Touch-screen won't cut it. I'm not being sarcastic, but it's really too much work to lift your arms all the way to the screen constantly. I think you're on the right track about adapting to specific tasks. A normal keyboard--or more specifically the smaller smart-phone/blackberry keyboards--would need to adapt hotkeys for different apps/tasks, since many things done are not straight text-input. A single key for Copy, instead of Ctrl-C, for instance. The ability for the *user* to map these keys and their context.
As far as text/speech input, I still think, at this point, that low-tech is better. I think we'll be touch-typing for a while. I've seen the PDA's that "remember" the last words you typed that, say start with "St" or whatever, and then list them for you. Boo. Wrong. Too disruptive. Too Microsofty: "We do it because we can, and it's 'cool' to program this stuff". But not usable.
And as has been discussed, speech input is way, way down the road, if at all. I think the concept of the keyboard will stay, but it won't be on-screen, it'll be something ingeneous (don't ask me what) that will be useable anywhere at a near-normal size, but physically portable in a very tiny size. Think of those rubber roll-up keyboards, but rolling up much smaller and folding out with much more ease-of use. Probably not even mechanical--but touch/optically based pads. I don't know. I guess that's why I'm not the guy inventing it and making billions.
--Jim
