[New Feature] Beta Testing AI Detection for Unwanted AI Content

XrumerGeek said:

View attachment 525044
7th or 8th time in the last calendar month. Yet members using AI for every reply does not flag.

Can we just accept the fact that this module is not fit for purpose. It's made things worse (which was a very high bar).

No matter how many times you try and bang the "work in progress" drum, it will not change the above. Get rid.

Click to expand...

bump (9 days)

easiest solution would be allowing certain users become a "content police".

not full moderators but somewhat giving them an option to hide/mark posts and shadowban clear Ai users.

it would allow to clean up forum quickly and with small margin for error.

XrumerGeek said:

View attachment 525044
7th or 8th time in the last calendar month. Yet members using AI for every reply does not flag.

Can we just accept the fact that this module is not fit for purpose. It's made things worse (which was a very high bar).

No matter how many times you try and bang the "work in progress" drum, it will not change the above. Get rid.

Click to expand...

XrumerGeek said:

bump (9 days)

Click to expand...

Thanks for the bump, we've made some additional changes since your last post. The module keeps a lot of AI slop off the sub-forums it's currently live on. However, you'll not see the ones that get blocked, so understandable that you'd think it's not fit for purpose; however, I can assure you it's catching a lot. So the work in progress remains and it's ongoing incremental improvements will be rolled out over the next few months.

Bloody hope so! Cheers for the response.

lucky.sparks said:

easiest solution would be allowing certain users become a "content police".

not full moderators but somewhat giving them an option to hide/mark posts and shadowban clear Ai users.

it would allow to clean up forum quickly and with small margin for error.

Click to expand...

This is something we've discussed, and will continue discussing. In fact, this was most recently brought up in a discussion about the upcoming relaunch of the Exec program... (shhh, don't tell anyone I mentioned it)

So we're not saying no!

At least 90% of the regurgitated replies mimicking previous responses in threads are made by new members / members with sub 100 posts. Just limit new members to 5 posts per day for X period. It's a 2 minute implementation on Xenforo and it will reduce the impact in threads by 90% overnight.

You could even allow new members to request the rate limit to be lifted for them - with a staff member taking 60 seconds to checking their recent activity - and deciding for/against. Optional add-on.

I'd challenge anyone to show me a live example of a new member who has signed up and made 25+posts within 2 days of registration that offer any value. So why are we so against flicking that simple switch to put an end to it? It's a solution to drastically eliminate the issues with zero downside.

Yes, it won't fully eradicate. But it will become immediately manageable. Expecting members to report 15 replies in a 20 reply thread is ridiculous. And that's the reality right now; a 20 reply thread will have 4 or 5 unique responses on a rolling average.

You should not confuse rewriting a post with ChatGPT to fix grammar mistakes with a post that was written by ChatGPT from scratch.

The first one is a good use case. It turns broken English into something understandable. The second one is just ChatGPT generated garbage.

I see many people using it to increase their post count, get jrvip, and start selling their services.

In the short term it might benefit you, but in the long run it will kill your forum.

I'd assume part of the AI detection is to look for glaring emdashes.