My prediction is client side voice recognition, as good as this, in >2 years. Powered either by mobile GPU's and more efficient ML algorithms than we currently

My prediction is client side voice recognition, as good as this, in >2 years. Powered either by mobile GPU's and more efficient ML algorithms than we currently have, or specialist chips like http://www.movidius.com/ http://www.movidius.com/