Comments
DeathToAmerica wrote
Yikes...
AlexanderReidRoss wrote (edited )
The underlying API used to determine "toxicity" scores phrases like "I am a gay black woman" as 87 percent toxicity
Wtf?
This is a lot like how facebook instructs mods to censor black kids but enable white supremacists.
NEOalquimista wrote (edited )
Imagine that after posting a certain amount of "highly toxic" posts you are temporarily blocked from posting any more comments. And since the centralization and monopoly of these comment services made almost every website use the same system, you cannot talk anywhere anymore. Also, you can't simply create a new account, as it requests your real ID and perhaps some official documents linked to it.
This is the wet dream of governments and companies.
tnstaec wrote
Just like that Twitter chat bot that swiftly started spewing racist shit.
Buge wrote
Adversarial machine learning is an interesting field. As machine learning increases, attacks on in will increase as well.
Pe95vxACL2Px wrote
they're working on Reddit integration lol
https://github.com/conversationai/conversationai-moderator-reddit