The Fermi paradox as evidence against the likelyhood of unfriendly AI — LessWrong