When will an artificial intelligence achieve the main ending of "The Talos Principle" without being explicitly told about this goal?