Posted on 2 Comments

Laughing/Crying on your Monday with Vista Speech recognition

Nothing to do with MySQL, but I just love observing how technology can make us waste more time. Speech recognition is a wonderful idea, but …. oh just watch this:

Of course, watching this video to the end will cost you about 10 minutes, but it could be worthwhile education. Technology can be very useful. But that doesn’t mean that all new stuff is useful, or even suitable for your purpose. I don’t think that guy was doing anything particularly complicated or wrong though, and I also commend him for his patience – although you can clearly note his frustration as the experiment progresses 😉

Posted on 2 Comments

2 thoughts on “Laughing/Crying on your Monday with Vista Speech recognition

  1. Rewind our memories back to 1996…

    “Merlin” – OS/2 Warp 4.0 Beta. With integrated ViaVoice speech recognition in the operating system. Your voice could be used to control every aspect of the system without any need for a mouse or keyboard.

    With one minor oversight: The US Beta was completely unable to cope with British accents. After going through the training process – we decide to give it a spin. A few false steps, we successfully open up Lotus Word Pro and try to bash out a short “letter”.

    Of course, this was 12 years ago when speech recognition was still quite primitive. A few badly transcribed statements later, we start laughing. The laughter is further transcribed into further hilarious sentences. By then, we are bent over double in laughter. Peeks at the monitor caused us further laughter and breathing is barely possible.

    I must admit, later when the GA release of OS/2 Warp 4.0, even though I had purchased the UK English version, I never did retry the speech recognition capabilities: The experience has caused me to write it off and I have been very skeptical of any other such “alternative” input technologies. OS/2’s handwriting recognition was much more mature but required a tablet to operate.

    Until it just works without any training and is able to cope with our ambiguous and context-sensitive language without doing the wrong thing, I don’t think it will work. For it to work, I personally believe that we need to develop some form of Strong AI. Strong AI is something much more sophisticated than a primitive program which can pass the Turing test.

  2. gold

Comments are closed.