Freebase is trying to create a database of knowledge with a collaborative Wiki-like approach. The differences are (1) there’s much less in it today, (2) it leverages Wikipedia as a source, (3) it’s far easier to edit entries in Freebase than in Wikipedia, and (4) it doesn’t follow a strict encyclopedia-like entry per topic.
As a simple example of point 3, despite having wanted to make Wikipedia edits I’ve never bellied up to the bar to do so. In 10 minutes of messing around on Freebase I updated the entries for both myself and Mark Logic in Freebase. It’s very easy. And they do a good job of both knowing what metadata is appropriate for which types of entities and for providing pick-lists for entering into metadata fields (e.g., dates of employment, job titles, companies). And they make it easy to add a new entry as well. For example, I had to add Versant Corporation to their database of companies before listing it as one of my past employers. (My only enhancement request is to enable “present” as a valid value for date types when listing dates of employment.)
When using Freebase I find myself doing primarily two things: (1) extracting data from text and loading it into metadata fields (e.g., the attributes about the San Carlos airport runway were added by me) and (2) creating web links / relationships (e.g., I added the link to the San Carlos airport website.) Basically, I’m structuring knowledge.
Freebase was originally launched to strong reviews in March, 2007 (with Michael Arrington saying: “this is cool unless it achieves consciousness and kills us all”) though Metaweb itself was founded in July, 2005. (And I just updated the founded-date field in Freebase for Metaweb which previously just said 2005.)
I haven’t spent that much time on GoogleBase, but I’m not all impressed with what I see when I go there. My take is that Freebase is what GoogleBase wanted to be.