By David McIntyre 2020-10-05
Amid widespread social distancing, many people are spending more time in their homes than ever before and with that, they’re taking a critical look at their surroundings. Scrutinizing my own home, I like to dream up improvements beyond new paint colors and framed photos. As a long time tech enthusiast and a market strategist for an edge inference company, my mind goes to smart home innovations.
Imagine a microwave that recognizes you as you walk up with a plate of bacon and, without constant monitoring, knows how long to blast it so it’s cooked just the way you like it. Or an autonomous vacuum that can locate your favorite pair of shoes, so you don’t have to search for them in a frenzy.
As futuristic as these scenarios may seem, from a technical standpoint they’re not so far off. With advances in artificial intelligence and machine learning, and new edge inference chips bringing unprecedented computing power on-device, innovations long-kept for sci-fi books and movies could hit the market very soon.
While we’re close to taking big steps in smart home technology, there is one hurdle we must surpass before old and new devices can lay a proper foundation for the next-generation smart home. The devices we already have—gadgets like virtual assistants and smart speakers, smart home security systems and network-connected appliances—need to operate at a higher standard. Until then, we risk delaying smart home advancement.
Smart home devices must be accurate
First and foremost, in order to be useful a device has to be successful in doing what it was designed to do. Put simply, it must be accurate. This means that your voice assistant understands the intent of your command the first time, not the second or third; your face recognition lock recognizes you even when you’re wearing glasses (or better yet, a mask); and your smart home camera doesn’t constantly trigger false alarms.
As obvious as these examples may seem, accuracy is an area where many of today’s smart devices fall short. The reason for this, I think, is in part because the standard neural network (NN) benchmarks technologists use to test a gadget’s functionality are often not reflective of real life.
As a result, it’s not uncommon that a device does well in testing but performs poorly in the real world. For example, a video motion detector that’s trained to successfully recognize a change in pixels could trigger an alarm in response to nothing more than a moving shadow. Built with the purpose of protecting someone’s home, that motion detector isn’t “smart,” it’s just seeing pixels change.
Companies are constantly refining smart home innovations in hopes that they can work flawlessly right out of the box. Until then, when today’s smart devices fall short on accuracy, users end up frustrated, and rightfully so. While fleeting, these moments of friction stand in the way of what smart home living should be: a seamless integration of technology in our daily lives.
They should never put consumers at risk
It’s no secret that smart home devices have struggled to remain secure—even devices designed to make homes safer have left consumers vulnerable to attack. It’s common knowledge, for example, that Ring, who makes some of the most popular home security products on the market, was plagued with countless security breaches last year.
Voice assistants have also had difficulties. Recently, researchers found that people can unintentionally trigger an assistant with more than 1,000 words and phrases, prompting it to record what you say and then send that recording to the cloud for wake word verification. This raises obvious privacy concerns. Any recording or transfer of data off of a personal device should be initiated by its owner, not triggered accidentally and without their knowledge.
Across devices, security and privacy are clear issues. To reduce risk, I believe we must bring more horsepower to the gadgets themselves and limit the need to expose data by sending it to the cloud. Until network-connected devices prove that they are secure, consumers will be left to question whether the home enhancements they promise are worth a threat to their privacy and security.
You have to be able to trust them
Having experienced accuracy blunders and aware of security concerns, many consumers are skeptical of smart devices at home. In fact, a recent PwC report pointed to lack of trust as one of three factors inhibiting voice technology experimentation. The report quotes a respondent expressing hesitation to trust a virtual assistant with sensitive information, saying: “The assistant can’t answer my questions half the time, but I’m supposed to trust it to help me with something involving money?”
These frustrations are only heightened in quarantine. As the title of a recent article comically made clear, “We’ve been isolated for months, and now we hate our home assistants.” Only after experiencing consistent accuracy, privacy and security will consumers learn to trust their connected devices and embrace the modern smart home.
Recent years have brought unprecedented advancement to smart home technologies, but there’s significant room for improvement. To pave the way for continued innovation, technology leaders should invest in refining the devices we’ve already welcomed into our lives. Connected devices that consumers can trust to be both accurate and secure will provide a solid foundation for the next-generation smart home. With that and sooner than you think, your microwave could know how you like your bacon.