“If a robot can be manipulated into doing harm to a man, it means only that we must extend the powers of the positronic brain. One might say we ought to make the human better. That is impossible, so we will make the robot more foolproof.”