Turn it on it's head: how else will you enter data?
I can see non-keyboard interaction becoming the dominant paradigm for browsing in quite short timescales, but for data entry/capture some form of keyboard (virtual or real) will most likely be around for a long long time, even with voice recognition systems becoming more and more sophisticated.
Why? Even with intelligent contextual understanding for grammar (e.g. 'weight' versus 'wait' etc) there will most likely be circumstances where talking aloud into a device isn't practical, for example in a noisy environment where ambient noise is a problem, or in a meeting where it isn't appropriate (imagine everyone talking into their notebooks!).
Ironically I'm guessing that many non-English languages with extended character sets that aren't naturally conducive to keyboard input (e.g. Japanese) would be faster to transition to keyboard-less interaction. I believe that many of the asian/pacific rim languages drove heavy adoption of the fax over e-mail because of this (anecdotal evidence) and this same phenomenon would be responsible for the keyboard-less interaction.