You're right, but I particularly believe that all our stored scientific information still has a lot to say, things that we humans haven't seen yet, and something that could decipher this, and very quickly, would be an AI.

What could bypass experimental validation would be quantum computing to simulate systems/environments.