Until recently, this forum encountered the occasional human troll. Easy to spot, annoying nonetheless.
Nowadays, we have to cope with AI bots which have become smarter and more difficult to detect by the day. Personally, I don’t want to expect a potential bot behind each forum persona and I try to give everyone the benefit of the doubt. However, I can’t help but notice that this is getting more difficult.
How do you guys deal with potential AI bot user accounts?
Does anyone know if Discourse (the platform of this forum) is planning to scan for AI bots or if there are any measures in place?
I’ve started to be wary of leading questions, not just in the OP but in amongst the responses to a thread:
“What do you…” or “What kind of…”, like you’d get in a questionnaire.
I know, that’s suspicious. On the other hand, it’s also a totally legit question, right? Like asking for a recommendation when it comes to audiointerfaces, monitors, etc. Although the account hasn’t been created just recently there are a couple of other flags. Nonetheless, I am still hesitant to brandmark a user as an AI bot. Benefit of the doubt or just me being naive. In that case we’ll see how the user responds.
See, that’s the dilemma. Nobody wants to chat with a bot. At the same time, it’s also unfair to new users to be regarded as potential bots.
I liked them old fashioned hot head trolls better…
Only last year it was easy to spot AI anything. But some of the AI generated films/reels are excellent - the days of 6-fingered, 3-armed characters are long gone.
And asking what gear others’ use is a legit question. Although guitar strings may be a bit too specific!
On the forum of a software company that does not, and never has, produced a guitar?
(Yeah, OK, I actually have more Yamaha guitars than any other brand, but … on the Steinberg forum?)
The more I think about this so-called “AI”, the more I have to ask “what problem, exactly, is this solving?”, and the more I end up concluding that it’s actually creating more problems than it’s solving, and worse, creating novel problems that could never have been imagined before.
Who would have thought that we would end up considering, perchance, whether to penalise the proper use of language?
No doubt, a certain bot will be along shortly to give me a “like” for that one!
So, how to deal with bot accounts? Someone in the other thread suggested linking the MySteinberg database of registered product owners to the Discourse accounts – is that not already in place, because that’s also how you log on here, is it not? Perhaps it could be linked to trust levels?
I would vote for that with all 18 fingers on my 3 arms..
Feel free to flag posts from users that look like AI bots. There are many hints that reveal it, location, recent registration, names that don’t make sense for whatever reason, like the wrong gender. But also be careful not to assume a user is actually a bot just from one post.
I delete many new users on a daily basis who are spammers or are suspect for other reasons, so there’s that.
Is this a real issue as of yet though? LLM bots that have posted in the past while have been removed.
LLM bots make up a fraction of nuisance posts and new users here. Much more prevalent are bots that are not LLM, that create new users for some advertising purpose. Discourse spam filters automatically removes most of them, the rest are manually removed by staff and moderators.
The spam bots I’m talking about have doing the same thing for a decade and a half or more
I would be more concerned with users mistakenly accusing other users of being so-called AI bots
