This is a really neat development, I thought. It's not really designed to compete with anything like Google, but the University of Washington is attempting to explore the "next generation of search engines." There are still a few anomalous entries that don't really make sense, but it's fun to play around with!
For example, you could fill in the fields:
Arg1 -- Einstein
Predicate -- Born
and it would return a list of tuples that it thinks describes your specifications.
There was a lecture on this on TV today, which is where I heard about it. I'm not sure how famous it is, but I found some of the details pretty interesting. For example, it is a "self-sufficient" application. It learns on its own. It decides what to learn and how to learn it. That's obviously a huge understatement of what it is actually doing, but none of these aspects were covered in much detail.
Anyway, I thought I'd share.
how does this differ from modern day search engines?
http://turing.cs.washington.edu/papers/ijcai07.pdf