Confession: I have an Amazon Echo. I really like Amazon Echo. I use Amazon Echo almost every day.
Everything about the Amazon Echo is great, except for the primary feature of the Amazon Echo: it is always listening. When I received the Echo nearly five years ago, as a gift, Amazon was not quite the Surveillance Capitalism behemoth that it is now. They packaged their new smart speaker with lots of information about privacy and what Echo can and can’t and won’t do.
Of course, none of that turned out to be true. In just the last year, Echos have been turned into permanent recording devices, listened to a couple’s conversations and then inexplicably sent those conversations to the husband’s employer, and sent 1,700 voice recordings to a totally random stranger. Amazon hasn’t exactly done much to help the image of Echos as Bradburian household horrors, unveiling an Echo Dot for Kids, filling patents for true always-on recording, releasing recordings to outside contractors, and, perhaps most egregious of all, embedding Alexa into a Big Mouth Billy Bass.
It’s reached the point where no one should feel comfortable having an always-on speaker in their home, but damn if these little things aren’t just so convenient. On top of being useful for quick searches, playing Baby Shark on repeat 40 times, checking the weather, and dozens of other little things, the original Echo was a really good speaker. It seems a waste to throw the whole thing away just because one feature is unacceptable.
If you’ve read any of my tech-related articles in the last few years, you know that my whole schtick isn’t that we should just avoid problematic technology, but rather that we should take control of the technology that dominates our lives. If you want to make and sell a smart speaker, you better be prepared to accept my terms for allowing it into my home. So, I asked the question:
“Is it possible to get all the benefits of an Echo without all of the problems inherent in allowing Amazon to have a direct wiretap into you home?”
Active Intervention: Building the Alias Parasite
My first plan was to build the Alias Parasite. Alias is a Raspberry Pi-powered AI platform that sits on top of your smart speaker, pumping white noise into the system until it hears its own wake word, then triggers the speaker underneath. Amazon is listening, but all it hears is noise. It’s a clever solution that has the benefit of letting you set a new wake word.
I built one, following the instructions provided, added my own hacks so that it could draw power from the Echo rather than a separate wall plug, gave it a new hat, and trained it to respond to the name “Orwell”.
As maker projects go, this is a really fun build that teaches you a lot about working with Raspberry Pi and voice recognition AI. The final implementation was not quite as fun as the build. The interface was finicky, and I could never get it to consistently respond to its new name. Even when it did respond, it was really slow. The white noise was subtle, but just loud enough that you could hear a low hum if you got close. Plus, this is an active system, so it draws power and adds an entire separate point of failure to the system.
Fantastic in theory, elegant in execution, and mediocre as a long-term solution.
Surely there must be a simpler way?
Passive Intervention: Removing the Microphones
While the Parasite was perhaps needlessly complex, the solution that ended up working the best turned out to be dead simple. An Amazon Echo has an array of precision microphones mounted around the ring of a circuit board. What would happen to Alexa if those microphones just… weren’t?
I pulled out my trusty iFixit kit, opened up the Echo, ripped the seven microphones off the main board (don’t forget the one mounted in the middle), and closed everything up.
I have the hand it to Amazon, they made the Echo extremely easy to open and disassemble. A+ for repair and hackability.
There’s a lot of ways this could have gone wrong. The microphones could have been critical to creating a complete circuit, causing the whole thing to shut down. The software could lock out if it doesn’t see the microphones. Or the Echo could completely glitch out without access to its main sensors. But I gambled on the fact that Amazon was smart enough to build the system around the idea that eventually a microphone or two might burn out and that the Echo could handle the trauma of losing all 7.
I was right.
Here’s what happens if you physically remove every single microphone from an Amazon Echo, rendering it incapable of listening to every word spoken in your home:
Nothing.
The smart speaker continues to function exactly as if it still had all its microphones, it just can’t hear anything. The physical button and dials do exactly what the physical buttons and dials always do. If you send it a command through the app, it does what it’s told. It can’t hear you, but that’s the easiest fix in the world because…
Amazon sells a remote for the Amazon Echo with a low-quality, never-on-unless-you-push-a-button microphone. It works perfectly, as long as you are close to the microphone and press the microphone button. It can’t listen constantly, because it doesn’t have a continuous power supply. The Echo is just as responsive (honestly, more so, because it doesn’t need to figure out the wake word).
Welcome to the future. There is absolutely no reason your smart speaker needs to always be on, and yet Amazon won’t sell you an always-off-unless-I-say-so option (Amazon could probably just re-purpose their crates of disused Dash buttons). If you can handle keeping track of a little remote, you can keep all of the benefits of an Amazon Echo without the drawbacks of wiretapping your home.
This is a really weird future. We’ll just have to keep hacking away at it until we get the future we want.
Southern Fried Science is free and ad-free (other than the occasional use of Amazon Affiliate links when we talked about consumer products). If you enjoy Southern Fried Science, consider contributing to our Patreon campaign.
