Hacker News Reader

Comment by marcinzm | Hacker News Reader

>all conversations

I believe Alexa only streams what you say after the trigger word and not all conversations. Assuming you believe Amazon.

View on HN

if its technically possible to do it without you knowing then all that is left is how much you trust Amazon with that. Note that Amazon may have to surrender its devices access to three letters agencies for random reasons as well. You have to live with that assumption.

It's widely understood how the wake-word system works. You can monitor how much data is being uploaded and when. So no, it's not possible to just stream everything you say undetected.

I believe 99 percent of Alexa users wouldn't notice the increased data usage. If "three letter agencies" carefully choose who they spy on, it might go undetected for a very long time.

If a three letter agency is interested in you then you would be lucky if your Alexa is the best bugging device they have.

Sure. But most people trust corporations with almost all their private correspondences, photos, internet browsing history, shopping behavior, etc. So at that point how much is there to lose by adding everything you say into the mix as well?

Some companies, notably Google and Amazon, have maintenance of customer trust as a core principle. This seems to be working. Generally, people trust a lot of their private data to these companies and the companies do not abuse it.

Microsoft believes in this too, but had to learn the hard way back around 2000 that customer trust matters a lot.

Facebook seems to have had a harder time learning this lesson. Perhaps they've learned it now, but it'll take a while before consumers really believe it.

I notice Amazon's wording about this is pretty careful. The device could still be processing conversations locally to some extent and still be adhering to the wording.

It records every conversation locally in a two second loop I believe, and then streams it after a trigger word.

Correction, it keeps a rolling loop going so that when the CPU hears the wake word it has enough audio buffered for voice recognition to have a clean starting point.

If it only started recording when the wake word was processed as being heard there would be an added delay before you could start speaking a command. Might as well push a button and wait for a beep for that added delay.

And the wake word detection is done offline, as can be easily proven by disconnecting your internet and saying 'Alexa report this conversation to the NSA' - she will wake up, but then blink red due to lack of network connectivity.

Minor correction but it could still do the wake word detection locally, then fail with the red light because the "report this conversation to the NSA" part could not be sent.

Exactly. Everything after the first trigger word you said when you unwrapped and plugged it in is streamed.

Reply on news.ycombinator.com