Interesting thoughts.

It seems like an attempt at Oracle AI, which simply strives to answer all questions accurately while otherwise exerting as little influence on the world as possible, would be strictly better than a paperclip maximizer, no? At the very least you wouldn't see any of the risks of "almost friendly AI".

You might see some humans getting power over other humans, but to be honest I don't think that would be worse than humans existing, period. Keep in mind that historically, the humans that were put in power over others were the ones who had the ruthlessness necessary to get to the top – they might not be representative. Can you name any female evil dictators?