When we talk about the existential risk of AI, many people imagine that it's as if machines will revolt and turn against us humans. But the risk is that machines will get so good (in many ways) that we will trust them too much and give them more power than we can then control.