If you look at google and search, we forget the past and the problem its trying to solve. Fundamentally, search is a solution for finding relevant content in a web universe of size X.
When the size of the web was two orders of magnitude smaller (say 1998), then directories were good enough (yahoo, dmoz). There just wasn't that much stuff there.
Google thrives because the size of the web grew 2 orders of magnitude.
The obvious question to ask is, what will happen when the web grows another two orders of magnitude? Will keyword navigation be the best way to find content still? (typing two or three words in a search box). I'm not sure.
We all know that the ideal is AI and natural language processing. But I think the web will grow faster than our advances in nlp interpretation.
I'm guessing that social relevance will be very important in helping guide navigation to content. How to blend all these approaches in a reasonable UI is a nice challenge. I'm excited to see how the web evolves.
By the way, if you like thinking of these kinds of things, I recommend you read nicholas carr's blog in www.roughtype.com. He has some interesting ideas.