Google Home knows who's talking.
When each recognized user speaks Home's wake words — which are "Hey, Google" and "OK, Google" — Home now switches to that person's account to give personalized information.
This is a huge deal, a technological marvel and a major leap forward in human-machine interaction.
It's also a cautionary tale about the downside of overzealous concern about privacy.
How Google Home knows you're you
Google Home is a virtual assistant appliance similar to the Amazon Echo that runs a version of the Google Assistant. Assistant is also available in Google's Allo messaging app.
Google Home now supports up to six users for a single device.
The new capability started rolling out to U.S. users this week, and Google promises U.K. availability "in the coming months."
When the feature becomes available to you, the Google Home mobile app will present a card that says "multi-user is available." By tapping on that message, the app will walk you through the setup process. Alternatively, you can tap on Devices, Settings, More, then Shared Devices to drill down to the multi-user options.
The setup process enables you to initially "train" Home to know your voice and distinguish it from other people's voices.
This user recognition is only for the wake words. It does this using neural network artificial intelligence, according to Google. While there's a required initial training process, the voice training continues in the background to continuously refine recognition, according to a Google representative.
Before we move on, let's savor this technological achievement for a moment. Google's A.I. is so good that with two repetitions each of "Hey, Google" and "OK, Google," the Home device will recognize you when you say those words in the future. The processing takes place not in some massive remote data center, but right on the Home device itself. And it gets better over time all by itself.
Why this is such a big deal
Google's announcement pointed out the obvious benefits of user recognition, which include the ability of Home to offer "my own personal playlists, my own commute time, my own schedule and more."
Already, privacy advocates are complaining. One expert told the The Washington Post that "consumers worried about their voice data being collected in general should think twice before picking up a home hub" and that the feature "could be used in ways that consumers may not realize — particularly if it's combined with other information tied to your Google account."
This advice is misplaced.
A Google representative told me that the identification of particular users on Google Home is "kept local on the device and not used for other services."
It's also important to realize that the voice recognition is used exclusively for switching from one account to another. Once the account is switched, Home stops recognizing your voice and the voice data (including the audio recording) is then associated with your account.
In the past, all data was associated with the account of the person who set up the Home device in the first place. For example, let's say mom bought a Google Home gadget and set it up in the living room using her Google account. From that point on, whenever dad, the kids, friends or the in-laws used the Google Home, all that data (regarding the content of the query) was associated with mom's account, and was therefore almost useless for improving Google's understanding and personalization of services for mom.
Now that Home can switch accounts with each user, data is theoretically usable in the same way that Google Search data or browsing history or YouTube searches are usable. Google could harvest data from your Google Home device, including your music, food, travel and search preferences, and combine this with other signals as you use the Internet while logged in to your Google account to offer better suggestions and more relevant advertising.
Google is transparent about recording your voice, and makes all your recordings available to you in a searchable format here on the Google Dashboard.
This is no different from using Google Assistant on Allo, where activity is faithfully associated with the proper account. In fact, when you use any virtual assistant on a smartphone — including Apple's Siri, Microsoft'sCortana, or Samsung's new Bixby — your spoken-voice command and queries are connected to your individual personal accounts.
In short, Google Home's new ability to recognize users does not represent a new privacy threat. Not even close.
It merely brings virtual assistant appliances to parity with smartphone-based virtual assistants in terms of associating activity with specific users.
In fact, the ability of Home to recognize users and switch accounts accordingly is a positive and inevitable development.
The glorious future of user-voice recognition
If someone uses your Google Home and the user isn't recognized, Home will work anyway.
The ideal scenario for Google Home and all voice-based virtual assistants would be for the devices to work only when they recognize the user.
This feature would require higher quality user identification technology, which is surely coming, and would enhance both smartphone- and appliance-based virtual assistants.
Burger King recently used a 15-second TV commercial to trigger Google Home devices. The actor in the add said, "OK, Google: What is the Whopper Burger?"
This shameless and irresponsible stunt was shut down by Google, which instructed Home devices to ignore the ad. But in the world of viral campaigns, there's no such thing as bad publicity. The add didn't trigger Google Home devices, but did trigger the media, which went on to write more about Burger King and Whoppers than it has in recent memory.
Still, if virtual assistant appliances didn't respond to strangers, then future ads (or podcasts, radio shows and TV shows) wouldn't be able to interact with your device without your permission.
A greater concern is when virtual assistant appliances are connected to home appliances. Right now, a stranger shouting through a window at an Amazon Echo or Google Home device can theoretically turn on whatever's connected — the lights, oven, TV or sprinkler system — or even unlock the front door. Today, it's a very bad idea to have your doors, windows or burglar alarms activated by voice.
But in the future, when such appliances are really good at recognizing voices, we'll be able to connect all of our home devices without fear of unwanted strangers triggering them.
Unless privacy fears prevent it, voice recognition will be so good we should be able to use our voices to log in to all kinds of services on other devices, including other people's virtual assistant appliances, but also ATMs and security gates. Eventually, we could start our self-driving cars using voice alone. (We could get into the car and say "OK, Google Car, take me to work," but if someone else does that it wouldn't go anywhere.)
Google's caution in how the new user identification feature is implemented reveals how gun-shy Google is about violating user privacy.
While actual biometric privacy violations are a real and serious threat, the secondary threat is that ignorance and fear will prevent us from gaining enormous benefits.
Voice identification can, should and must be made very secure, enabling us to enjoy what's possible without compromising our privacy or personal information.
It will always be more secure and less prone to abuse than face recognition, as I explained last month.
Most importantly of all, voice is a democratizing interface.
Millions or possibly billions of people cannot use smartphones to any significant extent. We all have older parents, grandparents or great-grandparents who can't or won't use most of the features of their phones and see them as too complex. Others don't have the physical mobility to use screen-based interfaces or keyboards. (I have one relative who never communicates online because he finds computers too complicated and typing too daunting.)
It can be easy to dismiss this problem as insignificant, even as it puts non-users at an increasing disadvantage over time.
Our mobile devices, computers and networks are sophisticated enough now to enable the full power of the internet to anyone who can talk.
As Google Home's newest feature demonstrates, "just talking" can now enable people to log in to their accounts as well.
So let's proceed with caution, but also perspective and optimism.
In other words, let's hold companies accountable for privacy. But let's also not pressure them into hobbling future technologies out of unreasonable privacy fears.
We need to continue down this path and enable the full development and wide distribution of simple voice-based user identification of the kind that now exists in Google Home.