Freebase is a new semantic web database of, well, everything. I think a kind of semantic Wikipedia would not be a bad name for it. I have been doing some things with semantic web technologies for my M.Sc. thesis and I got to see the importance of metadata. Tim O'Reilly has given Freebase a try and is intrigued. Michael Arrington also writes about it. Michael refers to it as a kind of Google Base done right. I think very few people actually ever understood the power of the Google Base idea, myself included.
The problem I see with freebase is that it is a completely centralized model. They pull in data from Wikipedia and lots of other sources. Sure everything is licensed under a Creative Commons license, but I fear that it is too ambitious for a startup to want to be the database of everything. Basically a new version of the internet (web 3.0?) in one place. Wasn't the power of the web that it was decentralized, do we not need that anymore all of the sudden?
Anyway, I signed up for the account waiting list on the site.