As someone with a physics background, I'm struggling to think of a theory under which AGI is impossible. We have an existence proof for GI, and no evidence from neurology of structures so unique that we could not build an analog of them.
Yes, I don't see why AGI would be impossible. All you need is Turing completeness, and a workable approach, right?
Still, coming up with a workable approach could be very difficult. We and other animals are the only existence proof. And that's all based on a billion years or more of ~random variation and selection. So it'll likely be way overcomplicated and inelegant.
