Ten counter-arguments that AI is (not) an existential risk (for now) — LessWrong