Skip to Main Content

Strategic Business Insights (SBI) logo

Connected Homes November 2017 Viewpoints

Technology Analyst: Michael Gold

Distributed-Computing Developments

Why is this topic significant?

Computation is not restricted to a single location—not the cloud, not the home, and not the smartphone. Increased use of distributed computing is reducing operating costs for suppliers and improving security for users, but all parties are somewhat poorly prepared to manage increased complexity.

Description

Many home-network applications execute mainly in the cloud. Virtual assistants (Alexa, Bixby, Google Assistant, and Siri) work only when cloud-connected. The same is true of various independent automation-logic services (If This Then That, Microsoft Flow, and others). The most popular mesh-network routers (Insteon Hub, SmartThings, and Wink Hub) also tend to require a working cloud connection at all times.

In contrast, home-networking solutions from a few other vendors make use of "smarts" in local devices for a variety of purposes. Separately, Securifi's Almond 3 router and PC-based solutions from HomeSeer respond to sensors, schedules, and smartphones in close proximity to control thousands of kinds of devices—lights, appliances, door locks, security systems, and so on—even if cloud connections are severed, as long as home networks remain connected. An Apple TV or always-on iPad can perform similar feats with a limited but design-conscious catalog of compatible devices.

Another such collection of devices is likely to emerge for partners of Nest. In a home equipped with a smoke detector and a thermostat from Nest, if the smoke detector activates, then the thermostat turns off climate controls, even if the home's broadband connection is unavailable. When a user leaves home, a Nest thermostat detects that the user's smartphone is not present and responds by activating a Nest security camera. These and other automation decisions occur inside the thermostat. So far, Nest's cloudless capabilities have been limited to its own products. However, as a result of recent changes in strategy at Nest, partners will now have an increased ability to exploit the computing capabilities in Nest's thermostats. For example, if a smoke alarm activates, such a thermostat will automatically generate a command to unlock a door that is equipped with Yale's new Linus lock, scheduled for a 2018 release.

Implications

One benefit of fortifying the computational capabilities of devices in the home rather than simply relying on the cloud is that applications continue to function even when broadband service is interrupted. Keeping messages off public networks also reduces opportunities for hackers to infiltrate home networks. Moreover, increased use of in-home smart devices reduces suppliers' own costs for communications and cloud computing. Those costs can be especially significant for makers of security cameras or any associated cloud-based video-storage services.

Impacts/Disruptions

Although a digital thermostat is not an especially smart device, it contains a microcontroller that is always on, chronically underemployed, and a genius in comparison with a light bulb or a door lock. Other potentially smarter always-on gear that could be constantly ready to respond to a sensor, to execute a scheduled event, or to monitor for the presence of a user's smartphone include set-top boxes (as in the case of Apple TV), routers (as in the case of Securifi's Almond 3), burglar alarms, and other home-security gear. In general, a supplier does not need to maintain an ongoing cloud service for the routine operations of a simple device that relies on a more complex device for control. But with automated setup procedures, routers, set-top boxes, and thermostats might quibble among themselves about which is in charge and which is the smartest. The makers of these various devices will surely do likewise.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: High

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: Now to 10 Years

Opportunities in the following industry areas:

Electronics manufacturing, embedded systems, semiconductors, communications services, web services, standards development, home security, cybersecurity, safety, smart grid, green buildings, illumination, infotainment

Relevant to the following Explorer Technology Areas:

Intelligent Video Surveillance

Why is this topic significant?

Home-security camera offerings are improving their ability to notify users when they detect events of importance—not just when they detect motion. Users must configure software for the devices, but suppliers' varying approaches to improving peace of mind point to a road map that promises to simplify user experiences.

Description

The vast majority of security cameras are used for monitoring by security guards and other personnel, and for reviewing recorded past events. But for more than 10 years, several companies have been selling artificial-intelligence software for businesses that wish to automatically generate alerts when software recognizes images that indicate security risks—baggage left unattended, impostors who trail behind a badged employee at a security gate, and so on. Recently, similar offerings for home-security cameras have proliferated.

For example, after configuration and some experimentation on the user's part, systems can use face recognition to determine that a child has returned from school, triggering a notification that is sent to a parent's smartphone. BuddyGuard, Nest, and Netatmo sell cameras with bundled cloud-based face-recognition services. Nest's software, FaceNet, is the same as the software that recognizes faces within Google Photos. Netatmo's website has a testimonial from a user who reported that a burglar was caught by police, whom the user called after receiving a notification saying "unknown face seen." The user was away from home when the indoor camera captured the image.

Other cloud-connected camera offerings do not recognize specific faces but do identify categories of objects, including people, vehicles, and animals, enabling users to configure systems to send a notification if, for example, a vehicle enters a driveway. Netgear's Arlo Pro battery-operated camera comes with a free video-storage service, and the company is gradually introducing a premium cloud-based offering that performs category-based recognition.

However, the most sophisticated offering for home users—despite the lack of face recognition—seems to be from start-up Camio, whose cloud software categorizes images from many different brands of cameras. With ongoing use, the software learns which events are important enough to generate alerts. Camio also lets users search for stored videos using basic natural-language processing; a user can search for images of a "package on the porch" or "deer in front yard."

Implications

Suppliers are likely to improve home security by crafting systems that learn from the ways users configure camera apps. Early adopters will therefore help improve offerings for later adopters. This is especially true for Arlo Smart, Netgear's premium service, which is currently available only to "a limited number of randomly selected Arlo users" who are helping to train the system. Camio takes the extra step of having users rate notifications as either helpful or unhelpful, with some product reviewers expressing satisfaction with the resulting rapid improvement in the relevance of such alerts.

Impacts/Disruptions

Improvements in machine learning are likely to further aid ease of use. Laboratories and university researchers have software that can describe activities in images—such as "man pushing car" or "dog eating ice cream"—not just objects in images. Security cameras seem to be likely enablers for commercializing such software.

Camio uses machine learning for additional purposes—notably, to make economical decisions about when to expend resources to recognize images. Camio also hopes to gradually reassign artificial-intelligence duties to users' hardware, thereby reducing Camio's costs for video-data traffic and performing cloud computing. Other organizations could benefit from similar uses of machine learning for both user-facing features and internal functions.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: Medium

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: Now to 10 Years

Opportunities in the following industry areas:

Electronics manufacturing, professional security monitoring services, web services, cloud computing, embedded systems, semiconductors, advanced research

Relevant to the following Explorer Technology Areas: