So I actually think this viewpoint vastly overestimates the ability for human reviewers to come to the right conclusion. I can tell you thats not really the cas

So I actually think this viewpoint vastly overestimates the ability for human reviewers to come to the right conclusion. I can tell you thats not really the case: paid trained human contractors for content moderation very commonly make far more problematic errors than "wpa2 cracking info is a bannable offense, but not this case because it's a academic research". This is the kind of edge case that even after escalation to higher tier human contractors will often still get wrong. I don't really see how it's actually much less dystopian just to have humans in the loop with the same outcomes.