Why Would AI "Aim" To Defeat Humanity? — LessWrong