IMHO this is irrelevant to what is being discussed here.
> enables devices powered by Snapdragon 800 processors to be “woken up” by a custom voice command (beginning with a custom phrase set by the device OEM like “Hey Snapdragon”, for example)
Because they are always on and have to use minimal power, these chips are used to recognise a very specific term (probably why these companies don't let you customise how you summon the assistant), and cannot do anything like voice transcription, network calls, ... (which would just kill your battery).
Why would audio have to be transcribed locally on battery power? You could do that server-side and/or keep the data stored locally as audio until the device is charging and on WiFi.
Record audio data continuously on disk?
I'm not sure if this is even worth debating, it is a fact that this is not happening (these chips are on most new Android phones, and they do not save audio).
Wait are you trying to say there’s no way to record audio on a smartphone app? And also, why are you calling it “disk”? Smartphones use flash memory for storage.
What minimal power? Phones are being charged at home all the time. Android apps can prevent sleep. Dsp to parse voice is there. There are sdks for all this.
I fully agree with you on that. However, please grant that mobile devices spend a substantial amount of the day plugged in and can draw full power (people mining coins, etc).
interesting. thanks... I wonder if they have something like this for ios devices? seems like they have similar capabilities via the "hey siri" or whatever it is but I am curious if there is a hack to allow it to use that all the time from userland.
https://developer.qualcomm.com/blog/using-qualcomm-hexagon-s...
"Industry wide adoption"
2013
The rest is PR.