Artificial intelligence (AI) is coming to the smart home. A few weeks ago, Amazon announced the Intel-powered AWS DeepLens, a fully programmable, deep learning-enabled wireless video camera designed for developers. This breakthrough solution puts AI at the edge. My colleague Naveen Rao recently discussed the value of AI at the edge, both in business and at home. I’d like to expand on the possibilities of AI as part of the smart home and what experiences it can deliver to end consumers.
Think of the peace of mind that comes with a security system that can tell the difference between a family member and a stranger. Or the convenience of knowing as soon as a package lands on your porch. How might your home help prevent a serious accident—for example, by recognizing and alerting you to a small child approaching a hot stovetop? Maybe your smart home will do a lot of little things that just make life better, like following you from room to room with your favorite podcast as you go about your chores. In short, your smart home can serve as a personal assistant, retrieving the information you need and performing everyday tasks.
But for any of these things to happen in a practical fashion, we’ll need to make smart home devices truly intelligent. That means giving them cognition. If we want smart homes to do everything we imagine, we’ll need to allow them to think for themselves within certain parameters—and they’ll need to think a lot. The sheer amount of data generated in the smart home every day makes it impractical to shove all analysis up to the cloud. Most smart home activities require analysis on an ad hoc, reactionary basis, with findings that need to be acted upon quickly. Depending on the cloud for this analysis would mean a longer wait for tasks to be completed, not to mention a drag on bandwidth.
There’s another important issue here, and that’s privacy. Whether it’s who’s visiting our homes or what images our cameras are capturing, there’s just some data we don’t want to share.
So, what does it really take to bring AI out of the cloud and into our homes?
Devices that think for themselves
To start with, smart home devices will need to hear and see. The first requires the ability to capture clean signals and recognize natural speech. The second requires vision, the ability to immediately process and understand video, sensor, and motion data.
Once they’ve captured the auditory and visual information, devices will need specialized compute to perform inference—the ability to make a judgment by applying real-time data to an algorithm.
The right AI in the right place
Of course, some AI will continue to happen in the cloud. Smart home systems will need to use data to continually refine their models. There will also be countless opportunities to extract value from other data sets—for example, knowing the timing of package deliveries or amount of pedestrian traffic in a neighborhood.
This is why making AI work in the real world requires a comprehensive approach. We’ll need sophisticated AI platforms to accelerate the training of massive data sets in the data center, as well as specialized processors and algorithms for edge compute to bring AI down to the device level.
Connectivity throughout the home
With so much focus on AI, it’s vital to not forget home connectivity itself. Within the smart home, many devices must be able to communicate with each other, share data over the network, and filter and send data to the cloud.
The smart home will rely on next-generation connectivity both into and throughout the home. This calls for investments in a wide range of technologies, including cable, fiber, 5G, and Wi-Fi, as well as continuing research in access point security.
Better innovation together
Bringing AI-enabled smart home products to market will be a collaborative effort. Consumer brands will need to work closely with technology partners to make these new solutions a reality. How will your organization forge new partnerships to bring AI from the cloud to the smart home of the future?