A 12-year-old girl in Tennessee has been charged with murder, accused of smothering her 8-year-old cousin as the younger girl slept. A relative said they had been arguing over an iPhone.
A security camera recorded the killing, inside the bedroom they shared on July 15 in Humboldt, Tennessee, the county prosecutor said.
The recording shows the older child using bedding to suffocate her cousin as the younger girl slept in the top bunk, Gibson District Attorney Frederick Agee’s statement said. After the child died, “the juvenile cleaned up the victim and repositioned her body,” Agee said.
A relative told WREG-TV in Memphis that the girls had been arguing over an iPhone after coming from out of town to stay with their grandmother.
Only if you choose to ignore the explanation.
The device listens locally for the trigger word. When it hears it, it wakes up and starts processing what you’re saying, sending that to Google or whoever and getting a response back. It isn’t constantly processing or what you’re saying beyond “was that ‘hey Google’?”.
That’s because the explanation is often a bit disingenuous. There’s practically no difference between “listening locally” and “constantly processing what you’re saying”. The device is constantly processing what you’re saying, simply to recognise the trigger word. That processing just isn’t shared off device until the trigger is detected. That’s the claim by the manufacturers, and so far it’s not been proved wrong (as mentioned elsewhere, plenty of people are trying). It’s hard to prove a negative but so far it seems not enough data is leaving to prove anything suspect.
I would put money on a team of people working for Amazon / Google to extract value from that processed speech data without actually sending that data off device. Things like aggregate conversation topic / sentiment, logging adverts heard on tv / radio for triangulation, etc. None of that would invalidate the “not constantly recording you” claim.
I will never trust that ever