Speech recognition software used to be awful. It couldn't deal with background noise, its vocabulary was limited and it was awkward to use in public.
But we've come a long way. A new speech recognition system can transcribe English or Mandarin about three times...
