The Edge Use Case You’ve Been Waiting For

Lori MacVittie 축소판
Lori MacVittie
Published June 02, 2021

Success on the Internet has always been positively correlated to cat videos, and edge is no exception. 

Lest I be accused of committing a post hoc fallacy, let me say I am only partially serious about cat videos being a significant indicator of edge success. For those not familiar with this logical fallacy, you might recognize it from an infamous chart proving the "decline of pirates in the world are causing global warming." (Data Science Central) This fallacy is in part responsible for the catchier phrase, "correlation does not prove causation."

Nevertheless, I will point out that a quick Internet search will find a robust index of articles, blogs, and videos promoting the use of cat videos and memes for marketing, social media, and general customer engagement.

Perhaps cat videos are the exception to the rule.

Regardless, I am definitely about to use cat videos to illustrate a use case for edge. 

Find My Cat

First, consider the high likelihood that the neighborhood in which you live has a high number of video doorbells installed. Estimates are that as of 2020, 16% of all US households were using one.

Amazon's Ring accounts for 40% of doorbells in use, followed by Google's Nest with 24%. All other brands each account for less than 10% of doorbells in use, including Vivint, Remo, August, SkyBell and SimplySafe. Strategy Analytics estimates that more than 20 million US homes now use a video doorbell. (Business Wire, Feb 13 2020)

Let’s say your cat escaped. Because cats are like that. You can’t find them and tasty treats are not working their magic today. Imagine that you could harness the power of all the video doorbells in your neighborhood to find your cat.

The ability to identify objects—including cats—is a common use case for machine learning. Research indicates a high success rate and its use has wide applications in the fields of ecology, wildlife biology, zoology, conservation biology, and animal behavior.

Why not tracking down your lost cat?

If the data—the video—from those cameras were automatically and inexpensively aggregated, it would be a fairly simple exercise to search them and identify a cat. Moreover, with the ability to extract location from such video, one could rather quickly be informed as to where your cat might be.

The challenges are not completely technical at this point. They are business and even societal (privacy) challenges.

Despite these challenges, the edge is ripe for such solutions given the changes in technology over the past several years. Growing maturity of machine learning and AI in addition to more powerful compute at the edge provide an opportunity to incorporate endpoints and edge compute in new ways.

Endpoints are Part of the Solution

This concept—that endpoints are part of the solution space—is part of the premise on which we base our belief  that edge computing must be redefined.  

edge 2.0 architecture

Much like cloud computing was initially born from the realization that idle compute could be utilized and turned into a service, so too can the idle compute and processing power of endpoints and nodes at the edge be utilized for new and interesting purposes. Like finding your cat.

This idle compute is increasing, thanks to the infrastructure renaissance that is delivering optimized compute in form factors capable of integrating with the smallest of devices. The exciting advances in GPUs and DPUs are a part of several of our internal innovation projects as we explore new ways to deliver, distribute, and optimize applications at the edge.

Endpoints—cameras, phones, doorbells—routinely contain more compute power than my first computer, my first gaming console, even my first "smart" phone. They are capable of being a part of the solution, rather than a passive player. What remains is simply the means to incorporate them into a platform that can properly respect privacy, ensure security, and operate as part of a larger pool of resources.

Now, will we actually see the development of a real-time, neighborhood cat finding service based on video doorbells? Perhaps not. But the concept is sound, given that we already use human power to review video from hundreds of public-facing security cameras. Machine learning and AI is harnessing the power of compute to automate and scale a human process—the review of video in the search for a stolen car, an abducted child, and yes, even a lost cat.

Imagine being able to conduct those searches in real-time.

That’s the kind of capability we see possible with the platform we call Edge 2.0.