Amazon wants to know everything about you by having a robot talk to you

Talk to me.
Talk to me.
Image: Annapurna Pictures
We may earn a commission from links on this page.

Recently, Amazon made its first foray into helpful home devices—Echo, the perky remote-vending-machine-disguised-as-smart-home-hub that answers questions, controls a few lights, and, incidentally, orders products from the mega-company. The introduction of this small black monolith has brought another voice into the home that people can have a chinwag with, Alexa.

If you want to know the weather, you can wake Alexa up by saying her name, asking a question, or giving certain commands—and the disembodied voice in the corner of a kitchen or bedroom (or wherever Echo lives) speaks up with an answer. If you’re lonely, you can spend a night in chatting with it. Echo’s voice, Alexa, has become the front line in Amazon’s big push to become not just indispensable for books and diapers, but for informing customers as well.

Amazon isn’t finished making things talk, however. In late June, the Seattle company quietly announced the release of a cloud-based version of Alexa as a voice service on demand, and a set of tools that software engineers can use to build new voice applications for non-Amazon products. In short, the company made it possible for compatible objects—things created in startup garages and R&D labs of major appliance companies alike—talk like Alexa. The company has also put $100 million behind new applications that will use the voice.

Until now, voice capabilities and the ability to both listen and respond have been largely built into specific software, or piped into specific products from the cloud. For example, Siri only lives in Apple’s cloud and currently only speaks through Apple devices, and learns from the interactions with those devices.

By opening Alexa’s speech logic and vocal capabilities up to the world, Amazon pushes voice out to a a vast array of new applications, and more importantly, learns from more than just your interaction with Amazon or Alexa. Services like ticketing market StubHub have already added it for customers who prefer to sweet talk that best seat for a Taylor Swift show out of a machine.

Talking devices are learning devices

Voice capability is a kind of honeypot—a way for a company to find out what people think, what they need, and to learn from the real world. As with your Google searches, every interaction is a learning experience that feeds the great database in the cloud with new knowledge about what’s popular, interesting, or misunderstood. Every question you ask helps the company behind the sockpuppet voice application find out what’s important to you. These voices are essentially a brand speaking to you.

And most of these types of services, whether Google Now, Siri, or Alexa, monitor ongoing sound, once activated, in order to hear their wake words, like “OK, Google” or “Alexa!” and keep small amounts of ongoing audio stored for analysis. This means they are sitting in a kind of hibernating listening mode once authorized, quietly waiting to be asked for a joke, horoscope, author’s name, or maybe where to dispose of the indisposable.

This potential listening activity has been controversial in privacy circles, as users may not know they are agreeing to having snippets of conversations stored online, and some developers may not know the capability has even been added to software. Such a case of conversation stored without knowledge emerged earlier this year around Hello Barbie, a connected, talking version of the strangely ageless doll which sent data from conversations with the doll’s owners—mostly children—back to the servers of a company called ToyTalk in San Francisco. Not creepy at all, nope.

Hello, hair dryer!

With an artificial intelligence arms race emerging among the big consumer internet companies—now including Facebook with its release of a personal assistant called M—we will soon have dozens of things competing for our attention. They won’t just want to show us ads and get our clicks, but chat us up, and discover our hidden needs. Make those things talk and it’s going to get crazy loud in here. And some of these devices may even sound like Stephen Hawking. It’s only a matter of time before we get licensed celebrity voices in our thermostats.

The Chinese won’t be immune either, if Microsoft has its way. The company’s super hot Chinese chat bot, Xiaoice—or “Little Bing”—may be added to household appliances, according to the company. And as this experiment with talking to several personal assistants at once shows, what seems like an awesome little companion now could become a noisy, argumentative mess if too many things listen and talk.