These're posts that are related to your question:
a note about "human evaluators":
http://googleblog.blogspot.com/2008/09/search-evaluation-at-google.html
term "universal search":
http://googleblog.blogspot.com/2007/05/behind-scenes-with-universal-search.html
http://googleblog.blogspot.com/2008/07/technologies-behind-google-ranking.html
http://googleblog.blogspot.com/2009/05/more-search-options-and-other-updates.html
To answer your question simply I think that you type your question to google and google tries to guess what type of content you're looking for - news, images, information on web, movie clip and so on. The problem is obviously how to "guess" the content. Well, you may use:
1] pagerank - what is popular
2] set of rules (don't forget that Google has huge databases so that the rules may be general - "guess the language of the query and search first in pages that are of that language")
3] several searches that differ very little => last search may be what user wanted to get with the first search. (some kind of evolution)
4] microformats
5] tags - Michael Jackson is probably a person (tags: person, celebrity, ...)
6] synonyms
...
So the system is very complex and tries to persuade the user that it knows you mean but it's no magic box. It's simply so complex that it persuades you wery well.
It's just my opinion how it all may work!