Genius because it should be obvious? Yes, it should be obvious for people familiar with both areas, and it is actually known for quite a long time. Hyperbolic geometry is used for visualizing hierarchical data since Lamping-Rao 1995 and Munzner 1998, then there were papers about hyperbolic SOMs (2001 IIRC) and lots of papers about the Hyperbolic Random Graph model for scale-free networks. I would say that the paper linked above introducing "Poincare embeddings" (a rather poor name IMO, BTW) does not feel as impressive if you know the details and the earlier work.