i think we should be very wary of ai-powered moderation but with gab joining the fediverse in 10 days, we're going to see a lot more fascist instances popping up. this gives moderators an even greater workload.
just a thought: perhaps something like @fediversespace, combined with a simple classifier, could allow moderators to proactively discover and block instances that harbor nazis.
i'm not talking about running machine learning algorithms on the contents of anyone's posts. rather, on the relationships between instances.
if a new instance shows up, and it primarily interacts with gab and other "free speech" instances, it could be a hint that it's something you'd want to block.
one spooky instance