Agreed. I don't see steering wheels disappearing from racing games or thumbsticks from first or third person shooters for example. It doesn't preclude a camera system from being an "enhancement" to the normal controls.
I should have cleared that up as to not meaning a replacement for the majority of traditional "core" games you usually see on PS3 and X360, but as a replacement for new types of games or certain types of current games.
Their use cases are pretty similar where existing gaming content are concerned. What you mentioned is already here today. That's why I said EyeToy is here first, not because of some meaningless "I'm here first 2D/3D" claim. e.g., In SingStar, you can use voice control or the DS3 to pick song. It makes sense because the players have a mic on hand. In MLB 2009, you use your voice to "teach" the crowd cheers. In EyeToy mini games, you flail around. In Eye of Judgment, the PS Eye recognizes playing cards. In EyePet, you draw objects for Milo^H^H^H^Hyour pet to play. Then in Eyedentify, err... it's like Milo. Many of these don't require accurate 3D positioning. But it's more usable if it does not rely on good lighting (though some applications will require reasonable lighting like color and handwriting recognition).
Traditionally, a key difference between MS's and Sony's approach is consistency. Even if natural interface/augmented reality is added to games "here and there" like Sony, MS will probably make the experience more common and consistent (e.g., integrated into the Blade and XBL UI). This gives the impression that the technology is implemented "everywhere". Since the same gestures can be used in isolated apps, it also simplifies the user experiences and makes it more seamless/pleasant. I think this is the key advantage for MS (Sony is pretty much helpless here).
In any case, a camera based solution is much much more than motion sensing. In fact, Milo -- which does not use 3D motion sensing -- generates more hype and appeals to more people than the flailing body mini-game. If you read the interesting papers in the PS3 controller thread, you'll find advanced image processing techniques to interpret a video too (e.g., deciding whether a burglary is going on in a gas station using a regular security camera). Personally, I am more interested in Natal's 3D imaging capability (even without the generated skeleton).
Eye Toy has had potential, but Sony haven't really done anything with it other than a niche game here and there. Hopefully that changes with the release of the PS3 wand.
PS Eye/EyeToy's potential is still being realized.
For broad use, some of the applications are actually filed under Cell programming rather than EyeToy games. The magic is done by the CPU. Unfortunately, for fast, precise, straining, or time consuming activities, the controllers have their advantages (Wii wins). This is probably why Sony chose to extend and integrate the controller experience better with PS Eye here. Their eyes are on gaming.
I actually think that Natal and PS Eye have more interesting applications elsewhere. I am sure we will see very interesting use cases in the near future. Some will be further out. I'm still waiting for Magic Mirror and Eyedentify for example. In particular, Magic Mirror is the perfect app for a 3D camera. I suspect this will be one of the launch apps targeting female non-gamers for Natal.
As for marketing, price point and focus are critical. I have no idea what Sony's plan is (other than trying to make existing and new games work with the new control scheme). Microsoft is going to push for one umbrella and concerted Natal concept to the masses, and that is its strongest card by far. As some pointed out earlier, marketing and software will be the key to winning.
