If you define AGI as something that can do 100% of what a human brain can do, then surely we have to understood exactly how brains work? otherwise you have have a long string of 9s as best.
To me true AGI is achieved when it gets agency, becomes truly autonomous and could do real things best of us, humans, do — start and run successful businesses, contribute to science, to culture, to politics. It still could follow human "prompts" and be aligned to some set of objectives, but it would act autonomously, using every available interface to the human realm to achieve them.
And it absolutely does not matter if it "uses the same principles as human brain" or not. Could be dumb matrix multiplications and "next token prediction" all the way down.