Bringing computing to the “cutting edge” of the edge may include actually incorporating sensors into the components used to build structures and machinery to bring true autonomy to the objects and equipment that will drive digital transformation (DX). Experts assembled for TDK Ventures’ DX panel on edge computing explored the possibilities over the next 10 years of mixing a bit “smart dust” into systems to monitor, control, and adjust the performance of these technologies.
Pete Warden, technical lead of Google’s TensorFlow team, noted that “compute keeps getting smaller and smaller and more and more energy-efficient.” Taking those trends to a logical conclusion over the next decade, we could see devices that are smaller than a grain of rice, with enough battery power to run things like machine learning computing. The potential use cases are compelling, Warden explained:
“From an engineering point of view, it seems pretty clear that’s the direction we’re heading,” he said.
Mike Vildibill, vice president of Cloud Edge AI at Qualcomm, said whether smart devices can be miniaturized to the size of “marbles, breadcrumbs, or dust, the point is, there’s going to be lots of them.”
Seizing on Warden’s concrete-sensor example, Vildibill said engineers should design monitors not for earthquakes and other rare events but rather specify the performance, data storage, connectivity, and bandwidth for the standard operational environment.
“I believe the future holds more nomadic computing using not only the main device but also devices nearby and powered by advances in connectivity,” he said. “Therefore, we need our devices to share and use resources contained in the smart devices around them.”
Like the electrical grid, he said, these devices will be able to borrow computing power and other resources from underutilized devices nearby to ensure efficient and constant operation.
Elana Lian, Intel Capital’s investment director, said smart dust could prove a worthy target for venture capital “as compute gets more power-efficient, lower-cost, and ubiquitous. When I look at it as a driver of my investment commitment, I want to see a different level of automation as we get to the smart dust level. As we close in on smart dust applications, we’re going to see Level 4 and Level 5 automation. Level 4 will bring full autonomy within a defined area or specific solution. Level 5 — complete autonomy in all situations. That’s linking our minds to use our neuroscience to control [our physical environment].”
FCC Technological Advisory Council Chairman Dean Brenner promoted the network effect of bringing connectivity and resource-sharing to the edge. The more cars, homes and neighborhoods that connect with each other, the more efficient resource allocation can be, he said. And the more smart devices can communicate with each other, the greater the economic and security benefits.
Chris Bergey, senior vice president and general manager of infrastructure at Arm, noted that the Helium Project and other initiatives are experimenting with “sharing and crowdsourcing connectivity over the blockchain and using cryptocurrency to compensate bandwidth contributors. He said full autonomy presents one of edge computing’s next big challenges. He noted that building prototypes in optimal, controlled conditions is easy compared to applying them to use cases.
“As a Tesla owner who paid for the self-driving upgrade, I can tell you that it’s nice, it’s convenient, but it’s very far from self-driving, even in the best of conditions,” he said.
Bergey compared edge autonomy’s current position to that of videoconferencing in the late 1990s.
“We tried to demonstrate that we could do it using 56k modems. Of course, we couldn’t really, except in optimal situations, not the way we use it today when we have gigabits of bandwidth to our homes and a totally different network build-out.”
Similarly, he said, edge computing has disappointed because the expense and compute limitations has limited its rollout to particular verticals employing massive bandwidth. With smart dust bringing computing power across the board, the rising technology tide will lift all application boats.
“But the promised autonomy is still very far away. There are controlled environments — lights-out factories, for example — but we’re nowhere near people being comfortable with pilotless airplanes,” he said. “There’s so much energy, money, intelligence being put into it that those things are going to happen, but prototype to reality is probably something like 20 years. The metaverse and virtual worlds drive a different level of computing and have different latency requirements. Those things drive edge computing, but you need to have it under a high percentage of use cases to be able to depend on it, otherwise you have to put all that compute into the device itself.”
Session viewers were intrigued by the idea of smart edge devices operating outside a network. Panelists discussed several scenarios where this concept would prove beneficial. Vildibill, for instance explored autonomous vehicles’ need for “disconnected intelligence.” He said that if a vehicle loses its connection to a communications or crash-avoidance network, it still must be able to operate, either autonomously or through driver intervention.
Other edge computing use cases such as motion sensors and smart switches do not need connectivity at all, he said.
“In the future, when you buy a premium version of your product, you might be paying more for additional features — like autonomous driving on a Tesla — but you might also be paying more for richer connectivity,” he said. “Richer connectivity might not just be latency and bandwidth, but rather a bunch of services that are integrated. So, we might see connectivity, in and of itself, being offered as an edge product or service that has different tiers of quality and value.”
Likewise, Vinay Ravuri, EdgeQ founder and CEO, said his company is developing edge devices that require no power onboard source. They need little energy and can harvest it from ambient light or electromagnetic waves to power the lightweight computing they require to function. He said these devices can be used currently for small light sources or low-capacity transmitters. In the future, they may have a place in human-machine interfaces where onboard power sources would prove cumbersome and external sources could limit mobility.
Making processes more responsive with less latency and greater performance by bringing computing power to the edge will go a long way toward the world realizing widespread digital transformation. Monitoring, sensing, and making decisions based on events and circumstances at the source are tools to help make this happen. TDK Ventures thanks its DX Week Edge Computing experts for sharing their insights into the issues confronting researchers, entrepreneurs, and investors operating in this space. Like all TDK Ventures programming, DX week sought to explore the visionary work being done on computing’s cutting edge. The discussion DX Week provoked will inspire and inform the next generation of innovators and impact scalers who will lead the charge to accomplish the goal of global digital transformation.