There's another Microsoft search beta out that's worth a look. I think the "results ranking" sliders (under "search builder") are cute but ultimately useless; I certainly consider myself a "power user" but I have never even used Google's "advanced search" page, nor seen the tilde operator. Why? Because the results without any tweaking are good enough to get me what I want. Whenever I see user interfaces like these I remember the section from Joel Spolsky's book about how your users have better things to do than learn the details of your program; they want it to just work. And depending on how well it works for them, people learn less and less, and ultimately that's the way it ought to be -- these tools are only a means for solving greater concerns.
However, they (back to MSN search) do have a question-answering system, though I can't quite figure out how to trigger it except for using the questions I see in the press about it. This sort of thing, if done well, can potentially revolutionize search: search engines are already information portals, but few have been able to figure out how to provide answers directly instead of a list of potential answers. On that note, I got to catch a talk by S. Narayanan yesterday at Stanford about his research at the Neural Theory of Language project, which combined some linguistic theory, some cognitive science (including a slide with MRIs to support an argument(!)), and a lot of computer science to make what sounded like a pretty sophisticated system. Unfortunately, he didn't present any results, though it looks like he has some on his website.
