The problem with voice recognition software that doesn't require a physical action like pressing a button (Siri) is that they do a terrible job at distinguishing between when your actually trying to talk to it and when you or audio/video say(s) something in conversation to someone else/playback that might sound like the command keyword.