Is that a requirement though? If the AI used existing understanding/methods (which I assume it did) then there's nothing really "new" there, right? Shouldn't it be possible to get understanding from how the existing methods were used? Or, even knowing it's true, cull an incredible number of paths from the problem space around it, to make finding understanding easier?

I'm not a mathematician, but I've always been able to find understandings from exploring the paths starting from either direction of the conceptual graph that exists between a "problem" and a "solution".

Third paragraph from the end of TFA:

Mathematicians are both frustrated and excited by OpenAI’s work. Sarnak, for his part, said in an email that he won’t be satisfied that it’s “a Theorem” until humans check and understand all the details. But “the general feeling is one of relief,” Radziwiłł said. There are scores of results across not just number theory but also geometry and theoretical computer science that the proofs of quasi-Riemann will, in a swift stroke, establish as mathematical truths.