After showing off plans for ambient computing and on-device voice processing at Google I/O in May, Google began to deliver on those strategies. At its Made By Google event in New York City, the company announced updated devices for the smart home and demonstrated local natural language processing brought about by the ability to run small machine learning models locally.
Among the devices introduced are the new Nest Mini, two Nest WiFi units (which we broke the news on in April), and truly wireless Google Pixel Bud earphones, not to mention the Pixel 4 handset and Pixelbook Go Chromebook. And to connect them all into a more cohesive smart home platform is the updated Google Assistant that Google also previewed back in May.
That was the main theme of Google’s event: A combination of ambient computing — with new or improved sensors — and faster response times thanks to a hybrid natural language system that uses both local machine learning and, when needed, the Google Cloud for intelligence.
This means that Google’s two main software platforms — Android and Chrome — truly have a sibling in the Google Assistant. The recently announced feature that lets you “move” your streaming music or video from one Google Home device to another in the house with a simple voice command is a great example of this. Instead of each Google device working separately, often with different Assistant features or functions, all of the Google devices in your home will work better together.
I also think the move towards on-device processing will be a big factor in making this ecosystem better.
By training the voice processing models in the cloud and then pushing the results of that model down to edge devices in the home, Google is taking a stab at improving the privacy associated with recorded voice commands. And privacy is not coming at a cost, but instead, with an added benefit: Localized voice interaction typically brings faster response times for the devices in our homes.
Google showed off that speed during a Pixel 4 demonstration using a new Voice Recording app. For about nine minutes, unbeknownst to the event audience, a Pixel 4 phone was recording what the presenters were saying. That’s not impressive, of course, but the fact that the recording was generating a highly accurate transcription in real time was. And there was one more thing: After showing off the transcription, we were told that the phone was in Airplane Mode. The entire demonstration took place while offline.
Of course, we typically don’t want our Google Home and Nest Mini devices to transcribe our every spoken word. The point here was to show that Google has made big strides in on-device voice services using relatively small machine learning models running at the edge. As this functionality makes its way down to Google’s smart home devices, we’ll see lights turn on or dim more quickly without relying on the cloud. As a byproduct, this also means that our smart home devices stay smart even if we lose our home internet connection.
As far as ambient computing goes, we’ve already seen a glimpse of that in the currently available Google Nest Hub Max. If you give it permission, the device uses its camera to detect who is nearby and then shows relevant contextual information for that person. The Google Nest Hub Max might show the individual’s next appointment or her commute to work, for example.
With the upgraded Google Assistant, things go a step further. If sensors in your home don’t detect movement and it’s late at night, the Assistant could check and see if you left any lights or appliances on and shut them down for you. Using the radios in your phones, the Assistant might determine that you’ve left home but didn’t enable your alarm system. It can either notify you to take action or even turn the alarm on for you.
Essentially, Google is getting closer to delivering more “smarts” to the smart home, which I noted earlier this year is a current gap in this space.
But again, it’s not just sensors, silicon, cameras or microphones that can accomplish a truly smart home. You need a system to put all of these ambient signals together so that the right things happen in the home at the right time. And that system is the new Google Assistant.