4 ms·

I understand how (a) discoverability and (c) recoverability are a problem, but what do you mean with (b) efficiency? Most people talk faster than they can type

• threeseed 2y ago
  But that assumes that the speech recognition is perfect. Which at least for those of us speaking non-US English is never the case. And you have only to ring your bank and try and transfer money between accounts with it reading out every account number and asking for confirmation every step of the way. Versus a few clicks with a mouse to see that for almost all operational tasks voice is cumbersome and inefficient.
  • danielbln 2y ago
    Whisper is incredibly robust, with a vast amount of language. I use it in German as well as English and it's incredibly reliable. Modern, transformer based ASR is a different ballgame.
    • threeseed 2y ago
      It needs to be perfect. Every single time. Because voice interfaces don't have the equivalent of a delete key or allowing the user to quickly select a different option.
    • soco 2y ago
      Time for the mandatory mention of the Scottish voice recognition elevator sketch: https://www.youtube.com/watch?v=MNuFcIRlwdc https://www.youtube.com/watch?v=MNuFcIRlwdc
• Karunamon 2y ago
  That assumes perfect accuracy. If a command is misheard then you probably need to correct whatever is now in the wrong state, and then definitely reissue the original command. If its text input then you have to do some select/correct dance. Both of these things take a lot of time.