So agent would have reached perfection. I'm not sure humans can create a machine capable of perfection. This is a philosophical question waiting for a definitive answer.
replyI don't think we will reach a world where agents make no mistakes. Because there are always tradeoffs. So an agent could choose the "wrong" tradeoff and we call that a "mistake", but is that really a "mistake"? Arguably not
replyIf the agent is able to make fewer mistakes but chose the wrong tradeoff it will be faster to fix forward than to look back, figure out why it happened, dig out the rollback, and then figure out how to change the decision.
reply