Dave,
You're a little behind in the voice recognition. Large vocabulary continuous speech recognition is already available from, e.g., IBM and Dragon, but its recognition is still only in the 90-95% range -- not good enough to really take off. Also, there isn't sufficient natural language processing (that is, not just recongizing ordinary speech, but knowing what to do with it once one has a string of words). Our user interfaces are still mouse and keyboard centric by design -- the ideal UIs will be designed from the ground up with casual speech in mind.
These improvements aren't here today, but they will be tomorrow. And when they do show up, the world will start to become a very different place.
Such a change is going to take enormous computational power (both CPU and RAM, almost certainly). But it'll be worth it -- easily.
--John
P. S. IBM's recommended system for its current ViaVoice continuous speech package is, as I recall, 32 MB with a 200MHz or higher MMX chip. We can expect these requirements to continue to grow rapidly as people demand and the software gains capability, and as speech facilitates greater use of other application software. |
