« If we're going to achieve compassion in the machines and also feel safe with the machines, to raise machines with human-like values, we need to make them human-like by simulating, or perhaps eventually imitating, human beings in high accuracy from top to bottom. »