Oh sure, I agree that we're not imminent danger. However, the field is advancing quickly with us getting glimpses of what autonomous agents can do when seeded with a goal function. It is not too difficult to imagine adaptive and flexible goals that enable broad unpredictability and learnt behavior.