any longer.
Before bots we had humans to detect human spam. There was just a short window of time where we had bots detecting human spam and humans detecting bot spam.
Edit: for human v. human, check out The Cleaners very interesting how under-paid Indonesian Filipino human-filters watch film material from over-motivated terrorists - day for day.
- I suspect from the same owner.
Not any more.
Surely discourse must have something like this already ...