Semantic (Web) Base
Assuming that Google gets around to offer decent API's to build applications on top of Google Base, the "Google Base Web" looks like this:
On the buttom are sites that add data to Google Base (using bulk upload - not displayed are people that are entering data by hand). On the top are applications that access all this data through the Google Base service - much like an internet scale data warehouse. The content model of this system is a simple entity-attribute model.
It should be obvious by now that you could build a similar designed system that uses a different content model - like relations or RDF - indeed, people that believe in the superiority of RDF should be starting to build such a system right about now.
But make no mistake about it - building such a "RDF base" / "Semantic Base" / "Semantic Web Base" is no simple matter, especially if it should work for internet scale data sizes.
And the difficulties in building such a system aren't only of technical nature - some conceptual questions need to be answered first: Do I really want to treat all the data from all sources as one giant graph / knowledge base? (this may work for RDF, but I don't think you would want to do this for RDF(S)) Do I want to treat each source as a seperate graph? How do I discover the correct graphs? ...