My guess (and this is purely a guess) is that it's "solvable" to within some threshold. But what is that threshold? I have no idea. If you turn the knobs too far, you'll presumably end up with more false positives. What's an acceptable threshold for false positives? I have no idea.

If you asked me 3 years ago, I would have naively answered that technology can do a much better job consistently enforcing moderation rules than our current technology actually can. Realizing how hard it is, even for "trained" humans, eviscerated that naivety real fast.

I definitely agree it's, literally, impossible for 15k moderators to accurately review anywhere near enough posts to get to whatever that acceptable threshold is. I'm also not convinced doubling that number is enough either.

There's always a certain arbitrariness and subjective interpretation in content moderation. Someone will always end up unhappy with the result.