Amazon Staff Listening to Alexa Recordings

Ron Burgundy II

https://www.rte.ie/news/business/2019/0411/1042068-amazon-staff-listen-to-alexa-recordings/

Amazon has confirmed that staff are listening to Alexa recordings in a bid to improve the AI ability to understand human voices.

If Amazon is doing this then Google has to be doing something similar?

Buttros

Sounds like an awful job. Listening to me asking Alexa to do x y and z and saying not nice things when it mis hears me a few times. 😂😂😂

Creamy Goodness

of course they are, no one is surprised by this.

facebook says it's not doing something, then 6 months later, yeah actually, we are doing that thing.
google says it's not doing something, then 6 months later, yeah actually, we are doing that thing.
amazon says it's not doing something, then 6 months later, yeah actually, we are doing that thing.

Ron Burgundy II

Is this Amazon's first announcement?

TCP/IP

Who cares if it makes the AI and NLP better.

Uriel.

I can't believe that this is somehow shocking news for people. What is important is that their anonymisation and data security processes are as sufficiently robust as they claim.

Despite having a spot I do remain concerned about the possibility of video "spying"

OldMrBrennan83

TCP/IP wrote: »

Who cares if it makes the AI and NLP better.

Exactly. All these things are middling at best now so hopefully it improves them all.

matrim

What likely happens is that a recording is flagged as problematic and is sent for human review to transcribe it and then process what went wrong. Recordings should be anonymous.

TommyKnocker

Ron Burgundy II wrote: »

https://www.rte.ie/news/business/2019/0411/1042068-amazon-staff-listen-to-alexa-recordings/

Amazon has confirmed that staff are listening to Alexa recordings in a bid to improve the AI ability to understand human voices.

If Amazon is doing this then Google has to be doing something similar?

Mobile phones are doing the same thing. Why do you think so many of the popular apps people install require access to the devices microphone? With the advancement of IoT it will only get more widespread.

wexfordman2

TommyKnocker wrote: »

Mobile phones are doing the same thing. Why do you think so many of the popular apps people install require access to the devices microphone? With the advancement of IoT it will only get more widespread.

That's not what you said as you typed .

Effects

Most people don't care that they are being listened to or spied on, that's why the companies keep doing doing it.

theteal

Not getting much from our house other than arguments about how to stack the dishwasher!

punisher5112

Jaysus I would hate to be the person looking at me phat a#s through the Alexa show.....

Alexa " put that away"

Clareman

How do they expect voice recgonition to get better if they don't review what the product is being used for nevermind the fact that normally there's an option of opting out of "Making the service better", this is a complete non-story in my opinion.

Atomic Pineapple

The naivety in this thread is testament to how tech companies have managed to institutionalize spying (and subsequently stealing). Most people don't know what the information they give up is being used for, if they did, they would care. The data is only barely anonymous, there are academic studies showing that whether it be voice or text, it only takes minimal effort and less then 30 minutes to figure out who any given person is in the data. The opt out argument is a joke, any ethical company would make it opt in. No one reads the terms and conditions because they have been purposely designed to be burdensome. Other companies have paid people for their voice data to help train their natural language processing. Amazon claim to only use a very small amount of samples, if so, they have no excuse for not doing the same. It is a very important story to be reported on because it will get some people questioning the practices of organisations like Amazon. That is a very good thing.

OldMrBrennan83

I know, and I still don't care. Whatever improves my services.

Convenience > Privacy for me. For anyone who doesn't feel the same then it's easy to log off.

Atomic Pineapple

OldMrBrennan83 wrote: »

I know, and I still don't care. Whatever improves my services.

Convenience > Privacy for me. For anyone who doesn't feel the same then it's easy to log off.

Ah your so innocent, thinking it's being used to improve your services .

It's not easy to "log off". There are many blogs where people have done so only to find it's almost impossible to get away from the data collection techniques of these companies. One woman was even accused of being a criminal when doing so. So, there goes the claim that someone just needs to go "off the grid", the tech companies have ingrained themselves so much in our lives that we can't.

OldMrBrennan83

Atomic Pineapple wrote: »

Ah your so innocent, thinking it's being used to improve your services .

It's not easy to "log off". There are many blogs where people have done so only to find it's almost impossible to get away from the data collection techniques of these companies. One woman was even accused of being a criminal when doing so. So, there goes the claim that someone just needs to go "off the grid", the tech companies have ingrained themselves so much in our lives that we can't.

Nope, improving my services is a by-product of what they're using it for. That's ok with me.

Atomic Pineapple

OldMrBrennan83 wrote: »

Nope, improving my services is a by-product of what they're using it for. That's ok with me.

Perfectly proving my earlier point that tech companies have managed to convince the general public that it's OK to spy on them, steal their data, and sell it to the highest bidder.

OldMrBrennan83

Atomic Pineapple wrote: »

Perfectly proving my earlier point that tech companies have managed to convince the general public that it's OK to spy on them, steal their data, and sell it to the highest bidder.

They haven't. Some people just don't care and others think their personal data is massively more important than it is.

Atomic Pineapple

OldMrBrennan83 wrote: »

They haven't. Some people just don't care and others think their personal data is massively more important than it is.

They clearly have, you just said you're OK with it as long as it results in some arbitrary improvement in some service you use. You claim you're aware of their practices. Thus, proving my point.

You're entitled to your opinion on your own personal data but your claim to be able to put a value on the personal data of others just offers more evidence of the institutionalization of the tech company's practices (and language). Why do you think Silicon Valley execs send their children to schools that ban the use of electronic devices entirely?

punisher5112

This business of selling it off though is mad....

I know what I want and don't bother with ads so can't really see what they could do with the data otherwise.

Obviously I don't want very personal information or important stuff been taken but talking to Alexa and meself I should be ok...

ashes2014

OldMrBrennan83 wrote: »

They haven't. Some people just don't care and others think their personal data is massively more important than it is.

I know my personal data is very important to me. Apparently its very important to private companies too-if it wasn't then they wouldn't be listening.

If a private company wants to listen to my conversations they should be open and transparent about it. I would tell them where to go, as would the majority of people, which is why they do it sneakily.

Atomic Pineapple

punisher5112 wrote: »

This business of selling it off though is mad....

I know what I want and don't bother with ads so can't really see what they could do with the data otherwise.

Obviously I don't want very personal information or important stuff been taken but talking to Alexa and meself I should be ok...

Alexa records everything, at all times, not just when you request something from it, are you aware of this? The idea isn't to grab all of what you say, it's to understand things like when you leave for work, when you're home, how long you spend in a certain room, when you eat, what you watch, etc.... The more you can know about someone the less you leave to chance when trying to manipulate them.

km991148

The fact this is headline news shows the disconnect between IT industry (how quickly its evolved) and the general public.

This is not some conspiracy by tech companies, just the way this sort of technology works (how else do you think Alexa etc is able train to understand deepest Kerry accents or heavily French accented English etc).

The issue arises when the technologists design this sorta stuff meet with the realities of a call centre full of predominately young uns doing the system training and whenever they hear funny, drunken / whatever clips they play them back for fun. But even then, they probably have no real way (or care to) link them back to the end user.

This isnt the worst kinda thing big tech do - I would be more concerned about selling of data, manipulation techniques to sell more stuff or the shutdown of competition that Amazon etc causes. Combined with aggressive techniques to reduce tax liabilities its a general downward spiral. Having some kid laugh at my drunken demands to play karaoke classics is well down the list of concerns. More than likely its probably not even Amazon/Apple/MS staff anyway and instead a few private companies that provide this service to Apple etc for use in Siri Cortana or whatever.

punisher5112

Atomic Pineapple wrote: »

Alexa records everything, at all times, not just when you request something from it, are you aware of this? The idea isn't to grab all of what you say, it's to understand things like when you leave for work, when you're home, how long you spend in a certain room, when you eat, what you watch, etc.... The more you can know about someone the less you leave to chance when trying to manipulate them.

Yes aware of it. Most tend to be unplugged and the little one is going through the faze of everything has to be queen due the the film.

She has it playing 24/7.

I have a bit of tape over the Alexa show too.

c.p.w.g.w

theteal wrote: »

Not getting much from our house other than arguments about how to stack the dishwasher!

Let me guess your wife is stacking it wrong...

My misses, SIL, Sister, mother and dad's partner all can't stack a dishwasher for the life of them...greasy frying laying flat on the top shelf

km991148

Atomic Pineapple wrote: »

Alexa records everything, at all times, not just when you request something from it, are you aware of this? The idea isn't to grab all of what you say, it's to understand things like when you leave for work, when you're home, how long you spend in a certain room, when you eat, what you watch, etc.... The more you can know about someone the less you leave to chance when trying to manipulate them.

It may record everything at all times, but more than likely only stores it for a few seconds in order to check for the trigger word (i.e. it listens to everything, but doesn't keep everything).
What it does with the successful triggers tho.. it probably keeps that and all sorts of other metadata, but Amazon keep sh!tloads on you already - what you have looked at, what searches you have done etc etc and probably uses it (along with Alexa metadata) to answer questions like - when you are most likely to purchase something, what will make you pay for prime, why do you always buy a giftcard or whatever around the time of you bros bday. How much you are likely to spend on said gift card when presented with default options. What product from the store meets your search with the highest margin that meets your specific search terms... etc etc etc..

punisher5112

It is acknowledged that conversation is kept up to a minute and then wiped but then if people are listening in then is this live or are they now keeping all conversation....

km991148

punisher5112 wrote: »

I know what I want and don't bother with ads so can't really see what they could do with the data otherwise.

You probably think you do know what you want. But Amazon probably often think they have a better idea.

Just because you can't see what they could do with the data doesnt meant o say there isn't all sorts of stuff they can do that you dont know about. Not to mention all the stuff they will be able to do in the near future with the data you have already given..

An extreme example would be - Fast forward 10 years. Amazon knows (because everyone has their app on their phones) that you frequent with people who buy more than average amounts of alcohol (bought via amazon groceries). They therefore conclude that you are more than likely a drinker and therefore sell that to the health insurance industry and affect your premium. Why would you give them consent to do that.. because in 2025 you bought the Amazon Health Watch that has that sorta stuff buried in its Ts an Cs. (This is a very contrived example, but not beyond the realms of reality..)

km991148

punisher5112 wrote: »

It is acknowledged that conversation is kept up to a minute and then wiped but then if people are listening in then is this live or are they now keeping all conversation....

If Alexa densest understand something (some accent etc), its sent to someone to transcribe so that over time it gets easier to understand that accent.

Other cases may be that when a query doesn't make sense its sent to someone who may categorise it "Sales attempt, Piss taking, Customer Support Query" etc.

People probably dont have the ability ti listen live, nor would they want to (but they would probably have no issue making a recording of the eejit trying to buy something stupid when drunk jsut for a laugh with their mates)

punisher5112

km991148 wrote: »

You probably think you do know what you want. But Amazon probably often think they have a better idea.

Just because you can't see what they could do with the data doesnt meant o say there isn't all sorts of stuff they can do that you dont know about. Not to mention all the stuff they will be able to do in the near future with the data you have already given..

An extreme example would be - Fast forward 10 years. Amazon knows (because everyone has their app on their phones) that you frequent with people who buy more than average amounts of alcohol (bought via amazon groceries). They therefore conclude that you are more than likely a drinker and therefore sell that to the health insurance industry and affect your premium. Why would you give them consent to do that.. because in 2025 you bought the Amazon Health Watch that has that sorta stuff buried in its Ts an Cs. (This is a very contrived example, but not beyond the realms of reality..)

I get what your saying alright.

Hopefully this won't be the case.

o1s1n

punisher5112 wrote: »

I get what your saying alright.

Hopefully this won't be the case.

My partner works in GDPR and as much as it does everyone's head in at the moment, it's scenarios like that which people in the field are now trying to prevent. She spends a lot of time at data privacy conferences and some of the projected scenarios they come up with like the above would actually scare you.

I have been a firm backer of Alexa since Echo was released but have now decided to get rid of it. Found that I was mainly using it as a bluetooth hub for my stereo and wemo switches - so going to try and see if I can find an alternative solution that doesn't involve an external connection.

Sure I'll be losing the voice commands, but after using it now for a while I don't think the tech is worth the level of privacy you need to give up in exchange.

Atomic Pineapple

km991148 wrote: »

The issue arises when the technologists design this sorta stuff meet with the realities of a call centre full of predominately young uns doing the system training and whenever they hear funny, drunken / whatever clips they play them back for fun. But even then, they probably have no real way (or care to) link them back to the end user.

It's incredibly easy to link it back to the end user (granted you're right, the kids listening are not going to do it). Even DNA that was sent into those ancestry type services was bought and traced back to the end user without any medical expertise required. You can buy vast amounts of browser history from a data broker and it is easy to identify who it belongs to from looking at things like when they look at their own profiles.

km991148 wrote: »

It may record everything at all times, but more than likely only stores it for a few seconds in order to check for the trigger word

It doesn't need to store all the audio, just the behavior it infers from it. This type of data about you is stored forever (Unless you do a GDPR request to delete it).

The Oort Cloud

I can't believe that people actually install spyware in their own home and bedrooms, it's beyond idiotic

The Oort Cloud

In the 90's GSM users were outraged to hear that there were many people listening into their private phone conversations using radio scanners, then 2 or so years later The GSM networks started to implement encryption and that put a complete end to the security breach. But look at how things have changed today, people basically want to give their privacy away to any company or person, while installing spyware boxes themselves in their own homes.

Technology really has imo made a lot of people stupid

km991148

Atomic Pineapple wrote: »

It doesn't need to store all the audio, just the behavior it infers from it. This type of data about you is stored forever (Unless you do a GDPR request to delete it).

Thats the metadata I discuss in the rest of the post.

wexfordman2

Atomic Pineapple wrote: »

Alexa records everything, at all times, not just when you request something from it, are you aware of this? The idea isn't to grab all of what you say, it's to understand things like when you leave for work, when you're home, how long you spend in a certain room, when you eat, what you watch, etc.... The more you can know about someone the less you leave to chance when trying to manipulate them.

1) smart speakers do not "record everything at all times". The speaker listens for a wake word, and only once it hears that wake word, does it forward the audio for analysis to interpret what you said.

That is the fundamental principle of how they work.

Anyone who thinks you are getting something for free with a smart speaker is a fool, and people in this day and age should understand that primarily, the user is the product, not the device, or the provider.

As far as privacy is concerned, on the proviso that the system is designed to only send recorded data to the cloud after the activation word, the I have absolutely no concerns with privacy. I don't care that Google knows I want to turn on my lights, turn up the heating, set an alarm, or ask it how many lbs in a kg. I am fully aware that anything I ask it, is not private, that it is listening and responding to me and my questions.

When I type something into a web browsers, I do so k owing that it is not private, the same when i ask my smart speaker a question, the only difference, is one method is done using my fingers, the other using my voice.

km991148

wexfordman2 wrote: »

As far as privacy is concerned, on the proviso that the system is designed to only send recorded data to the cloud after the activation word, the I have absolutely no concerns with privacy. I don't care that Google knows I want to turn on my lights, turn up the heating, set an alarm, or ask it how many lbs in a kg. I am fully aware that anything I ask it, is not private, that it is listening and responding to me and my questions.

Thats not the real concern when discussing privacy. What happens when they use the data (and specifically the metadata around your queries) to influence your future bahaviour. Or the behaviour of other services you buy. Currently (especially for google) thats all advertising sales. But its not the only thing.

o1s1n

wexfordman2 wrote: »

Anyone who thinks you are getting something for free with a smart speaker is a fool, and people in this day and age should understand that primarily, the user is the product, not the device, or the provider.

Likewise, it's also foolish to think this is about Google 'knowing you've turned on your lights, turn up your heating, set an alarm, or ask it how many lbs in a kg.'

i would imagine most people couldn't give a fiddlers if Google know they've just switched on their lights.

It's what Google do with that data. Are you okay with them building up a profile on you with this data and selling it to third parties?

Are you okay with one of those third parties perhaps being Facebook, who may then adjust what you see on your daily feed in an attempt to see if they can alter your mood? (They have admitted doing such) - the length at which this kind of thing can reach is massive and it's only in it's infancy.

Secondly, there's also the data it's sending that falls outside of the set parameters.

One thing that really freaked me out to be honest was, while looking though my Alexa saved recordings, how many times there was an accidental activation and how much conversational data has been transmitted that shouldn't have been.

That, coupled with the idea of staff listening to such recordings is enough to completely turn me off the idea - (whether the intent is malicious or not)

wexfordman2

km991148 wrote: »

Thats not the real concern when discussing privacy. What happens when they use the data (and specifically the metadata around your queries) to influence your future bahaviour. Or the behaviour of other services you buy. Currently (especially for google) thats all advertising sales. But its not the only thing.

On, expand on that for me, give me an example of what I should be worried about, give me an example also then of how it is different based in me typing something as opposed to using the spoken word.

o1s1n

wexfordman2 wrote: »

On, expand on that for me, give me an example of what I should be worried about, give me an example also then of how it is different based in me typing something as opposed to using the spoken word.

Big difference for me anyway would be a lack of control.

You type something into Google, you have complete control over what you are sending.

A listening device can mishear things and transmit incorrect data, accidentally activate as I mentioned above and send personal conversations, can pick up content in the background which is not to be transmitted - there are far more failure points in the process.

km991148

wexfordman2 wrote: »

On, expand on that for me, give me an example of what I should be worried about, give me an example also then of how it is different based in me typing something as opposed to using the spoken word.

I gave an example above and I am not saying its any different to typing or using any other device above (I also gave examples involving apps) and played down the whole "Amazon staff listening" panic.

I am not trying to be antagonistic. Nor am I being over dramatic about dystopian futures or whatever. Just trying to highlight that there are use cases that people in general have not thought about (nor have tech companies, yet) BUT you have already given data to to enable.

wexfordman2

o1s1n wrote: »

Likewise, it's also foolish to think this is about Google 'knowing you've turned on your lights, turn up your heating, set an alarm, or ask it how many lbs in a kg.'

Which is what I said also.

o1s1n wrote: »

would imagine most people couldn't give a fiddlers if Google know they've just switched on their lights.

It's what Google do with that data. Are you okay with them building up a profile on you with this data and selling it to third parties?

In so far as they do so already with .most of my internet use via a web browser, yes i am. I am fully aware that they use this data to try and sell me other products via third parties. O see no difference between the level of privacy i am afforded when i type something in to a web browser, or request it vocally.

o1s1n wrote: »

re you okay with one of those third parties perhaps being Facebook, who may then adjust what you see on your daily feed in an attempt to see if they can alter your mood? (They have admitted doing such) - the length at which this kind of thing can reach is massive and it's only in it's infancy.

Again, in the same way as it occurs when I type something, I see no difference

o1s1n wrote: »

econdly, there's also the data it's sending that falls outside of the set parameters.

What data exactly ?

o1s1n wrote: »

ne thing that really freaked me out to be honest was, while looking though my Alexa saved recordings, how many times there was an accidental activation and how much conversational data has been transmitted that shouldn't have been.

That, coupled with the idea of staff listening to such recordings is enough to completely turn me off the idea - (whether the intent is malicious or not)

Mm, I am aware of accidental triggers of the awake word, I get an audible.indication, I know exactly.when it happens, as it happens.

NIMAN

The people who will be outraged by this will be the same numpties who have been on Facebook for 10 years and telling them and the world their every desire, thought, preference etc.

km991148

NIMAN wrote: »

The people who will be outraged by this will be the same numpties who have been on Facebook for 10 years and telling them and the world their every desire, thought, preference etc.

yep - but generally people are not Numpties because they dont know this. Technology has moved much faster than people can even imagine. I don't just mean the technology and the gadgetry, I mean the business models, the ethics, the privacy concerns.

Even to the point that a lot of people may even think they get it and are comfortable with it, but not fully realising all that can happen in the background.

EDIT - The worst ones are the "journalists" who dine out on this sort of sensationalism and mock outrage. A whole industry that didnt move along with the modern world is now conspiring against us because people are not driven enough to do basic fact checking. Outrage sells headlines, semi sensible statements about Amazon not giving a sh!t about your private conversations doesn't.