Some of the hottest techno gadgets of the holiday season have now been opened and are positioned somewhere in your home just waiting for further instructions. If you’re one of the many shoppers who tried to purchase one of this year’s hot-ticket gift items only to find out they’re out of stock until after the New Year, that might not be the worst news to some privacy-savvy consumers.
Several companies have released home models of their virtual assistants (VA), as well as third-party accessories to go with them. Amazon’s Echo and Google’s Home are both compatible with their own lines of smartphone-driven, wi-fi-enabled outlets and appliances. With the right setup, you can tell your VA to turn on the lights in the living room, open or close the garage door, play your favorite song, or look up the show times to a newly-released movie. There are literally hundreds of functions that the devices can perform, depending on the model and the accessories you’ve chosen.
How do these devices work so well? They rely on lots of complicated artificial intelligence (AI) interface, but there’s an even more mundane mechanism: they’re recording and analyzing everything you say.
In order to understand your preferences and commands, these mini-audio sponges soak up your commands and send them to their servers where engineers can tweak the devices’ capabilities based on your voice patterns. They can also look at whether or not your command was successful—as in, “Alexa, play The Nutcracker Suite by Tchaikovsky”—and help the device learn from its mistakes. If your Amazon Echo played the Pentatonix version and you had to correct it, the device can “learn” which one you really wanted for the future.
Here’s an actual interaction with an Amazon Echo device from December 23, 2016:
- “Alexa, play Dance of the Toy Soldiers by Pentatonix.”
- “I can’t find dance songs by Pentatonix.”
- “Alexa, play March of the Toy Soldiers by Pentatonix.”
- “I can’t find the song March of the Toy Soldiers by Pentatonix.”
- “Alexa, play The Nutcracker by Pentatonix.”
- “I can’t find the album The Nutcracker by Pentatonix.”
- “Alexa, play Waltz of the Sugar Plum Fairy by Pentatonix.”
- “Dance of the Sugar Plum Fairy by Pentatonix.” And the music begins.
There are several shifts in the dynamic during that “conversation.” The device knew to look for a specific song by calling up previous information from its stored servers. When the command switched to “Nutcracker” in hopes that it would recognize it, the device knew that it referred to an album instead of a song; unfortunately, the group didn’t release an entire album called The Nutcracker, but rather only one song.
However, when the command was to play the semi-accurate song title—in this case, “Waltz of the Sugar Plum Fairy” instead of the correct title, “Dance of the Sugar Plum Fairy”—the device was able to make that adjustment without further input from the user. How did it learn to do that? Through its AI machine learning, something that is improved every single time any user around the world issues a command.
This listening and recording has some privacy experts on edge, mostly due to the potential for as-of-yet-unknown ramifications. Are servers storing our voice patterns and connecting those voices to our user accounts? Absolutely, it’s how these fairly expensive devices are improved upon. Are hackers or the government using our vocal patterns against us? No. Could they ever do such a thing? Well, that we’re less sure of.If you take issue with having your voice stored and analyzed, your only current course of action is to not purchase one of these home devices. It’s also very important if you do opt for a virtual assistant that you read the fine print and make sure you’re comfortable with the terms and agreements before you install it.
Questions about identity theft? Contact the ITRC toll-free at (888) 400-5530 or on-the-go with the new IDTheftHelp app for iOS and Android.