Ask HN: Should BMI preceed AI for goal alignment?

If you've got an AI that is out of control and human-indifferent or -hostile, all BMI will do is make it easier for the AI to control us.

Now, I presume you're hoping that improved BMI will make it so that the AI understands us better, and therefore helps it not become human-hostile. And in fact it might work out that way. My guess, however, is that any human-equivalent or better AI will think so radically different from us that empathy isn't realistic, and therefore improved BMI will not decrease the probability of a human-hostile AI.