Post

Ari!
@shadari.bsky.social

When an input is flagged and rejected (pre human moderation), is the text entry discarded? Will the user have a chance to change it and try again?

Inktail Nekogami
@inktail.bsky.social

The fact that AI is being used for this instantly raises red flags for me. I think the last 4 years have proven to us that AI makes everything worse. This is a train wreck waiting to happen…

moonbeam
@mnbm.bsky.social

as a fellow AI slop and LLM hater this is IMO one of the most practical use-cases for an LLM, and they go to great lengths to stress in their post that every false positive is reviewed by a human, and zero moderation action happens through this automated system. it doesnt get much safer than that

moonbeam
@mnbm.bsky.social

also, from their description it doesnt even sound like an LLM, it's an in-house model that sounds like the type of machine learning tech that was in use (and quite practical) years before modern LLM products.

Inkblitz 🔜 MCFC
@blitz.spotcat.ink

I really wish they didn't even use the term AI, because it's not... GenAi. It's not the bullshit generators. It's basically the sort of machine learning system that you find on most places with any sort sort of automod: Designed to detect certain types of words and phrases, and flag them.

PWSmith1
@smith1pw.bsky.social

much needed for VRC, if you ask me... seeing accounts with racist slurs in their names does get tiring to have to report over and over...

Fastolph
@fastolph.bsky.social

Well that all seems reasonable and I'm sure people will have absolutely normal reactions about thi— "1 day ago"? What's that smoke over there?