Everyone is talking about that Windows Vista Speech Recognition demo at the Financial Analysts Meeting. For a while the CNBC coverage was the #1 download on Google Video; on Youtube it's been viewed over 40,000 times. There are at least two T-shirts for sale. Rob already explained what happened, and so did Larry Osterman (the developer from the team who does Windows Audio).  But my Mom doesn't care. She read about it in the paper and wants to know if I'm in trouble.

Well, Mom, the trouble is that Speech Recognition is easy to pick on, because it produces hilarious results when it fails.  Did you see the Saturday Night Live skits?  Look at the attention Paul English gets for his tips on how to talk to a human instead of an IVR system. If something gets in the way of the microphone, the computer doesn't know what you intended to say any more than if you have your hands incorrectly on the keyboard or somebody bumps your mouse accidentally.  That will always be the case, no matter how much R&D we do in teams like mine.

People ask skeptically if speech is "ready for prime time" but when you actually try Vista I know you'll be impressed.  Maybe we can do some more with the UI to, say, alert you better if we notice your gain suddenly changed, or if your mute button is on.  But if you've set things up correctly, it'll work great.

So am I in trouble? Maybe I'm working in a field that will always get more attention for its failures than its successes.  Should I have become a doctor instead?

(by the way, I dictated this post using Windows speech recognition.  I’m using build 5472, which is actually an earlier build than the one that was shown at the meeting.)