AI needs to get some flack for how it’s working.
2 Likes
Our AI right now, watching from the corner… 
2 Likes
I don’t think this forum needs AI moderation at all. We don’t have banable behavioural issues and we’re very small so don’t attract any loons.
I get that it’s fun to play around with as an admin, but having innocent posts flagged by an AI is not good for the community.
There’s also a lot of healthy scepticism of ’AI’ here, so I’m not sure AI moderation will be popular. The summarise stuff is fine as it is optional.
2 Likes
Our AI moderation is more like
and it flagged this post… 
Because of a gun?
OK guys…I’m going to stop posting for a while…I can’t seem to attach ANYTHING it doesn’t think is a copyright violation - this is FAR WORSE than YouTube takedown notices…
BTW this is how it looks behind the scenes (it will flag this, so give me a minute…). The funny thing is that it flagged it because it is seen as NSFW, not because of the gun, and not because of a copyright violation. Maybe because Elmer is holding a flaccid gun? This AI has a dirty mind! Also interesting, the two detection models have a very different degree of confidence: 83% vs 16%.
2 Likes
I agree 1000% with Tams - time to dismiss the frat mom…
2 Likes
So, we can’t talk about our gun problems here? Asking for a friend.
Indeed. I will disable OpenNSFW since it is clear that it is not as accurate a model as the other model.
2 Likes
@Hifihedgehog fyi one of @Bronsky 's posts got flagged for being toxic. It was a decent attempt (there were some trigger words in there) but I wonder if we can switch to a mode where posts are never automatically hidden, but instead just flagged as potentially toxic, and marked for review.
That way people can freely post what they will, and the AI still keeps an eye on things (“AI on things” #dadjokes) and helps us with moderating, without annoying the user base.
1 Like
I’m not annoyed at all. I think it’s kinda funny.
Good for you. Getting flagged briefly spikes my blood pressure, glad you’re OK with it. I’d still like a less aggressive approach, with flags but without auto-hide.
Count me as annoyed BECAUSE I don’t think our tribe has EVER gone out of bounds (new site and old)…it’s one of the biggest reasons I come here more than any other site…
3 Likes
Understood and agreed! I’ve gone ahead and switched off the toxicity and NSFW filters since there is no need for it. Apologies if it seemed like it was pointed at the original crew who we all know and trust. My intention was not for this to be directed at our core group from old TabletPCReview but to be smart enough to filter if or when we grew and some immature folks or naughty bots got inside. Plus I just like trying new things but this is definitely not a value add at all. We’ll revisit this in the future should site traffic deem it so and we get crazy people trickling in. In fact, just blame me for going crazy with a new toy squirt gun, accidentally filling it with ketchup and me getting everyone in the neighborhood messy in the process. Pardon the mess to the conversation flow.
3 Likes
Agree with revisiting…
STRONGLY DISAGREE with any need for an apology - you have done an incredible job with this site, so play away…(as long as the AI overlord is in prison)…
2 Likes
Thanks everybody including you Bronsky for understanding and taking it in stride. The bot obviously is too stupid to get context or meaning, so it is off as it should have been. We’ll switch it back on only if the AI improves and/or we have a legitimate need for it from newbies joining who can’t handle themselves.
1 Like
The great part is the NSFW and toxicity filters are both self hosted (as in our server runs it and not some third party processor). And now it is self quarantined to an indefinite off status since it proved itself too dumb to figure things out for itself. I monkeyed with several different LLM options but it clearly needs more revisions and testing from Discourse before they set it loose on the masses.
2 Likes