Even if you start off with messy complicated AI, it could modify itself to become a more efficient optimizer. And it would do that because it would be beneficia

Even if you start off with messy complicated AI, it could modify itself to become a more efficient optimizer. And it would do that because it would be beneficial for whatever set of values it has, no matter how complicated. Regardless I don't know why you think a messy AI is safe. Having a really complicated utility function doesn't mean it is going to result in anything desirable for humans.