I talked to an Amazon Echo engineer about how the sound recording works. They said there is just enough hardware on the device to understand "hello Alexa" and t

I talked to an Amazon Echo engineer about how the sound recording works. They said there is just enough hardware on the device to understand "hello Alexa" and then everything else is piped to the cloud. Currently, ML models are too resource intensive ($$) for always on-recording.