Over the last couple of weeks, an increasing number of users on social media have reported that their Amazon Echo devices are laughing, seemingly at them, at unexpected and unwanted times. Not a ton of people, but enough to gain some media attention, largely because it’s a fun story — the device reliant on artificial intelligence that is always listening to you developing the capacity to laugh at you to your face. It contains all of the ingredients for a minor tech panic, with ample opportunity for riffing, the kind you might see in a 60-second clip on the local news.
A day after the viral posts reached a fever pitch, Amazon produced a statement claiming that the problem was the Echo devices incorrectly hearing themselves being commanded to laugh. “We are disabling the short utterance ‘Alexa, laugh.’ We are also changing Alexa’s response from simply laughter to ‘Sure, I can laugh’ followed by laughter,” the company said in a statement.
You’ll just have to take its explanation of misheard commands at face value: None of the processing is done client-side, there is no way for third parties to look at how Alexa devices really work, to poke around in the guts and discover causes and effects.
The story of Amazon devices laughing at users is easily taken as a parable about our anxiety over increasingly sophisticated consumer-facing artificial intelligence, or maybe an object lesson in bad user design. But there’s another warning to pay attention to. The laughing Alexas lay bare, more than ever, the need for transparency from the companies that make these devices, whether that company is Amazon (Alexa), or Google (the Assistant), or Apple (Siri), or Microsoft (Cortana). The problem with smart speakers is that they are, for reasons both incidental and intentional, almost impossible to troubleshoot.
To understand why this is important, think about how you might deal with a problem on a computer (or how the person who you call to deal with problems on your computer would deal with a problem on your computer). Maybe you’ll run diagnostic software, or take a look at the log files or activity monitor, or try to figure out which component is buggy and install a patch for it. Maybe you just turn it off and on again.
You can do none of these things with smart speakers (save for turning it off and on). Much of this is attributable to the fact that these devices are “headless” — they operate without a screen or inputs such as a mouse and keyboard. It’s very difficult to diagnose problems when you literally cannot see how the computer is processing information.
Yet even if we could poke around the internals of a smart speaker, there wouldn’t be much to find. The entire premise of the product line is that these devices are “cloud-enabled”; that they’re connected to the internet 24/7 and do very little computing locally. When you say “Alexa …” or “Okay, Google …” the speaker records what you say next and sends that audio to a remote server controlled by Amazon or Google, respectively. It is on that server that the audio is analyzed and the speaker figures out what you want, and then the response is sent back to the speaker on your living-room table.
Nearly all of a speaker’s ability to parse and respond to commands comes server-side. The only client-side computing a smart speaker performs is understanding when it hears the activation phrase (and other obvious functions like playing audio). In other words, if you were to look at the software in your Amazon Echo, you would likely find a small program whose only task is to figure out when you say the name “Alexa.”
This is a remarkable shift in how we interact with technology. Rather than installing apps, and maintaining software versions, all of the maintenance and upgrades are performed remotely on servers outside of your control. The ease of use explains why the devices are so popular — talk to it like you would a person, rather than tapping and clicking through menu interfaces. But it also means that you are at the whim of these companies. They can make any change, at any time, and you are powerless to stop them. When you buy a smart speaker, what you are really buying is an umbilical cord that ties you to a technology monolith.
To illustrate how useless these devices are without a server connection, consider this: If your Google Home or Alexa speaker loses connection to the internet, any alarm you set on them will not go off (I know this from personal experience). These things cannot even function as a clock without access to the mother ship. Maybe “cannot function” is the wrong term — surely these devices have the basic components to act as an alarm clock locally, and keep time even when they lose internet access. Yet they are intentionally designed to cease function entirely.
Smart speakers are made to get you used to the idea that you literally cannot function without these large companies. They are the only services capable of running your alarm clock, or playing music for you. So when your Amazon Echo starts laughing at you, there is something slightly nefarious at play: complete helplessness. There is nothing you can do to fix it.
The tech industry, centralized around a handful of gargantuan companies, has moved far away from openness to a model of “walled gardens,” platforms that give users a modicum of control, not nearly as much as they used to. On the one end is your standard personal computer or laptop, which you can tinker with and monitor as you see fit. Somewhere in the middle is the smartphone, whose operating system is overseen by a gatekeeper who often gets to decide what software you can and can’t run on the device. On the other end is the smart speaker, a black box of voice parsing and artificial intelligence powered by software that only its developers in Silicon Valley really understand. Moving consumers toward the smart-speaker end of the spectrum is the long game.
Whenever you connect a Wi-Fi light bulb to your smart speaker, or link another product to it, you are assembling a powerful program that can control many different parts of your digital identity (and your home!) on a computer that you don’t own and can’t access. No wonder Alexa is laughing at you.