You know it from your phone: tell a name, and I'll call that person for you. Example: I speak to the phone "mum" and it calls dad :). Joke aside, voice recognition on the phones never worked for me. But apparently the Windows 7 voice recognition is way better of what we had so far. And is usable to dictate text after some training.
When I think about voice recognition, narrative interfaces come to mind. The Quill prototype, for example finds a document [PDF] if you type in the story about it:
Another tool that comes to mind is Feldspar. Instead to typing the story, user incrementally specifies associations [PDF] about the searched-for information.
It would be nice to see how such interfaces would work with a good voice recognition. In the mean while, we cam dream about the possible future to be able to talk to computers like in Sun's Starfire.