Announcing the AI Alignment Prize — LessWrong