I guess because they do not have a common memex yet. :-)
But yes, RDF should fit very well.
As others pointed out already, I also dislike the idea of being locked into a proprietary file format. This should not be a problem using RDF - I guess we all should think about that idea.
RDF is the technology behind schema.org, so that would be a good success story. There are quite a bunch of companies working around RDF based technologies, like TopQuadrant, StarDog, Ontotext, and more. But if you want to start playing with it, I recommend you start with Jena/Fuseki [1], it is super easy to run.
Example: say your database contains a note about "doves", and another one about, say, "seagulls". Later, you can search for things you have written about birds, and the query would find both your notes about doves and seagulls (something that you wouldn't be able to do with a fulltext search). There's a whole world of things you can do with inference, so using RDF is a no brainer when it comes to finding relationships between things. It also makes it trivial to import knowledge annotated by other ppl.
If you are interested, here are some other resources I found useful to get started:
This is book is old but Chapter 2 and 3 are a good intro on "machine-readable meaning" and Inference. The author also has another cool book on "collective intelligence" I still haven't worked through yet :-)
> Is there a general interchange format for these systems?
I don't understand why none of these people is using RDF. Seems like a perfect technology for a memex.