By IANS
To filter out background noise, a user's voice is sent from the device to a Google data centre, which is where it goes through the machine learning model.