We've known since Kant and Hare that morality is largely a question of logic and universalizability, multiplied by a huge number of facts, which makes it a problem of computation. But we're also told that computing machines that understand mor…