Earlier quoted context omitted.
And I view the voice command thing as a thing to be avoided in most or all cases. I simply don’t want to talk to electronics. The privacy implications alone scare me away. No thanks.
> I simply don’t want to talk to electronics. The privacy implications alone scare me away. No thanks. I think you should consider the tradeoffs here. Practical safety on the road today, where I don't die because I'm fiddling with buttons, trumps theoretical unmeasurable fearmongering about vague privacy concerns. The reality is that if Tesla were to record every command I ever give the car and post them online again…
The way echo speakers are implemented is that the device only determines if you used the wake word and then sends it to the cloud service. However this wake word detection is sloppy and turns on more often than it should. It's filtered out on the cloud service side, but it creates a situation where extremely private moments are uploaded. For end users this means that you can't be certain when it's recording vs when it isn't....
IDK how it's implemented in cars, but the STT features I've seen require the explicit press of a button and IIRC it's inferenced locally on the car's hardware. Which is possible thanks to the car's high energy usage and price. But it might depend on implementation and it might send telemetry back to the manufacturer.