Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I personally assume that they wouldn't be processing all sound at all time from a device because it would be massively expensive for them. Obviously, it's technically possible that they could be, but I don't see how it would be financially possible. If we talk about targeted surveillance, then that is certainly financially possible. But I find it unlikely I would be targeting among the millions of people using these.


Advances in AI are bringing down the cost of processing audio by perhaps 5x a year, and has been doing so for several years now.

The risk is that it becomes worthwhile for someone to record everything now and process it later at their own leisure.


The primary cost in this case is raw storage and bandwidth. What is AI doing to reduce that cost? Also what source confirms that “processing audio” is 5x cheaper because of AI? Seems like a dubious claim.


I don’t know why anyone in tech believes the myth that their devices are streaming everything.

It’s trivial to inspect traffic from a device. You can see that your devices aren’t constantly streaming data somewhere. You can see a burst of activity after a command.

Beyond that, it would require a massive conspiracy for everyone who ever worked on these programs to never leak details about always-on surveillance recording. It would be a bombshell revelation and you’d see it everywhere.


> It would be a bombshell revelation and you’d see it everywhere.

What a strange idea. From what I remember about the "privacy nightmare on wheels" disclosure [1], Americans have apparently already consented to being recorded at all times by simply being within range of cameras and microphones in cars they don't own, passengers consent to losing control of their DNA by sitting in the car, etc.

It's better for Europeans for the time being. But short of collaboration with extraterrestrials to beam up millions of New Yorkers for medical experiments next Tuesday, it's pretty hard to imagine at this point that any example of corporate abuse will qualify for "bombshell revelation" ever again. The overton window has well and truly shifted, and most people will basically argue with you that they enjoy the abuse.

[1] https://foundation.mozilla.org/en/blog/privacy-nightmare-on-...


> What a strange idea. From what I remember about the "privacy nightmare on wheels" disclosure [1], Americans have apparently already consented to being recorded at all times by simply being within range of cameras and microphones in cars they don't own,

Your memory is incorrect, or at least misguided. There's nothing special about cars, doing things in public places does not come with a reasonable expectation of privacy.

The flip side of this debate is public photography: Should photographers (including people using their phone) lose their right to take photos in public places because they can't collect consent from everyone around to have their photo taken?

> passengers consent to losing control of their DNA by sitting in the car, etc.

I don't know where you're getting this one. It's definitely not supported by the article you linked.


> There's nothing special about cars, doing things in public places does not come with a reasonable expectation of privacy.

You're probably talking about pedestrians on the sidewalk (which I also disagree about) but are you going to say the interior of the car is not private? Why would a home be private if a car is not? Spying in the kitchen but not the bedroom? Spying in the bedroom as long as it's audio only? Is my backyard private? Is anything?

> I don't know where you're getting this one. It's definitely not supported by the article you linked.

Dig deeper then. I'd find the reference for you, but when you're confronted with it then you'll just probably just argue it's standard for shrinkwrapped "all rights reserved" kind of legalese and it doesn't matter that we're expected to tolerate this kind of bullshit because we don't have the technology to stick a DNA sequencer under the seat. But that's exactly the point. The abuse is SOP now and for whatever reason, some people will always make excuses for it.


So record everything and upload it along with the legit uploading. Just hide it in there. Seems trivial.

You could even convert to text first to make the packet super tiny.


The more scalable way would be to have reconfigurable wake word hardware, then push target words, then send back analytics instead of audio.

Especially since Amazon already has advertising profiles on users, so could generate monetizable target word lists per-user.


> So record everything and upload it along with the legit uploading. Just hide it in there. Seems trivial.

You really think this would go unnoticed?

> You could even convert to text first to make the packet super tiny.

This entire debate came up in the context of them having to upload audio clips for server side processing.

So which is it? They’re uploading Audi clips? Or they’re converting everything to text?


Doesn't it have to listen to everything locally for a command though? I wonder if only commands are keywords.


The 'wake' (eg 'Alexa') trigger match is very fuzzy (which is why other words can sometimes set it off) and there's no analysis beyond looking for that particular trigger


The first point is obviously true, the second point is just an assumption you're making and trying to dress up as fact.

Can't imagine why anyone would want to do that, unless maybe they work for Amazon in a capacity related to this data?


The devices are trained to trigger on wake words or phrases.

It’s much easier to run an on-device model to match a very specific phrase than to do speech to text of the entire language.


Yeah, a more likely exploit would be it starting to listen for other wake words (different from the original one) and then recording whenever those words are said. Still would be a massive conspiracy lol


> Still would be a massive conspiracy lol

Exactly. Tens of thousands of people have worked on these products. Many of them ex-employees now. Many disgruntled, even.

And we’re supposed to believe they’re all united in keeping this a secret? Nothing has been leaked, no security researchers have found anything, just a perfectly hidden conspiracy sitting in millions of homes?


> Tens of thousands of people have worked on these products

I agree with the sentiment, but that seems like a vast overestimation on who would actually know what the device collects. More sophisticated things than this have been made with less. Programmers often have siloed concerns, and PMs only know what their incentives are. It's not hard to imagine less than a hundred people knowing, and what percentage of those people care about us?

Do we have some data that suggests tens of thousands of people have had an integral role in Alexa?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: