Not just another complaint about AI reply-bots

Plenty of people have already complained about the automated AI replies, and I think the admins are right to close those threads and point to the report button.

Even if people do report them, the harm remains though: every time I read an AI GPT reply, it dilutes the value of the forum: it happens quite a lot, and reporting them doesn't instantly remove all their posts.

Could I instead suggest an automated solution? Could the forum automatically pass the new message content to a GPT-detector API, and ban accounts that hit a very high threshold? Could even run it only on new accounts to save cost.

I'd love to see a zero-user-effort, instant solution implemented.

I'd be happy to volunteer to code it for free, to contribute to the community.

Thanks!

hardlinephone said:

Plenty of people have already complained about the automated AI replies, and I think the admins are right to close those threads and point to the report button.

Even if people do report them, the harm remains though: every time I read an AI GPT reply, it dilutes the value of the forum: it happens quite a lot, and reporting them doesn't instantly remove all their posts.

Could I instead suggest an automated solution? Could the forum automatically pass the new message content to a GPT-detector API, and ban accounts that hit a very high threshold? Could even run it only on new accounts to save cost.

I'd love to see a zero-user-effort, instant solution implemented.

I'd be happy to volunteer to code it for free, to contribute to the community.

Thanks!

Click to expand...

Very nice to offer it for free. Hope the forum gives it a thought.

hardlinephone said:

Plenty of people have already complained about the automated AI replies, and I think the admins are right to close those threads and point to the report button.

Even if people do report them, the harm remains though: every time I read an AI GPT reply, it dilutes the value of the forum: it happens quite a lot, and reporting them doesn't instantly remove all their posts.

Could I instead suggest an automated solution? Could the forum automatically pass the new message content to a GPT-detector API, and ban accounts that hit a very high threshold? Could even run it only on new accounts to save cost.

I'd love to see a zero-user-effort, instant solution implemented.

I'd be happy to volunteer to code it for free, to contribute to the community.

Thanks!

Click to expand...

I think a decent number of people use GPT to translate to english though. A human could probably differentiate GPT spam vs translations, but there's a chance something like this would false flag people genuinely using it for translations.

I'm not discouraging this idea as I'd love to see these spam posts disappear, but something to think about.

Deleted member 15724 said:

I think a decent number of people use GPT to translate to english though.

Click to expand...

Idk if I am wrong, but doesn't BHW have an automatic translation feature? I could have sworn I've seen screenshots of BHW not in English, as well as people quoting posts that translate into a different language altogether. Idk if it is true, but if it is, wouldn't that make ChatGPT redundant?

----
I don't think adding a GPT detector may do that much justice. Could potentially lead to some legitimate messages being misinterpreted for AI and will only clog up the mod queue of reports when users complain, which does no one a favour. Right now, the best thing to do would be to report. Though I do think AI messages should just have very minimal existence altogether. Not like you will die for not speaking good English (it will even improve writing skill to be less reliant on AI).

These are all good points and go along the same line as we already take which it doesn’t matter the source, we always ask does it add value to the conversation? If it doesn’t then please do report it, whilst we don't warn for AI content we do warn/ban for many other reasons such as spammed content, selling outside of the MP and trolling etc etc.

As mentioned above the reports that are reported as “AI” are sometimes found to be under the threshold when run through AI checkers such as Grammarly and do add to the conversation so we keep them in. Others we have removed as irrelevant content. It's all based on context and using human brains to verify the AI output.

On top of this you don't see the stuff that doesn't get through and is already caught by our existing tools! It goes in waves but it looks like dentists have found spam bots at the moment! Sheesh!

In short we do appreciate your offer but we've got this covered. Leaving AI tools to work out if something is AI is clearly a loop within a loop that could becomes quite Machiavellian !

We always appreciate the help we receive from the community and never forget that the community comes first.

If you'd like to help further we're always looking for moderators so please do hit the link below to apply if you're interested in joining the team.

https://www.blackhatworld.com/form/staff-application.1/select

The only catch is GPT detectors aren’t super reliable yet, so auto banning could hit real users by mistake. Still, focusing on new accounts could be a smart way to ease the load.

Not the only catch but one of many for sure. That’s why we will still use human to verify the outputs.

Thanks so much for detailed reply Damien, and the encouragement.

Onwards and upwards

Thanks @hardlinephone

As this has been answered and there have been no further comments, we'll close the thread now.

If you have further questions, please visit our newbie guide summary here https://www.blackhatworld.com/seo/s...ter-list-of-all-newbie-guide-threads.1389081/

If that doesn’t answer your question, please reach out to support.blackhatworld.com for further guidance.

Answer supplied - thread closed.