MSNBC deceived the public: Vista’s speech recognition demo

Unless you’ve been hiding under a huge stack of free Ubuntu CDs, you would have known that last week MSNBC reported an ‘one-hit-wonder’ story about the failure of Windows Vista’s speech recognition technology. MSNBC called the demo an embarrassment, when technical glitches created many recognition anomalies resulting in several amusing representations of spoken words. Many mindless Microsoft bashers quickly erupted over this story, creating much harsh slander in and around the blogging networks.

But what MSNBC didn’t tell you is that the clip they showed was nitpicked from a largely successful demo. And here is the full video of the speech recognition demo.

As you can see, not only was the presenter able to complete much more complicated tasks in the beginning without any problems, but the technical glitches were quickly resolved and he was able to dictate the letter as planned.

Shame on MSNBC.

21 insightful thoughts

  1. You are correct when you say he performed difficult tasks earlier and was able to finish his letter as planned. Don’t people realize it’s still a BETA? I saw a news program and the news anchor slandered Microsoft on live television saying, “Now Microsoft isn’t happy we showed you that video and wants you to know that they are blaming ambient noise for the failure, but as you can hear it was quiet until the product didn’t work and everyone started laughing. Well live television’s rough, welcome to our world.”

    Very unprofessional.

  2. @Zef: MSNBC is a partnership between MSN and NBC. I presume NBC writes the news, MSN publishes it to the web.

    @Mike: Yeah, I thought it was quite unprofessional how they presented the story as well. Almost like a little kid making fun of someone else.

    @Richard: I’ll keep an eye on it!

  3. Beta version or not, fact is that when you present to a group of market analysts who will make a buy or sell recommendation on your companys stock you better make sure what you demo actually works. If it’s beta feature and you know it might not work 100% then don’t show it.

    The real problem was that from the presentation you would assume it doesn’t work at all.

    Reporters being un-professional because they only look for the negative. Comeon’ guys… I hope you are not that naive that you believe in the professional ethic of mass media journalists? Especially journalists working for a US media outlet.

  4. I’d say the criticism was deserved for the performance. One line commands such as “open word” are not “much more complicated tasks” and it’s not surprising they worked. The first attempt was miserable, and the second attempt was almost as bad. How can you say the technical glitches were resolved when they plainly weren’t?
    In the second attempt, the software made many mistakes.
    he said “Dear mom”, it wrote “Dear ma”
    he said “comma” twice, it never put a comma in
    he said “I hope you like the DVD”, it wrote “I hope you let the DVD”
    he said “Shannon”, it wrote “Shanen”

    4 errors in as many lines is pathetic for a speech recognition engine. From your demonstration it seems the presenter just had bad luck, but the fact remains that the demo was a disaster.

  5. @Leo. True. There were lots of errors, and I think it was still the equipment to blame. But speech recognition is a complicated process, and I think it has come a long way even if it still quite incapable in that demo.

  6. Fact from a Microsoft insider: The real screwed up was that we used a newser build for that demo and the gain on the input was screwed up. You can see that on the screen the red button is constantly lighted. So basically that poor Vista machine must have felt like being blasted by a 1Kw loud speaker at one foot range, thus couldn’t make up what was being said :)

    – A Mac loving Microsoft employee

  7. no offence… but who runs a new product demo to financial analysts none the less without a dry run making sure everything works 100%. You can’t blame the equipement… you have to blame the operator.

    oh, shouldn’t using a newer build actaully work better than an older build? And how can the input gain be screwed up on a new install? I hope Microsoft doesn’t have all volumes levels set to maximum by default.

  8. why is this coming out now, all this was avalabul 3 years ago with Dragon Naturally Speaking, all the same things, this is just a copy of it

  9. @Leo: “he said “Shannon”, it wrote “Shanen””

    Actually, that’s how the guy doing the demo spells his name.

  10. I think most of us tried Speech to Text at least once
    I trained my PC to recognized my voice (reading hours of famous phrases).
    Although about as similar as the demo it will be awhile before speech-to-text becomes truely up to speed. Put it this way, take “any” speech-to-text dictation software and turn it on, then call someone on your cell phone and talk naturally to where your PC can hear you.
    What you said on the cell and what the PC interpreted are probably 2 totally different things

    I wanted to do dictation like that demo – but realized that
    Dear Mom,
    We hope you like the DVD
    We miss you

    …takes about 3 minutes where as typing take about 6 seconds

    and god forbid you get a cold…
    CUT and UP sound alike to the computer
    Cool technology but still in infancy stage….well I guess I am comparing it to Star Trek to where you walk in your house and say “Computer…Status report” and then you hear “You have 9 email messages 6 appear to be junk mail, The episode of Simpsons where Homer is on “the Island” is on at 6:30 pm tonight, several updates to installed programs are available but not installed by your request, and there was several power fluctuations reported by your UPS’s which coincide with the weather report of thundershowers this afternoon”

  11. Looking at the big picture… in the final years of his life, my father had Parkinsons Disease and was dependent on ‘Dragon Naturally Speaking’. Since he was disabled it was always an inconvenient task to have Dragon be a separate program to install on his PC etc. Microsoft “needs” to work on this path because it will benefit so many disabled people in the future once voice recog is fully integrated into Windows. Even if only HALF the demonstration had worked, my father STILL would have been excited and happy about this demo.

  12. What a bunch of pseudo ‘experts’! Spelt? News altering/making media pukes, is more the norm than the exception! Lorelei, you were right on target with your comment. I’ve been involved with computing technology since the 1960’s, wrote my first program in 1K of memory. I did extensive work in the area of AI and heurististics. Microsoft was a prime competitor of my former employer but I applaud their effort in making accessability a part of their computing platform. Don’t sweat the small stuff, millions of people are still in the technological stone age.

  13. Unfortunately integration into windows usually means less competition and therefore less improvement. If windows would offer a good interface for programs to take advantage of then third parties could easily develop programs to take advantage from that interface. Microsoft could of course also offer a product of its own, but integration into windows is usually killing for competition (e.g. media player and internet explorer).

    Yes, speech recognition is good for a lot of people (without it I couldn’t work), but please no integration… Will I be able to use a competitive product in the future that in some aspect is better for me? Who can give me that guarantee? Perhaps dragon naturally speaking will die because of this… but at the moment it works better for me than windows!

    On Linux developers create interfaces too, giving us, the consumers, much more freedom of choice. I hope a competition authority will prevent overintegration (I’d prefer to install a program separately once to having no – realistic – choice).

    By the way, I donate speech to the VoxForge-project ( to enable more competition in this area and I encourage everyone who wants to see more progress in this area to do so too.

Leave a Reply