3 ms·
Except that someone would have noticed if all sound was sent to NSA HQ (or whomever). With these mics, that's by design.
by lucb1e 10y ago
Except that someone would have noticed if all sound was sent to NSA HQ (or whomever). With these mics, that's by design.
- bryanlarsen 10y agoIt's fairly easy to tell that these devices do not send data until woken up by their keywords.
- lucb1e 10y agoThat's different from what I previously heard. Mobile devices perhaps (because limited bandwidth and battery), but about these mics (and the "smart" TVs) I heard they upload everything and detect the trigger word online as well. More processing power, everything centralized, I get it from a software engineering standpoint, but it's a privacy nightmare. Next time I read someone claiming this I'll look or ask for a source. Do you have one that says they don't start sending until after the keyword?
- wccrawford 10y agohttps://www.amazon.com/gp/help/customer/display.html?nodeId=201602230 https://www.amazon.com/gp/help/customer/display.html?nodeId=... > 1. How do Amazon Echo and Echo Dot recognize the wake word? Amazon Echo and Echo Dot use on-device keyword spotting to detect the wake word. When these devices detect the wake word, they stream audio to the Cloud, including a fraction of a second of audio before the wake word. Maybe the others are different, but Amazon has claimed that they detect the watch word locally and then start sending, but with a buffer that sends starting just before you said the watch word.
- lucb1e 10y agoThanks!
- dingaling 10y ago> It's fairly easy to tell that these devices do not send data until woken up by their keywords. Indeed, we can confirm that they don't upload until woken. Now, can we determine what they do upload at that point? Perhaps the previous two minutes of conversation, recorded for context to 'improve the consumer experience'? We just don't know. Back in the early days of HTML forms on the Web, we were told to double-check the domain and SSL status before pressing 'submit'. Then JavaScript arrived and websites gained the ability to stream the data from text fields before the user had even submitted the form. I see a parallel with these in-house devices, they have a great deal of privacy-violating potential that runs contrary to people's expectations that a device only starts to perform a function when commanded to do so.