• aeharding
    link
    fedilink
    English
    arrow-up
    11
    arrow-down
    1
    ·
    13 days ago

    they quite literally are recording constantly. how else can they detect the trigger phrase? they only difference is that they are supposed to delete these recordings after the phrase isnt heard.

    I guess that depends on your definition of recording? An onboard microprocessor waiting for a trigger word is not storing or transmitting anything while waiting and that’s acceptable to me.

    • lseif@sopuli.xyz
      link
      fedilink
      English
      arrow-up
      6
      arrow-down
      9
      ·
      13 days ago

      like i said, are you confident it’s not storing or transmitting anything ?

      • offspec@lemmy.world
        link
        fedilink
        English
        arrow-up
        10
        ·
        13 days ago

        You realize it’s trivial to isolate and monitor traffic for a device on your network, right? Like this isn’t magic, we have the tools to check whether or not it’s physically possible for these devices to be exfiltrating 24h of audio a day based on the bandwidth they consume, and the variability in the transmitted data. There are free, fully sufficient tools to do this at literally every level of your home network, if these devices were actually recording all the time people would be noticing it and reporting on it.

        • lseif@sopuli.xyz
          link
          fedilink
          English
          arrow-up
          2
          arrow-down
          8
          ·
          13 days ago

          they can encrypt the data and bundle it with other requests. regardless, is it really easier to 24/7 monitor your web traffic than to just use a computer/phone instead of a voice assistant ?

          • offspec@lemmy.world
            link
            fedilink
            English
            arrow-up
            12
            ·
            13 days ago

            Encrypted doesn’t mean magically violating the laws of physics, data uses bandwidth. There’s no reason for these devices to be using the amount of bandwidth it would take to make what you’re implying even close to feasible.

            • lseif@sopuli.xyz
              link
              fedilink
              English
              arrow-up
              3
              arrow-down
              7
              ·
              13 days ago

              text transcriptions take up barely any bandwith, i guarantee u wouldnt notice it

              • offspec@lemmy.world
                link
                fedilink
                English
                arrow-up
                2
                ·
                12 days ago

                Yes because security researchers and hobbyists would never compare an isolated device to one plugged in next to a TV or in a crowded room, there just isn’t any data to support that anything like that is happening. As many other commenters have said, these devices are less complex than a cell phone and most people have no issues carrying one of those around. If one of your devices was “tapped” it would be that one.

              • Hexarei@programming.dev
                link
                fedilink
                English
                arrow-up
                1
                ·
                12 days ago

                Most text transcript services are too complex to run on the tiny little processor in these devices. They would be sending audio or nothing at all