You are right! But feels like it's a much harder problem to find hyper realistic materials and geometry at scale than actually rendering them with PB algorithms.

True, but one can argue that if you don't scan real materials and real geometry you are losing some signal too. There needs to be a balance between the domain gap and the readiness/cost of data. Imo the problem is more that we don't have assets and less about how we render them.

It's great to see a lot of the research community moving to bluesky! It would be a great boost if also the twitter handles that share papers would move here too! Does anyone know if any of these accounts are here?