Many people are already psyched about Natal's potential. Microsoft had a few tech demos to show what it has in mind, including a painting application, a block-breaking game where you have to knock balls with your body, and Milo.
Milo, an experiment put together by Lionhead Studios, shows how an eerily life-like 21st century digital boy can interact with a user. Using a camera, Milo detects motion, spacial relations (if you're near or far), and can recognize vocal patterns (are you scolding the wee lad?). One thing I'm doubtful about, though, is the alleged ability of the camera to scan a hand-drawn image in the blink of an eye.
However, the project is far from finished. I spoke with one of the people behind the techology that makes Natal possible, and he confirmed that the demo has a couple of "tricks" that help it work right now. No way could you wave an image right up close against the camera (the focal point of the lens wouldn't be able to adjust) and in a second have it process the image. This capability is more of a target goal.