Amazon Is Using Your Conversations With Alexa to Train AI
(gizmodo.com)
from FlyingSquid@lemmy.world to technology@lemmy.world on 26 Sep 2023 19:56
https://lemmy.world/post/5777847
from FlyingSquid@lemmy.world to technology@lemmy.world on 26 Sep 2023 19:56
https://lemmy.world/post/5777847
I’m never putting one of these in my home.
threaded - newest
I will be the last person to not have a smart home. There will be a banner over the doorway: “Welcome to Stupid House”.
There will be a small cover charge.
I’m with you. I hate how they expect me to control everything from my phone or with voice commands. I’m fine walking to a light switch or walking to the thermostat.
There’s a middle ground as well. I refuse to put Alexa or OK Google or whatever on any of my stuff, but I run home-assistant with zigbee smart devices. My entire setup runs completely cut off from the internet. I could in theory even air gap it, although that’s a little overkill. It’s a “smart” house, but one I’m 100% in control of.
Is that self hosted? I’d just about fuck with a FOSS self hosted smart home setup, but even then I could barely be arsed
Yes. You can host it on a pi if you want
That’s badass. I’ve got one lying around actually.
Be careful running it in a Pi because it’s a little heavy for that depending on how you configure it. A Pi model 4 is probably OK, but you wouldn’t want to run it on a model 3 or something even older, and you’re going to want to use one with at least 4GB of RAM.
Ah shit, the raspberry I’ve got is old as hell. Thanks for the heads up.
It will probably run even in a Pi Model 1, it’s just going to be a bit slow to interact with, and you’re not going to be able to do anything more complicated like enable the voice support (which you probably don’t want anyway, because I think it’s dependent on internet access for that, and then we’re back to the same problem as Alexa, although I don’t use it myself so I can’t say for sure).
Saving the idea for a spare weekend. Cheers!
For when you get bored:
www.home-assistant.io/installation/raspberrypi/
I’ll get a lot of hate for this but when you say pi you mean pi4. I kept seeing this HA on lemmy and tried it on a pi2. I don’t know if it worked or not, it’s a very bloated piece of software. After an hour of waiting I installed docker and the HA instance on my main server (which is ancient) in under a minute.
It’s cool and all but my feit dimmers require some pcb work and flashing to be compatible so verify what devices you have before you hop in.
I used to have an automated building running on a bare 386 and a floppy drive. Hate on me all you want but sending simple commands like turn device on shouldn’t require a giant software package but otherwise HA is neat, just a lot of overhead i can’t exactly justify.
Worth trying out though.
I think reflow stole a lot of their code.
No hate from me.
Just about every project I’ve started with a pi has ended up working out a lot better as a vm on an x86 host. But lots of people seem to love them.
To be fair It has its uses i suppose. I’ve had one running pihole since the original pi came out. Used PI2s in the past for OSMC and, even better, ambilight.
I think now a cheap android TV box you can flash is probably better for a simple less than 5watt device.
Besides the HA test I’ve been trying to use one to be an openvpn TAP interface but it’s been a fight and i think you just convinced me to do it in another docker instance on the server and save myself some headaches.
PiVPN was actually one of the things I thought the HW handled pretty well. Other than how much it ends up getting throttled by the 100Mbps link.
It seems like it would but the pivpn install script always hangs on me whether i I select openvpn or wireguard. Based on some reading I was lead to believe I needed to just use raw openvpn for a TAP interface. I’ve tried a few times but I always end up with CA issues or just flat out failure to connect.
It should be pretty simple, I’m just trying to bridge my network to a single remote device connected via cellular gateway. Maybe I’m just out of my depth. I’ve done it before with an old NAS years ago but I’ve tried a few step by step guides and no dice.
If you have any tips that’d be great!
Wish I could help.
I used it a few years ago for a simple wireguard link but eventually moved the vpn into my opnsense instance
Yep, I run mine on an ODroid XU4, but you can run it on just about anything including a docker container on a generic Linux install.
I’m using z-wave stuff but similar setup. Home Assistant does reach out to the cloud for some things like weather forecast and Google calendar but otherwise it will operate 100% without internet if needed. I also have cameras that while they aren’t air-gappend they are blocked from Internet access and can only talk to the NVR.
The last thing I want is to talk to a computer. Buttons are fine. The roboto phone customer service is bad enough.
When the hallway light was left on again it’s really damn nice to simply say “Turn Off Hallway Light” while staying under my nice warm covers. It’s also pretty swank to have the garage lights turn on when the garage door opens then turn themselves back off 5 minutes after the garage door closes. Someone left their closet light on? No problemo, my automation catches that and shuts it off.
Window coverings like blinds and drapes? Yeah, those are opening and closing automatically based on the position of the sun, even when I’m not home to do it. Did it rain while I was at work? Automation keeps my sprinklers from running tonight.
All of that is being done by Home Assistant and absolutely no Internet is required to make it work.
Can you use a custom wake word? The only reason why I’m still using Alexa is for the “computer” wake word.
HA doesn’t have “Wake Word” yet. It supposed to be coming before the end of the year. Right now you have to PTT (Push To Talk) to make it listen and for the privacy minded this is actually better than an always listening device.
Still, a lot of people want WW support and it is on the road map.
When skynet comes online, I’ll die quickly, being mopped to death. You’ll have to struggle in the post apocalyptic hellscape where humans fight robots with A-10s for some reason.
You can have a privacy-first smart home. I have. I run Home Assistant in a docker container. No external services/plugins. My smart doorbell streams to my local nvr. If my internet is down, everything keeps working. And it's not even that hard anymore. It's become a lot easier over the last 2-3 years. Still not for non-techie users, but a lot better.
That sounds pretty reasonable.
Edit: Still kind of want to call my place “Stupid House” for myriad other reasons
I’m not tech illiterate by any means, and everything after “home assistant” in that post is Greek to me
Docker is a way to run containers. Basically lightweight virtual servers. That makes it easy to run multiple servers on one machine. An NVR is a network video recorder. It's like a video security system like they use in stores where all cameras are viewed and recorded in a single place. I assume you know what a doorbell is 😄
Have any resources to get started with that? Been looking into security systems but don’t fully trust nest/ring/simplisafe etc
Just start with a local Home Assistant on. Raspberry Pi and go from there: https://www.home-assistant.io/installation/
Teach me your ways please! Setting up a Home Assistant seems like such a daunting task. I'm stalling converting my devices into it. Any tips for a (home assistant) beginner?
I just followed the steps on their site. Containers give me cancer, so I did a real install on my home server.
Caveat: I am a professional software engineer (but I didn’t really have to hack anything)
Start with a Raspberry Pi and just follow the docs: https://www.home-assistant.io/installation/
That's the easy way. I did it the hard way, but that's because I run on on a big dedicated home server together with a dozen other services.
Would’ve been newsworthy if it wasn’t the case
Yeah, that’s kinda the point. They literally tell you that your voice interactions are used to improve the service.
I gItz NufIN Ti HIIIDE
It’ll be really good at telling people to shut the fuck up if it’s using my data for training.
Of course they are. If you are surprised by this, then you are an idiot.
Harsh but true. We need some tough love in our relationship with tech.
Why people always so ready to call other people idiots? Idiot
Because most people are idiots.
Ok see that’s actually a really good point.
They literally tell you when you go through setup.
Well,that’s the thing with “news” right? Just scattered information without context for clicks. If people start connecting the dots and things make sense, most of the news become pretty uninteresting and would not evoke anger, prompting you to click and share.
I work for Amazon.
This has been the case for many years. Amazon has used AI in Alexa and other services for many years as primary providers, and has told it’s users it’s used it’s data for as long. We’re talking from close to inception here, so 6-7 years, at least. Hell, LLM’s aren’t even new to most big tech companies!
I’m all for privacy, but if you want privacy then you probably shouldn’t have a fucking tin can in your house that actions every conversation to a cloud service!
Not every conversation, just statements following a detected wake word.
You trust that?
Considering I set up one of the content types that relates to wakeword and utterance text analysis for Alexa, I trust it completely.
.
But can I trust you? Are you willing to share the source code?
Edit: Tell me why I’m suppose to trust an internet rando?
You’re right to be distrustful, but there’s a fine line between a healthy distrust of a closed ecosystem and blind worry/cynicism.
Obviously I’m not going to share proprietary source code. Even if I did, it would mean very little without knowing the upstream and downstream services. What I will say is that Amazon is at least honest about what it’s services do, even if it’s in the fine print. Customers are able to delete their data when they choose to, and if they do, there are serious (internal) consequences when stuff like data deletion and DSAR aren’t followed.
Also, it would very little without also inspecting every chip on the board. You could have easily written safe code, but the audio signal could also be intercepted before it gets to that point.
Alexa doesn’t solve any problems and only exists to make consumption easier. It’s not something I need to trust because it’s not something I or anyone else needs.
There’s this study for those interested in knowing more about how often these devices mistakenly record conversations:
…northeastern.edu/…/smart-speakers-study-pets20/
I bought an Alexa but I disable the mic. Do they still listen?
Who the fuck are buying these ? I really want to meet the person who has one of these in their home.
Hi I have one! The echo is actually a very useful smart speaker and produces good quality audio. Mainly I just use it to voice activate Spotify and for quick easy things like weather forecasts, trivia that comes up in conversation, and updates on package orders. Yes I realize it’s probably spying on me. No I don’t care that much.
Thats great. Try the google one as well. Its better than Alexa. I would say install one in the living room and second in in your bedroom. It’ll be a perfect home🥰
We either avoid the spying like the plague or we accept it into our homes. If we want a future with robots, those things are going to have microphones for ears, a speaker for a mouth, cameras for eyes, and some sort of smell detector for the nose. We’re just in the training phase for that experience.
Kids will adapt to it, older curmudgeons will try to stay away with their lamplight oil. I personally want a robot to take care of cooking and cleaning so I can enjoy other things in life.
Yes, it’s an ongoing struggle between convenience and privacy. I’ve noticed that the voices for privacy tend to be the loudest and angriest, yet they still choose to have cell phones and computers and complain on the Internet. Unless they’re Amish or in the deep wilderness, AI will eventually know a lot more about them than they would like.
A crazy number of people are buying products with Alexa in them. I don’t really get it to be honest. It seems like mostly just a novelty.
So retarded! I’ve stopped buying anything that connects to internet. The only thing in my household that connects to internet is my PC and phone.
I bought a phone mount for my car that had Alexa built in. The software fucked up after 2 weeks, but it’s still the best physical design for a phone mount that I’ve found. So it’s unplugged and got a dead battery and just holds my phone.
I have a firetv cube that doubles as an echo device, but sorry I don’t want to meet you.
I have 7 in my house. I have my lights, plant humidifier, TV, air filter, Aquarium equipment, and others all hooked up to them. You can run almost anything you need to just by using your voice.
Getting ready to watch something on TV? “Alexa it’s movie time.” Turns on the TV and Switches the input to the Shield and opens up our go to media app. Shuts off all lights in the house except the living room. Dims those to 10% and changes them to red. Just enough light to see your snacks and the remote by. Turns off the sump filter for the Aquariums in just the living room but leaves the in tank bubble filter going for better silence. All with 4 spoken words and no need to work with multiple remotes and apps in your phone.
That’s just one example. We have an obscene amount of routines setup for nearly ecery scenario we have run across and adding new ones is easy.
Hell… We even have a routine where if you say “Alexa, do you know Desi naach?” It plays Naatu Naatu from every speaker in the house at full volume.
Show me a FOSS system as plug and play, powerful, and as EASY to use as Alexa and I’ll drop Alexa in the trash tommorow and switch.
Lol suckers born every minute
Genuine question. What part of what I said makes me a sucker?
I only have so long to live on this earth. I’m not going to expend excessive amounts of energy on things that are unlikely to have any REAL effect on my life. I don’t give a shit if they listen in on me. I talk about work as a delivery driver (not Amazon), my wife’s work as a hotel front desk help, trackmania and Baldur’s Gate. Lots of state secret stuff there for me to be worried about.
Again as I said… find me an easy to use plug and play FOSS voice assistant system and I will throw Alexa in the trash.
I enjoy the home I have set up and I feel relaxed when I return from a long day at work. The Alexa system has helped make that a reality.
The best market for these will always be older people, and people with disabilities, both mental and physical. Voice interaction is a great interface when there is friction from a screen or physical input.
It’s also useful for those with families, as you can use it for shopping reminders, as an intercom around your house, etc.
They are listening to you even when you’re not talking to alexa, did you know that
What about just a firestick? That uses Alexa
Obviously. How else would it hear you say, “Alexa?”
I’d love a citation on this, outside of wakeword usage (a local device waiting for "Alexa* to begin recording).
Source: their ass.
Alexa devices use an onboard DSP to detect the wakeword and maintain a rolling audio buffer. On a positive match, the DSP wakes the main CPU which combines the saved buffer and any following speech and uploads it to the cloud where Alexa lives so she can try to figure out what you meant.
No audio is uploaded without being triggered by a wakeword. Also, the “mute” button physically cuts power to the mic, and the indicator LED is hardwired to the power rail as a failsafe indicator.
Oh, I know. For reference, I work at Amazon, so always interested in where these stories come from 😂
That was more for everyone else’s edification. I worked on a plethora of Amazon devices over the course of a decade.
Not going to get much out of me then, most of what Alexander hears is what’s on the TV or music I listen to. If they want to train alexa on that, their fucked
So who thinks this conversation here on lemmy isn’t being used to train an AI? Maybe not right now but later?
Sure the relatively small size of lemmy means it might not be scooped up and trained on. But the point still stands. All that is publicly online is food for the big-corp AI builders. And while Alexa invading your home privacy is obviously a shitty thing, I’m not sure we’ve all thought through the new relationship between us, the internet and the big AIs.
Well I know I have no expectation of privacy here, but I’d rather open source LLMs train on my words along with proprietary ones, than some company hoarding information and selling it to each other.
Stop usieng such grammar gud, and ur spellang soe good you be teach to them 2 good
The only thing an AI trained on Lemmy will ever be able to do is discuss the merits of socialism and talk about Linux lol
And none of it has paid off because Alexa is still super trash
I love being able to dictate a grocery list but god damn is she stupid.
Good luck asking for cream cheese and chive crackers without ending up with cream cheese as one item and chive crackers as another. Or worse peanut butter and honey crackers as peanut butter and then honey crackers
“chive crackers with cream cheese”
“honey crackers with peanut butter”
?
The problem is that Alexa isn’t actually parsing the meaning of the total phrase, she’s taking each individual word as it comes. With that context, she would just as easily interpret your phrasing as “thing with thing on the side”. You’d still get chive crackers, honey crackers, peanut butter, and cream cheese.
Edit: I thought about this a bit more, and it seems to me the only way Alexa could actually understand what you wanted is if you said “chive cream cheese crackers” or “peanut butter honey crackers”. You have to implicitly make it one item and not a potential combination of multiple items.
Yeah I had the same realization as I was reading this:
you can not add Boolean terms and expect it to not separate it.
So run on nonsense sentences are your friend and will definitely make the training from you so much more useful.
No wonder she’s been getting more stupid recently
Yeah, I realized these things are terrible about a year ago. So, I hacked them into computer speakers using some cheap amps and a 12 volt power supply.
not sure how much they’ll learn from me screaming “you dumb bitch” at it
Yep, most of my interactions lately that are non trivial are just sigh-inducing. Smart assistant my ass.
The new Amazon AI is going to be remarkably foul-mouthed. Every time it screws up (and it screws up a lot) I have to curse at it to make it shut up so it can hear the command again.
I brings me joy when I tell her “Alexa, shut up you dumb bitch” and then she responds with that sad minor tone dejected sound.
haven’t we all known this since product launch ?
I think most people, me included, underestimate the scale of the operation. When you hear “company will use private data to do X”, you imagine what a reasonable person would do, like random sample a few conversations here and there. In reality they record everything permanently over months and years, far beyond what would be necessary to run the service.
It’s kind of crazy how we get this level of surveillance while still having software that will lose your data if you don’t hit
Saveoften enough.that’s fair. i work with data for a living so that probably biases my perspective
What’s fucked up is if you try to regulate it and make these companies have data retention policies. It creates a giant moat around them where no newcomer can have a chance to compete.
That’s because you are not enforcing data portability at the same time. Having studied and discussed the GDPR at length within tech circles, I became convinced that data portability is the ultimate right and the key to ensure continuing innovation
Interoperability in general is the solution to walled gardens and monopolies that harm competition, consumers, and innovation.
Noooo reaaally?
Do things on your own as long as you can
I don’t necessarily disagree with you, but maybe you could tell me why?
it is a link to another post with few, but not limited to, reasons
WTF is this bullshit. Not everyone is equally capable and this reductionist shit is hurting those that do need the help.
well the title says “as long as you can”, not sure what is the issue here, just do whatever you can, don’t fold if you don’t need to, that’s it
We always knew that. What they don’t tell you is your phone is also secretly listening. “Ok Google” <- turn that thing off too
And Pokémon Go uses your location for its gameplay
That’s kind of how this shit works…
These types of projects are driven by metrics, and teams have some kind of quota/goal that they need to reach by a certain date to keep the project on schedule. Bonuses or job security may be on the line here, and so you may see some desperate employees “going the extra mile” to reach their goals.
Relatedly, Alexa’s voice activation sensitivity is essentially a tunable number. It can be changed to be more sensitive, so that it will activate more easily (e.g. maybe you say “Alex” instead of “Alexa”). The people who control this are likely on the team with that deadline, so the incentives are there to lower this value in order to collect more data by recording personal conversations “accidentally”. Maybe a bad update goes out that causes Alexa to activate randomly, and they quickly fix it after a few days when they collected all the non-Alexa personal conversations they need for their AI.
That’s maybe a bit too deep into the paranoia/tinfoil hat spectrum for some, but history has shown that you can’t give big tech the benefit of the doubt. Especially when you see some of the documents from the Google trial, where executives discuss rolling back new features to improve arbitrary metrics in the short term so that they can get their bonuses for the quarter, even if it hurts consumers.
An always on microphone connected to a company that is mostly known to exploit their customers and employees! Say it ain’t so!
Is this a surprise to anyone?
This was already my understanding when I got the first pre-release one in 2014.
In that time, it has mainly learned how to"dim the living area lights to 50%" and “set the AC to 22 degrees”. That is about 99% of it’s use.
Wonder if that’s helped it’s AI much…