1) it's baked into operating systems as features that most people don't use:
http://lifehacker.com/215764/hack-attack-make-your-macs-spee...
http://windows.microsoft.com/en-us/windows7/set-up-speech-re...
2) there's already an ecosystem of dictation / voice-assisted software that the overwhelming majority of the world doesn't want:
http://alternativeto.net/software/nuance-dragon-naturallyspe...
Joe Random Public can convince Google they want this as soon as someone convinces Joe Random Public they need this.