It is new, yes. It is different. Even clever.
But that doesn't mean folks are actually going to use it so much the game actually changes. The basic issue I think, oddly enough, is not of the interface, but of power.
Siri, as I understand it, it *not* hands-free. You have to touch the phone to activate the program, which defeats about half the purpose. And this is understandable, because having your phone *always* awake, always listening, will kill your battery in short order.
So, you're saved some typing and some mucking about choosing apps, which is nice. But game-changing? I'm not sure.
If you bring the phone to your face, Siri activates. If you hold the button on your headphones or bluetooth it also activates. Bear in mind, when I said hands free, pressing a button to activate it is not the same entanglement as having to hold a phone up to my ear to keep talking to you or holding it so FaceTime can see me and I can see it.
I can have the phone in my pocket and my headset on an generally have my hands free to drive or work on something.
As to game changer, my xbox has voice recognition. It's like playing an old text adventure with 2 word commands. Siri has natural language recognition which alone makes it easier to use.
Right now Siri is coded to connect to specific apps. So it is obviously limited in what it can do. Apple coded it to handle food, movies, appointments, and just about everything else is routed as a search to bing or wolfram alpha.
In theory, Apple could provide an API so Siri can connect to any registered app's command structure. So that bar can be minimal. though I think Morrus's point about which app to use muddies that up.
coding-wise, it's basically voice recognition wired up to natural speech parsing (so this technology could be wired up as a search engine). From that, it can recognize whats a command and what's a question. Questions go to a search engine, commands go to its command matrix.
In the command matrix, it's not as simple as performing a simple action like "Call Mom" which requires no memory, just recognition of the action "Call" and the subject "mom". Whereas, "Make an appointment for tomorrow" gets recognized and it knows it has missing fields (subject and time) so it will propmpt for them. The next part (which I'm not sure if it works as I don't get to play with Siri) is "Make a grocery list" followed by your reciting the items for the list which go into Notes. Then you could tell it "remind me the list when I get to the store". This requires context memory that you've been workng with the List the entire team.
In a way, lunquistically, pronouns are variables. Short term buckets for the topic of current discussion. While software likes variables, coding a parsing system to "remember" things is the big deal. All of this being wired to a command structure gives it the oomp it needs.
In any event, I think it's cool and a good many steps more powerful than Dragon, Xbox's kinect, or the usual voice commands on cell phones to launch a call.
I haven't kept up on voice recognition or natural speech technology. Dragon had long been considered the king, and it languished for a decade or so. Its still mostly a dictation software, not a natural speech processor.