I think there is a legit problem here. High quality publically accessible research can provide huge gains to society as a whole, but it's effects are diffuse,

I think there is a legit problem here. High quality publically accessible research can provide huge gains to society as a whole, but it's effects are diffuse, many people get a small benefit. It's not the school that is the primary benefactor of great research. On the plus side, in machine learning, most of the big companies are doing a good job of publishing their cutting edge research.