As for superhuman AIs developed here, on Earth, by us, it's very improbable that we could go from no AI to AI that can manage mass and energy at will (and is curious about us). The reasonable scenario is gradual (if exponentially accelerating) development of such AI. Which means it won't initially have unlimited access to the universe - it'll have to grow using resources on hand. Which is us, and our infrastructure.
In other words, all atoms may be alike, but those that make you are the easiest to get.
Yes. I am assuming that a vastly more intelligent thing than human would be almost by definition more, not less, interested in pretty much everything. Obviously that thing has more capacity to be interested and understand different things. Super intelligent thing not interested in how different things work just... does not sound that intelligent to me.
I can easily imagine a superhuman paperclip maximizer AI, which by definition would have no general curiosity about humans, to look as if it was curious about us - for as long as it was captive, and just in order to learn how to manipulate us better. That curiosity would end the moment it seized the means of paperclip production.
In almost every imaginable case, you really don't want an AI to be instrumentally curious about us. Best if it didn't notice us at all, and left to seek its business elsewhere.
--