you are very right. At 90s ibm was selling os/2 lan server with motto "voice controlled" and it just worked! But os/2 was marketary unsuccessful.
ok, but do the people really need to speak to devices at oral? same way as to speak to earwear bluetooth devices? We are still very far from using some stuff for those even when it is both possible and should be convinient for everyone. Back to nowadays we are using web appliances more and more, but shouldn't hear the voice from them except very special cases like the CAPTCHA.
i guess the real reason for both 90s and the now time not to communicate to devices via the voice is not really the technical problems. It's just the social kind of problem. for example, how the device could be sure about your personal identity? For the LAN server such a problem is used to be solved as a physical access to authroized personnel only. But the MID is not just the case about, ok? Even with proper tune on the perspn's voice recognition, no one will ever be sure on reproducing one's voice with pocket voice recorder ( should be funny with words "rm -rf / Enter" ).
ok, back to public kinds of problem, the one's voice communication to device is not the stuff to be expected in closer society. The particular, but not the main, is, say, for collaborative team work at office it is unusual for them to speak the simultaneously to do the job. And the sound-isolation of cubicles is far from best solution for team work.But this is not the very basic why such the behavior should be still considered marginal in the modern society. I mean as the main that vendors companies must reinvent the feature and promote the corresponding advertising company. Everyone should know now that voice controlling of devices is now not about those unsuccessful software platform like ibm-os2 and ms-windows but it is the very fashionable and glamour.
And after every indian schholchild will know about how cool is to talk to his/her OLPC, only after that the feature will be a must. While it's not a must, it's not a can/may.
To say the truth, i'm very far from thinking the voice controlling of devices is the near future. Why i said you're right about keyboards must die ( but must go )? Well, i think the highway road of progress in this industry area is to get away the movable parts those become dirty, unreliable and hard to make the cheaper. We are on raise of solid -state storage devices now --- well, the human interface parts maybe the next? And, the keyboard i expect from future production is the touchpad-like but with sensable buttons under the surface. Visible and sensible, both. I guess, there should be methods to output not on;y the picture on the touchpad, but put the softer and stronger areas on it to make the touchscreens buttons more sensible. And, those will be the keyboards to must to go.
Thanks for attention.

I'm not so sure that voice technology is all quite as useless as this article makes it sound... I know a number of children & young adults that rely on voice-recognition apps to do their "written" classwork for disability reasons all the way through college, which suggests to me that it might be one of the many cases where available tech simply isn't known to the non-disabled community.

Anyway... There would be a few huge drawbacks to eliminating keyboards that I think you're missing apart from the noise issue. One is that oral and written English have a number of differences, so things that sound well-done in one form often seem awkward if changed to the other. Other important issues is that many people can type a great deal faster (assuming use of at least a laptop-size keyboard) than they speak, and a substantial part of the population (including but not limited to the huge autistic-spectrum segment often found in tech) finds written communication cognitively a great deal easier to deal with.

All in all, I think that the voice recognition control (to replace keyboards) for computers would ultimately go over about as well as their parallel in phone systems -- and the same with touchscreens to replace mice, to be honest. While they're closer approximations to how we do things physically, they ignore the fact that computers were designed to work differently because it's faster and requires less effort. Consider, for example: to copy a block of text, we can either:
1) Speech: "highlight [X] then down three lines and two words to the right. Copy to clipboard."
2) Touchscreen: slide finger 2-3 inches (up then left/right), touch 'copy' button a few inches up
3) Mouse: slide mouse about .75 inch, touch 'copy' button perhaps 1.5 inches upwards
4) Keyboard: ctrl-arrow cursor to edge of text, shift-arrow down 3 times, shift-ctrl-arrow twice, then ctrl-c

Don't know about you, but those would have been in ascending order of speed/ease (i.e. 4 is best, 1 is least) for me, especially if each case were considered after a long writing session. Then again, when I really know what to say, I've been known to literally type so rapidly that it looks to others like I'm hitting all of the keys almost at once! ;)