Speech and Home/Office of the Future

Over the past month or two I've taken a few guests on walking tours of the Microsoft Home and then the Office of the Future.  You can see much the same thing for yourself on line, but now Patrick pointed me to this article from The Times UK that claims Microsoft's Jim Alchin (aka "Vista Godfather"--though that's the first time I've heard anyone call him that) says:

machines that actually understand speech - rather than using sounds as prompts for pre-defined actions - would define the next industry breakthrough.

I'm not sure I understand what that means.  The general case, of course, is what we call "AI-complete": if you can make a computer that really understands speech at the level of humans, then you will have yourself a human, not a computer.  We understand a lot of dialog because we are humans, who grew from infancy to adulthood in the presence of other humans, had experiences like getting dumped by a girlfriend in high school, etc. etc.  -- and all of those things combine to make us who we are, and able to communicate with other people. 

I recently listened to a podcast by Jeff Hawkins, who sounds like he's found a hack that gets us close. But there's a difference between the kind of super-fast recognition tricks that Numenta is reportedly working on, and the much more complicated activity that happens when we speak to other people.