Alexa feels like like one of those devices that everyone (tech & non-tech people alike) understand to be an always-on recording device, and a proof point that many people do, in fact, choose convenience over privacy.
Not to be facetious, but why exactly do you trust Apple to delete your data after some time? What has Apple done or said to earn your trust? Genuinely curious.
They've got a pretty stunning record for not keeping data around longer than needed. They build their products at the hardware level (secure enclave) so they can't even access certain types of user data.
Their business model doesn't involve monetizing user's data, and they've been public and explicit about that.
Sure, someone like Apple may have a strong brand. But once it's built up, it's an attractive asset to deplete, the way corporate raiders go after a company with a lot of cash (which Apple also has). Many companies have followed a pattern of building up a brand over a long period of time and then hollowing it out and degrading it. GE home appliances come to mind.
The other thing is that sometimes a company does something by accident that would damage the brand if anyone finds out, and covering it up is the only way to save it. I don't think there's any way to prevent this from happening somewhere at a huge company from the top, because people are always going to make mistakes and bad decisions and then they don't want anyone to know for fear of consequences.
1. Non-tech people, who don't understand the immediate technical aspects of what's happening, and haven't had much occasion to think about the implications.
2. Tech people, many of us who do understand a lot of what's happening in technical levels, but are conditioned by all the practices around us (and our paychecks/startups depending on it) to not think of this as bad for society.
I already have phone in my pocket with Google assistant (yes, you can switch it off, but is it truly off?) Thats why I decided that Alexa is not a big deal.
Phones have a security/privacy advantage, because being battery powered and (potentially) using metered data, the cost of recording everything and shipping interesting parts of the audio to the corporate mothership would be immediate, in customer complaints about battery life and data usage.
Wifi/ethernet and AC-connected devices change the risk/benefit calculation significantly.
Alexa or Google Assistant on a phone is therefore not really the same thing as Alexa or Google on a standalone plug-in device.
For example, a google phone (depending on model/setup) and Google mini might both listening all the time for "Hey google". Both respond by listening and recording the next collection of sounds. However, the bar for deciding when to ship that audio off to Google, or when it's confident enough it can handle processing on-device, could easily change depending on whether the device is battery-powered or whether it's on wifi/eth instead of mobile. A wall-powered device might opt to keep a running history in ram of the last N seconds of audio, so it has more context with which to answer questions if Google Assistant is addressed. If it's battery powered, there are obvious reasons not to do that unless the phone has a dedicated extremely low powered continuous recording loop circuit.
What people may not understand is that this huge data set is easily consumable.
It's one thing to think "Oh, sure they record me but nobody is going to sit through everything I said".
It's another thing to understand that with machine learning Amazon might be able to parse this data almost instantly and categorize you, rate you and analyse you without breaking a sweat.
I've never met anyone who thought it was always recording. The non-tech folks don't even realize the audio processing is handled on a remote server, and the tech folks understand how the wake word functions together with remote processing.
Yeah pretty much and it's quite stupid for Amazon to keep recordings forever. What use could they possibly be, a year or two max for retention should be the case.
I don't think that people in fact choose convenience over privacy, because the impact of a loss of privacy can be, and usually is, latent indefinitely. There's no way to evaluate what might be given up eventually, and no alarm to tell you when you've passed the point of no return.
For instance, all the US government security clearance investigations were leaked to someone unknown several years ago. The consequence for most people was a short period of credit monitoring and nothing notable otherwise. But if someone now has a database of all possible compromising information, fingerprints, and other information for everyone who's had a position of trust with the government, then some day when a previous "nobody" is suddenly of interest, then that data can be leveraged.
I disagree - Tech folks are aware and people do freak out over it. Then companies come out assuring people and blogs are written showing that nothing bad is happening. So, slowly people forget about it even though they have doubts.
Non-tech folks have no idea what is going on and if they were told this was an always-on recording device, some of them might actually freak out.
In other words, they don't have a way for consumers to set a retention policy, like businesses typically do for email. You have to delete things manually.
It seems pretty clear that there are lots of people who want this and I'd be surprised if more companies don't start doing it. Snapchat made its name with auto-deleting videos.
This is obviously bad, but I think the deeper problem is with Android that it's impossible to turn off the long-home-button press behavior. The wiretapping statutes really need to be updated to cover "negligent wiretapping" and require manufacturers to provide an opt-out toggle that prevents it.
Alexa speakers, of course, are surveillance devices and if you have it on it seems fair to say you have opted in.
What do you mean that it's impossible to turn off the long-home-button press behavior? I've always hated long-pressing home by mistake, so I always prevent it from doing anything at all. Maybe it's because I use a different home app though.
I get that there should be logs.. for debugging and feature tracking etc... I don't think its reasonable for the "voice" to be stored and available to 3rd parties. The system beyond speech to text doesn't respond to voice, it responds to text. So couldn't they just have the logs of the commands put into the system in text? And once the voice is converted + is processed to make the system better, automatically deleted? The voice should only be accessible by the amazon core service responsible for speech to text, the voice shouldn't leave that system, and again, once in text, the text, or commands, can be logged / shared with relevant "skills" providers, since they'd have logs of your interactions with their systems anyways.
They have tons of people working full time to classify the recordings to use as training data. Of course they aren't going to delete them automatically and waste millions of dollars of work.
28 comments
[ 3.0 ms ] story [ 80.2 ms ] threadTheir business model doesn't involve monetizing user's data, and they've been public and explicit about that.
If their promise is broken, their brand will suffer.
Amazon has no such promise or brand. They offer low prices and high convenience.
The other thing is that sometimes a company does something by accident that would damage the brand if anyone finds out, and covering it up is the only way to save it. I don't think there's any way to prevent this from happening somewhere at a huge company from the top, because people are always going to make mistakes and bad decisions and then they don't want anyone to know for fear of consequences.
When has Apple ever done that?
Apple's style would be getting everyone to buy into the feature that has privacy implications with some mitigations.
1. Non-tech people, who don't understand the immediate technical aspects of what's happening, and haven't had much occasion to think about the implications.
2. Tech people, many of us who do understand a lot of what's happening in technical levels, but are conditioned by all the practices around us (and our paychecks/startups depending on it) to not think of this as bad for society.
Wifi/ethernet and AC-connected devices change the risk/benefit calculation significantly.
Alexa or Google Assistant on a phone is therefore not really the same thing as Alexa or Google on a standalone plug-in device.
For example, a google phone (depending on model/setup) and Google mini might both listening all the time for "Hey google". Both respond by listening and recording the next collection of sounds. However, the bar for deciding when to ship that audio off to Google, or when it's confident enough it can handle processing on-device, could easily change depending on whether the device is battery-powered or whether it's on wifi/eth instead of mobile. A wall-powered device might opt to keep a running history in ram of the last N seconds of audio, so it has more context with which to answer questions if Google Assistant is addressed. If it's battery powered, there are obvious reasons not to do that unless the phone has a dedicated extremely low powered continuous recording loop circuit.
It's one thing to think "Oh, sure they record me but nobody is going to sit through everything I said".
It's another thing to understand that with machine learning Amazon might be able to parse this data almost instantly and categorize you, rate you and analyse you without breaking a sweat.
1. good audio compression and selective removal of empty data with time stamps
2. Speech recognition algorithms that convert recorded data to text content
you could hold a non-trivial amount of data in cold storage.
For instance, all the US government security clearance investigations were leaked to someone unknown several years ago. The consequence for most people was a short period of credit monitoring and nothing notable otherwise. But if someone now has a database of all possible compromising information, fingerprints, and other information for everyone who's had a position of trust with the government, then some day when a previous "nobody" is suddenly of interest, then that data can be leveraged.
Non-tech folks have no idea what is going on and if they were told this was an always-on recording device, some of them might actually freak out.
It seems pretty clear that there are lots of people who want this and I'd be surprised if more companies don't start doing it. Snapchat made its name with auto-deleting videos.
It’s honestly ridiculous and only government intervention will prevent this from getting out of hand.
Alexa speakers, of course, are surveillance devices and if you have it on it seems fair to say you have opted in.
I set mine to search in Firefox: https://support.mozilla.org/en-US/kb/how-replace-google-assi...