The three biggest players in voice assistants –– Google, Apple and Amazon –– have radically different approaches to profiling users, Northeastern University researchers say.
My Google speaker only hears me say “hey Google, set an alarm in x hours” and “stop”. Good luck profiling me.
And I turned off the Google assistant in my car. It was more a nuisance than a blessing. It would trigger if you said “eierkoeken” which was hilarious when we were talking about those things during a road trip.
To hear you say “Hey Google” it has to listen to everything you say, all the time. While they pinky-promise they aren’t doing anything with all the voice data they’re getting while listening, do you trust them?
While this is true, it’s not difficult to verify. You can inspect the network traffic through your router from that device and see whether it’s communicating more often or with larger data packets while you’re talking near but not to it.
That could be obfuscated with a powerful enough device that’s able to maintain the user profile locally and by sending the full profile each time, or by trickling the information out bit by bit as part of a regular heartbeat traffic, but that would require more expensive hardware and would eat up a lot more bandwidth.
Not impossible, but not undetectable either. I’m willing to bet that studies have been done.
Skepticism is good but without thinking through it, experimenting, or doing research to back it up it’s just paranoia. Conversely you could say my trust that others would have looked into it is naive.
studies were done and found similar to what you’re saying.
also the secret listening does not comport with any of the business side of profile marketing, either. So it would have to be an incredibly well kept secret on top of all of that.
Good point, but given that I don’t say anything in my bed room (I live alone, and I don’t date), I wish it good luck hearing anything.
eierkoeken
But why tho? That’s such a random thing to trigger on. Is it anywhere near “hey google” in Dutch or something?
Yeah it kinda sounds like it. The eier kind of sounds like hey, and koeken sounds a bit like Google.
I’ve had a few other accidental activations, that I couldn’t explain that easily, even from podcasts, that I decided that I didn’t need it. But in Android Auto I couldn’t find to option to turn off the activation phrase, so instead I turned it off completely.
Ah, pronunciation must be different than I expected. In my head I did “eye-er” like in German, but I looked up Dutch and apparently that’s “ey-er” there, so way closer to “hey.”
It’s not surprising. Most of Google’s profiling is though the web. They don’t bother with the speakers and stuff too much because they don’t need to.
It’s not very useful anyway, any number of people could be using the speaker, regardless of who is signed in
Yeah, but the nsa can listen to you through it.
That’s not what this is about. That’s a different story
So even the most sophisticated profiling is wrong a significant part of the time.
Great. Glad its been so worth it to scoop up all my data and leak it everywhere just to not know how to use it for your stated intended purpose.
One thing we overlook in ideological debates is that most forms of socialism and communism would have no use for advertising and marketing. Imagine all this effort going to productive endeavors and having an IRL adblock.
May my last online purchase be a beret.
I loved my Google Home when I got it in 2017, but when I got into home automation, I realized it is dumb to have to tell a device to do things. Motion sensors basically replace the main thing Google Home does and a Bluetooth speaker is cheap to buy.
I never trusted that they weren’t listening to me all the time with the speakers and I never looked back after I donated them away.
Yeah, just having a microphone in the house with some predefined voice controls which you can go and change gives you all of the benefits of a Google home with none of the Google bullshit.
Especially now with LLMs getting so big, just go set up voice-to-text ollama session with predefined prompts and responses
That handles automation, but can’t freestyle questions.
“Hey Google, convert (metric) to (Imperial).”
“Hey Google, weather today?”
“Hey Google, what’s the capital of Kakistan?”
I have a ceiling-mounted mini in almost every room and just toss questions around while I work or play. Or, just ask it to play music. (Which went to shit when I cancelled Spotify.)
Or, just ask it to play music. (Which went to shit when I cancelled Spotify.)
I learned how to get Home assistant working JUST to restore this feature. Fuck you Google for not supporting other services despite supporting controlling them once it starts going ugh
It can if you use an LLM as the interface with guidelines instead of rules
I use Home assistant with the Preview box for local control, so I can spy on myself.
Have you ever caught yourself doing anything suspicious?
Only when I took Ambien
This incident will be reported
These devices are almost always listening, and the companies behind them are already collecting our data.
Uh oh! Clicks scary link
The good news is, they aren’t recording all the time … and when they are activated accidentally, the recordings are typically short.
:/ alrightythen
So what it’s saying is that it doesn’t need 24/7 recordings to profile us in a way that’s profitable for them because we are some basic bitches, right.
Just pointing out clickbait. They don’t even accurately reference their own articles. I stopped reading after that.
