Much of the buzz at last week’s Consumer Electronics Show (CES) in Las Vegas was around the new crop of voice-controlled “smart home” technologies. Dozens of companies touted voice-activated televisions, light switches, thermostats, showerheads—even an “intelligent toilet.” All rely on Amazon, Google, Apple and Samsung digital assistants as their main interface. Marketing hype for trendy tech is not exactly new at CES, but the voice-control frenzy begs a closer look what the technology can (and cannot) do—not to mention the obvious privacy concerns over companies continuously gathering data from nearly every room in your home.

Amazon Alexa, Google Assistant and other digital voice assistants rely on speech-recognition technology to interpret commands and determine appropriate responses. These assistants have been embedded in iPhones and many different Android smartphones for years. Taking this a step further, voice is the only way to communicate with the recently introduced Amazon Echo and Google Home “smart speakers” used to search the web and control other devices.

Speech-recognition software converts the sound waves in a person’s voice, captured by a microphone, into different vibration patterns. The software’s algorithms then use machine learning to train a device to identify those patterns as words and phrases. The more the algorithms catalogue spoken language, the better the software becomes at interpreting speech. Add in GPS, mapping software and data gathered by cameras, accelerometers and other sensors, and a voice assistant such as Alexa or Apple’s Siri begins to build context that helps it, for example, suggest a nearby restaurant and give driving directions to get you there.

In Demand

Demand for this technology has grown sharply, despite its shortcomings, which include difficulty picking out commands in a noisy room and understanding accents. In a Pew Research Center survey of 4,135 people in December the majority of respondents reported their voice assistants accurately responded to commands “most” (39 percent) or “some” (42 percent) of the time. Still, 16 percent of users said the assistants accurately respond to their commands “not very often.”

The technology’s greatest strength at the moment is its ability to almost instantaneously provide users with information or entertainment, says Andrew McStay, a professor of digital life at Bangor University’s School of Creative Studies and Media in Wales. “Digital voice assistants remove the need to type out requests—it’s a much more natural mode of interaction,” McStay says. But he adds the devices sometimes lack important contextual understanding of what a person is trying to say—which “can be quite frustrating.”

There are still kinks to iron out before digital voice assistants are as useful as touch screens, agrees Murray Goulden, a senior research fellow at University of Nottingham’s School of Sociology and Social Policy in England. “My main experience after using Alexa in my home for two days was telling the kids to be quiet so that I could communicate with [the device],” Goulden says. “Alexa’s voice recognition is impressive, but if there’s any background noise, then she’s hopeless.” The technology seems to be designed for individual users but “that is not necessarily the dynamic when you bring technology into the home,” Goulden notes.

Regardless of the technology’s faults, Google announced just before CES that its Assistant is now available on more than 400 million devices including Google Home, Android phones and tablets, and even iPhones. The company says it sold “tens of millions” of Google Home, Home Mini and Home Max smart speakers in 2017, and that Assistant can be used to control more than 1,500 smart-home devices such as air purifiers, precision cookers and alarm systems. Amazon claims to have sold “tens of millions” of Alexa-enabled devices worldwide during the 2017 holiday season alone. And during that time Alexa helped users mix tens of thousands of cocktails, switch holiday lights on and off, and search the Web for recipes, holiday music and jokes, according to the company.

The Cost

More revealing than what Alexa was asked to do, however, is what Amazon learned about its customers. Alexa-gathered data indicated the martini and the Manhattan were the most-requested drinks, chocolate chip cookies were the most-requested recipe and “Jingle Bells” was the most-requested song. Alexa also informed Amazon that the most common person people called this holiday season was “mom” in the U.S. and Germany, and “dad” in the U.K. Such intelligence gathering is not necessarily nefarious, but it certainly helps Amazon market products on its shopping site.

Voice interfaces are a way Amazon, Google and Apple can gather information—not just on shopping preferences and other internet activities—but also about how they behave and interact with one another in the home itself, McStay says. “Those companies already know what we do online and when we make purchases,” he notes.

Privacy concerns are not necessarily deal breakers for many consumers but they are certain to grow as the technology proliferates. Apple announced its HomePod smart speaker last June and intends to start selling it this year. The company already builds Siri into a number of its products including the iPad, Mac and Apple TV. Samsung introduced its Bixby voice-controlled digital assistant last year and plans to include the technology on all of its devices by 2020. The Family Hub refrigerator, a smart icebox that Samsung introduced at the CES, includes Bixby as well as a touch screen to control an internal camera, which can send notifications to household members when it needs to be restocked. Moen, a maker of faucets and other plumbing equipment, introduced at the CES a $1,200 digitally controlled shower system whose water temperature and flow can be controlled with voice commands via Alex or Siri. Home goods retailer Kohler took things a step further by introducing its Numi intelligent toilet, which includes voice-controlled ambient lighting in variable colors, Bluetooth connectivity and a heated seat and foot warmer.

All of this may seem invasive, to say the least, but for their part Amazon and Google insist their smart speakers do not record voices until someone directly addresses the device with a “wake word” such as “Alexa” or “okay Google.” It is possible to accidentally “wake” such devices, however, which means it is not always clear when they are listening.

It is also unclear how families will be able to easily manage data collected in the home, Goulden says. “You can’t manage access to that data the way you have in the past such as with passwords tied to a single user account,” he adds. “It isn’t clear who should have access to what data, because it will be contributed by multiple different members of the household. The privacy boundaries between us and those we live with are complex, both highly nuanced and changing over time—for example, as kids become young adults.”