Ask anyone to describe a computer of the future and they probably won't even mention that it can understand speech - they'll take that for granted. It's only the clunky computers of the real world that need keyboards. Science fiction movies always show users interacting with computers by speaking with them. Twenty years ago there was little doubt that by today, universal speech recognition would not only be possible, but standard. What happened?
Yes, there are well-known systems like Dragon Dictate and IBM's ViaVoice. I'm told they work efficiently, but I've never had any success, so I guess I talk funny. Now we hear that a number of PC manufacturers will be putting speech recognition systems into their 1999 models as standard. How will that be for a support nightmare? Imagine trying to explain to a dissatisfied customer why the computer isn't doing what they expect. Studies have shown that people perceive a computer with voice input to be much smarter than one that only has a keyboard. They see the manual-input machine as something they have to "operate", but they see the speech input machine as being able to do things such as create a document, without them even having to start a word processor.
Let's take for granted the obvious, excellent uses of speech input for people with disabilities. But for the vast majority who could use conventional types of input device, if speech input is used to control the PC, what will they expect it to do? Training users to replace mechanical input with vocal equivalents is silly. Imagine an office where everyone spends the day going "Tab, tab, open, file, open, down, down, down, down, down, down, down, open."
The real power of speech control will come when the computer DOES try to work out what the user wants to do. When the user says "start a fax to send to my accountant" you can imagine the system working out that it's supposed to start the word processor, decide what sort of document template to use, load the details of the accountant, check that the modem is still connected and so on. The trouble is, none of this has anything to do with speech input - it could just as easily have been typed in.
So what will you be telling customers next year when they ask why they should ditch their old machines in favour of the latest model that has a microphone built into the keyboard? Oh, and by the way, if you think some people get a little worried about what personal information the vendors learn when the computer does an online registration of new software, imagine trying to convince someone that the microphone built into their new PC isn't being monitored somewhere!