Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Patents Privacy Software Technology

Amazon Patent Reveals Plans To Allow Alexa To Listen, Record Users At All Times (sciencealert.com) 171

schwit1 shares a report from ScienceAlert: A newly revealed patent application filed by Amazon is raising privacy concerns over an envisaged upgrade to the company's smart speaker systems. This change would mean that, by default, the devices end up listening to and recording everything you say in their presence. Rather than only record what is said after the wakeword is spoken, the system described in the patent application would effectively continuously record all speech, then look for instances of commands issued by a person.

In the patent application, the authors explain that your Echo device would only ever record between 10-30 seconds of audio at a time, before wiping it from the local memory buffer, and recording a new 10-30 seconds of audio over it (again and again). In each of these 10-30 second recordings, the device would continuously scan looking for commands involving the wakeword, and if it didn't find any, they'd get deleted forever -- in theory, at least. But because of the potential privacy implications of having a device that records you all the time, it's understandable that some people might not be thrilled about what this patent application represents, especially since Amazon has a mixed track record with Alexa recording things it wasn't ever supposed to.

This discussion has been archived. No new comments can be posted.

Amazon Patent Reveals Plans To Allow Alexa To Listen, Record Users At All Times

Comments Filter:
  • by Joe_Dragon ( 2206452 ) on Thursday May 30, 2019 @09:04AM (#58678468)

    The EU may force them to stop this.

    • This is only a patent. It's possible Alexa will not incorporate it. They'll still patent it because they've developed it.

      But I think people are being a little too wary of this. It sounds like it's just recording locally in a small circular buffer. Not sharing the data. There's no data protection concerns to worry about. At least not for now.
    • by gweihir ( 88907 )

      It may actually be a criminal act to have one of these in Germany, unless you put up warning signs.

  • by Anonymous Coward on Thursday May 30, 2019 @09:05AM (#58678480)

    Dont buy one.

    • Computer with microphone can turn microphone on.

      I'm shocked! Shocked.

      Sadly, Your advice about not buying one will never occur to the people who actually are shocked. And it isn't needed for people who already assumed this might happen

    • And don't have friends that have one? I even know of event spaces that have these.

      Your voice will be fingerprinted even if you don't buy one of these.

  • Listening=Recording (Score:5, Informative)

    by Immerman ( 2627577 ) on Thursday May 30, 2019 @09:07AM (#58678502)

    How exactly does anyone imagine Alexa can hear the wakeword if she's not recording? To a computer, listening is a matter of recording sound and then analyzing it, I've never heard of any other technique suggested.

    • by forkfail ( 228161 ) on Thursday May 30, 2019 @09:13AM (#58678552)

      My understanding of the current design claimed by Amazon is that the wakeword is detected from an audio stream that is not written to any storage at the local device.

      Once that is spoken, however, the stream is sent to AWS for processing of whatever commands follow.

      • by 110010001000 ( 697113 ) on Thursday May 30, 2019 @09:19AM (#58678594) Homepage Journal

        Data has to be stored to be analyzed. It might be stored in RAM or whatever, but it is stored.

        • Yes, but that may even by stored on separate RAM module that isn't accessible by the main processor, and processed by a separate controller that isn't accessible from the main controller. These sorts of systems exist in cell phones to cut down on power usage. You have a dedicated chip who's only job it is to listen for the key word so that the main processor can run in low power mode and go to sleep. I then gets woken up once the sound processing chip sends the wake signal. The sound processor only needs

        • You are silly, you can analyze it on the fly. After all the "wake words" are super short. Unless you want to call two bytes with a 8kHz 16 bit sample "storage".

      • 30 seconds of audio easily fits in memory. The real question is why would anyone think this is something that is eligible for a patent? Even if granted it would be revoked as trivially obvious the first time it is challenged in court.
        • The article doesn't seem to give much detail on the patent, but my takeaway is that this is an implementation detail "revealed" in the patent, rather than the actual subject of the patent. At least I certainly hope so.

        • by kenh ( 9056 )

          As I read it, the "30 second recording" technology is not the focus of the patent, it is a part of a greater function, but it has caught the eye of "privacy experts".

          The 30 second window is likely to accommodate a long "action word" (like, say the German word "Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz" [thoughtco.com]) AND figure out what you want - for example:

          "Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz, order another 12 roll pack of Charmin Ultra Soft Tissue

        • 30 seconds of audio easily fits in memory. The real question is why would anyone think this is something that is eligible for a patent?

          What we need is a patential version that compares the original spoken waveform against an AutoTune version of it. If they mostly match (ie, you're in harmony) THEN it activates. If not, it doesn't

          Just THINK of all of the accidental activations this could prevent. "Open the pod bay doors, Alexia." "I'm sorry Dave, you're not in tune and so obviously stressed. Why don't you order some Stress Pills instead? There's a sale on right now. Or I could play some nice background music while I wait for your ai

      • I don't imagine anything has changed in that respect. There's almost certainly currently a circular buffer long enough to detect the wakeword and at least enough of what follows to still have it available to send after the wakeword has been recognized, so that everything from that moment on can be sent to Amazon.

        Sounds like the change is just that the circular buffer has been made large enough to record the sentence said before the wakeword, and that gets sent too.

        I mean sure, it slightly increases how muc

        • by kenh ( 9056 )

          But if you've gone out and bought one or more purpose-built corporate surveillance devices to install in your home, it's a little late to get creeped out by a few more seconds of surveillance.

          Next up, the "revelation" that Ring doorbells are continuously recording everything that occurs within their view, even people walking in front of your door that never stop and press your doorbell!. Sure, Ring engineers "claim" they only store the last 30 seconds of video" - in theory at least, but privacy advocates are concerned!

      • by nagora ( 177841 )

        Exactly. And that can be changed by a remote command at any time. The patent is just trying to patent buffering which is an obvious and well-established method of holding on to things for a while - in this case so that you can say things like "order me a pizza, Alexa" instead of "Alexa: order me a pizza".

        Given how simple and obvious this is, I'm sure the US patent office will 100% approve the granting of a government-back monopoly on the idea.

        Because the US patent office is shit and anyone that cared about

      • And that understanding is completely incorrect.

        My parents have one. One day I got alerts that they were trying to reset their password. (My email account is setup as a recovery address.)
        I found out they were really trying to reset their Amazon password, but didn't (and still don't) know the difference between Gmail and Amazon.
        I ended up resetting their password for them, and in the process of verifying the Alexa device was connected and registered to their account, I found that everything you fucking say

        • We should thank the anonymous Amazon engineer who decided to make those recordings visible. No doubt management would be livid if they knew about it. A small, quiet act of resistance from inside the belly of Big Brother.

    • also Echos do have a hardware mute switch, so just leave it muted, then when you want to use it, unmute and ask your question. This is quite easy, and I don't see anything in this patent which, while moronic, would mess with that.
  • by Anonymous Coward

    It's a waiting game. Wait for the tech literate users to age out of the system, erode the privacy of ever younger users. Rinse and repeat until people have been subjugated beyond your wildest dreams.

    People will be domesticated enough before this patent expires in 20 years, but the march of surveillance-capitalism will have advanced so far beyond this will look quaint in comparison.

  • Recording (Score:5, Insightful)

    by 110010001000 ( 697113 ) on Thursday May 30, 2019 @09:12AM (#58678536) Homepage Journal

    Hmmmm, wouldn't the current devices need to "record all speech" in order to analyze it to see if it contains the "Alexa" keyword? I mean it might be recording it to RAM and then throw it away, or it might be transmitting to to Alexa, or the Chinese, or Russia, or the White House. And what it does can change on any software update pushed to the device. The point is, with closed source software you have zero idea what it is doing, and what it does can change instantly if the software is updated. This might be obvious, but the closed nature of the device is the problem.

    • by Anonymous Coward

      Two questions.

      1. Are these things capable of recording everything you say, private or not, and sending it back to the mothership to build a permanent profile on you? The answer is obviously yes.

      2. Do they have a financial incentive to do it? The answer is obviously yes.

      Conculsion: Yes, these things are recording everything you say and sending it back to the mothership in order to build a permanent profile on you.

      Short version: If it's technically possible, and financially profitable, they're doing it.

      • Two questions:

        1. Are you capable of buying a gun and shooting your neighbor in order to take all the money and valuables from his home? The answer is obviously yes.

        2. Do you have a financial incentive to do it? The answer is obviously yes.

        Conclusion: you are guilty of murder and theft. Proceed to the nearest prison to be executed.

        Short version: just because something is possible doesn't mean it is being done.

        • Question 3: Are there any penalties against doing so? Yes, in the case of murder and theft. No, in the case of Amazon data harvesting and profiling.

          Conclusion: Likely no murder. Likely data harvesting and profiling.
          • Yes, in the case of murder and theft. No, in the case of Amazon data harvesting and profiling.

            You seem very sure about that "no". I don't think you are right. But you bring up the point I was getting at: there are other considerations besides "can it be done" and "can we make money doing it" that are in play. Amazon is not exempt from such concerns. Jumping to conclusions based on an incomplete consideration of the concerns is inappropriate.

    • by kenh ( 9056 )

      The point is, with closed source software you have zero idea what it is doing, and what it does can change instantly if the software is updated.

      If only there was some way to monitor the traffic flowing in and out of the device over your own, private, internet connection...

      This might be obvious, but the closed nature of the device is the problem.

      No, it's "closed nature" isn't the problem, it's that anyone is so stupid to be surprised/upset that a device that says in all it's marketing materials that it monitors conversations to detect an action word is actually monitoring your conversation to detect an action word.

      Do you really think "The Russians" or "The Chinese" care about your inane conversations? Do you imagine that

      • If only there was some way to monitor the traffic flowing in and out of the device over your own, private, internet connection...

        This is why I'm generally comfortable with a smart device in most rooms. I know it's not transmitting everything to the cloud because that would be obvious in the data.

        What is concerning though is that as the processing power increases, soon these devices will comfortably be able to translate everything to text. At that point all conversations could easily be compressed and upload

  • I bet they are already doing that (like Android phones, iPhones etc. etc.).
    • They are, which is why I'm surprised this is somehow news. When you replay the recordings on Amazon's website, they start .5 seconds before the uttering of the "wake word". Either they have some advanced precognition built into that thing (in which case "Alexa, what are tonight's lottery numbers" might be a worthwhile question) or it just clips the recording before uploading (or more likely - after).

  • by kenh ( 9056 )

    How exactly can Alexa continuously monitor for the "action word" without, you know, continuously monitoring what is said in it's presence?

    The thirty second window is likely nothing more than a big enough window for Alexa to handle "excessively long" action words, like, for example "Hello Alexa please wake up" or other extreme "action words".

    Lets say your worst fears are true, and despite the promises of the engineers, instead of a revolving 30 second recording, the device actually, secretly, records every s

    • You're overthinking. Recording everything you say gives them data. Data that can be sold. For ads, for government, for whatever reason.
      It's valuable. There will be pressure to capture and use it.
      I guess if you don't care that everything you say in your home is recorded, it's not an issue.

  • }}} In the patent application, the authors explain that your Echo device would only ever record between 10-30 seconds of audio {{{ --- Then the next patent comes along that patents continuous recording.
    .

    Then yet another patent comes along that describes a digital audio stream being sent back to amazon servers continuously.

    Then another patent comes along that describes a continuous video stream being sent back to amazon servers.

    The phrase slippery slope comes to mind.

    • The phrase slippery slope comes to mind.

      Speaking of slippery slopes, have you gotten one for your bedroom yet?

    • by kenh ( 9056 )

      Your "slippery slope" is well documented in a series of patents, for all the world to see - that's not how most nefarious plots are carried out...

  • by Anonymous Coward

    Who is the bigger Bozo? I think it's those who chose to have one of these 'assistants' in the first place.

    As schmuckerberg said:

    schmuck: People just submitted it.
    schmuck: I don't know why.
    schmuck: They "trust me"
    schmuck: Dumb fucks

  • by kenh ( 9056 ) on Thursday May 30, 2019 @09:32AM (#58678662) Homepage Journal

    Were I to design an Alexa-like device, I'd record sounds nearby, and between breaks in conversation, digest the last chunk of the conversation and determine if there was anything for the device to respond to.

    For example, using AI, Alexa could monitor my discussion of the song "The Downeaster 'Alexa'" and not falsely-activate once, but if I utter the word Alexa in a different context, it would properly activate. I suppose the hardware could be designed to be hardwired to respond to exactly one "action word", and mechanically respond every time that one unchangeable "action word" is uttered in it's presence, but that seems incredibly inflexible and less than optimal.

    This is a technology you have to actively purchase and activate, if you are concerned, don't buy/activate it. You could literally rewrite this story to say "A device you buy that continuously monitors your conversations looking for it's 'action word' has been found to be continuously monitoring your conversations - privacy advocates express concern."

  • It's impossible to not record continuously. The act of listening for the so-called "wake word", which is for triggering transmission to servers, means the devices are never asleep and never stop recording. Just like motion detection in CCTV, the cameras record everything and only afterwards does the motion detection throw away the unneeded frames.

    The patent isn't about allowing anything. If anything at all, it will be about formally ensuring deletion, after processing for possible commands to be sent to

  • Comment removed (Score:4, Insightful)

    by account_deleted ( 4530225 ) on Thursday May 30, 2019 @09:51AM (#58678802)
    Comment removed based on user account deletion
  • Microphones are all around us. Not just in smart speakers. They are in our phones, laptops, tablets, elevators, and cars etc... What would Amazon do with 30 seconds of recordings, get us to buy more crap? What about the other microphones? Especially the ones we don't own. Privacy is just plain inconvenient.
  • People that own these spy devices don't mind that it's listening, recording or whatever them. They expect it.

    Most of the same people are just drones anyway. 95 % of them probably do and say same shit every day, who cares about the content of their conversation, other than for purposes of exploitation (commercial purposes) of course. You know...keep the consumerism alive and well.

    Same reason website trackign etc.. nobody cares about individuals in these cases, it's just ends up as aggregate data, usual
  • ... and yet some people wonder why I would never have an Alexa device in my house....

    • No. Most people don't care whether you would have one or not, and the rest don't care why you wouldn't. The very tiny number of people who might care in any way, all employed by Amazon in marketing, fully understand paranoia and fear.

      Can we all just assume that every time there is a story about Alexa and Amazon that all the people who don't want to have one don't want to have one, and avoid all the noise of them telling everyone how smart they are because they don't want one?

  • by doubledown00 ( 2767069 ) on Thursday May 30, 2019 @11:16AM (#58679516)
    Amazon wants to collect and analyze more data. This shouldn't be shock to anyone.

    Endgame: Previously they (supposedly) looked at only the wake word and what came after. With the "ability" more fully analyze what was said and for a longer period, they then have the capability to incorporate other "wake words". For instance Sony could pay them to find out how many times their brand is mentioned in a given household. Marketing campaigns can pay to see if their ads garner any discussion just after they are run. There are a million different revenue generating applications for this.

    And lest one say the above is all paranoid talk, Amazon has not given us reason to believe they are honest. Several times now they have been caught doing things with the assistant that they previously denied doing. They have not earned the benefit of the doubt.
    • Amazon wants to collect and analyze more data. This shouldn't be shock to anyone.

      And yet, this patent doesn't do that. Strange way to accomplish the goal you think they have.

      With the "ability" more fully analyze what was said and for a longer period, they then have the capability to incorporate other "wake words".

      They don't need this patent to use other wake words.

      For instance Sony could pay them to find out how many times their brand is mentioned in a given household.

      They don't need this patent to do that.

      And lest one say the above is all paranoid talk,

      It is all paranoid talk. It is a PATENT, not a production device. It is one small part of a patent, to boot.

      • Amazon wants to collect and analyze more data. This shouldn't be shock to anyone.

        And yet, this patent doesn't do that. Strange way to accomplish the goal you think they have.

        With the "ability" more fully analyze what was said and for a longer period, they then have the capability to incorporate other "wake words".

        They don't need this patent to use other wake words.

        For instance Sony could pay them to find out how many times their brand is mentioned in a given household.

        They don't need this patent to do that.

        And lest one say the above is all paranoid talk,

        It is all paranoid talk. It is a PATENT, not a production device. It is one small part of a patent, to boot.

        I'm well aware this isn't *yet* a production device. You blithely waive all this off without understanding that patents are as much product roadmaps to future implementations as they are bludgeons against competitors.

        • I'm well aware this isn't *yet* a production device.

          Then don't complain about it like it is. This is a part of a patent. It's an idea. It's such an obvious idea that it isn't even patentable by itself. It's something that any competent software engineer would do under similar situations.

          You blithely waive all this off without understanding that patents are as much product roadmaps to future implementations as they are bludgeons against competitors.

          Sorry, but you are wrong. Patents are patents. They are not product roadmaps. In this case, it's not even a "product roadmap" that is required to accomplish TODAY everything you assume they are going to do tomorrow based on one small part of one patent filing. You are comple

  • Is anyone actually surprised by this?

  • I'm afraid I don't have time to go looking at the actual patent application right now... but if this application isn't written just right, (and dated pretty far into the past) it's probably going to fall very easily to a challenge. As described in this /. post, it sounds like "record" is only loosely defined as "written to memory", regardless of whether or not it's persistent memory. I would imagine that it would be very easy to argue that Apple and Google (as well as Alexa) are already doing basically the

  • The need for people to willingly pay to give up their privacy.
    "Oh, you can tell me if I'm related to the Masai people in my ancestry? Here's #100 and my DNA for you to test and record, and you can do what you want with the result!"
    "Oh, i can tell this machine to turn on my light instead of doing it with my finger? Here's some money, hook up that recording device that constantly monitors what I say".
    "Oh, I feel unsafe in this unsafe world? Put in cameras everywhere on the off chance I get mugged, and let the

    • by ebvwfbw ( 864834 )

      I know a security guy with very high levels of clearance. He thinks it's wonderful. Sets alarms, wake up, lighting, heating/cooling, etc. He doesn't think it's a big deal to have a great big microphone in his house recording his every word. I told him I thought he was nuts.

      Of course the ironic thing is that it seems as if some houses are recorded more than the White House.

  • by WillAffleckUW ( 858324 ) on Thursday May 30, 2019 @01:21PM (#58680598) Homepage Journal

    The Computer is your Friend!

    Have you had your soma today?

  • This is actually a less technically involved (or developed) way of doing things, so it's more likely that such a patent is designed to enable Amazon to sue inferior competitors out of existence for patent infringement.

And it should be the law: If you use the word `paradigm' without knowing what the dictionary says it means, you go to jail. No exceptions. -- David Jones

Working...