Page 39 - EE Times Europe Magazine | April2019

EE|Times EUROPE — Boards & Solutions Insert 39

Let’s Talk Edge Intelligence

types of connected devices. While these can independently perform don’t tend to use the word ‘endpoint’ in our
machine-learning functions, they might also feed into a gateway where own communications, perhaps definitionally,
more advanced ML processing occurs.” ‘endpoint’ is aligned with the edge devices,”
Chris Bergey, general manager and vice president of the infrastruc- said Adesto CTO Gideon Intrater. “AI in these
ture business at Arm, had a somewhat different perspective, citing the devices would typically be some amount
increasing levels of intelligence in both edge servers and endpoints. of local inference, with the algorithms
“Basic devices such as network bridges and switches have given way running as a program on a processor, using a
to powerful edge servers that add data-center–level hardware into the dedicated accelerator, through near-memory
gateway between endpoint and cloud,” he said. “Those powerful new processing or in-memory computing.”
edge servers making their way into 5G base Gideon Intrater, He added that edge AI “is becoming a
stations are plenty powerful enough to per- Adesto reality across just about every application.
form sophisticated AI processing — not only We see a great opportunity in industrial
ML inference but training, too.” and building implementations where AI can provide benefits through
How is that different from endpoint AI? predictive and preventive maintenance, quality control in manufac-
Bergey explained by offering an example: turing, and many other areas. The industry is just getting started, and
“Due to their powerful internal hardware, every day that passes, we expect AI to do more for us. When our older
smartphones have long been a fertile test- devices without AI don’t intuitively understand our needs, we often
bed for endpoint AI. As the IoT intersects get frustrated because we have other devices that will provide intuitive
Chris Bergey, with AI advancements and the rollout of capability. The end consumer doesn’t know what goes into making an
Arm 5G, more on-device intelligence means AI solution work; they just expect it to work.”
that smaller, cost-sensitive devices can be
smarter and more capable while benefiting from greater privacy and IT’S STILL EARLY DAYS
reliability due to less reliance on the cloud or internet. So we are clear on the definition: You either sit in the camp that says
“As this evolution of bringing more intelligence to endpoints con- the edge is everything that’s not in the cloud or with those who clearly
tinues, the boundaries of where exactly the intelligence takes place identify the endpoint as the meeting point of the physical world with the
will also begin to blend from endpoint to edge, stressing the need for a digital world, mostly the sensors. But the specific application will deter-
heterogeneous compute infrastructure.” mine the point at which the intelligence might need to be added, with
There are others for whom edge is everything that’s not in the cloud. an increasingly blurred line between edge and endpoint and a somewhat
For example, Jeff Bier, founder of the Edge AI and Vision Alliance, said heterogeneous compute infrastructure.
that the group defines edge AI “as any AI that is implemented — in The next questions are: Who would want
whole or in part — outside the data center. The intelligence might be it, and what are the market expectations for
right next to the sensor — for example, in a smart camera — or a bit edge AI? “This is something we’re all still
farther away, such as an equipment closet in a grocery store, or even figuring out,” said NXP’s Levy. “The industry
farther away, such as in a cellular base station. Or [it might be in] some leaders are well engaged in implement-
combination or variation of these.” ing it; I can’t name names, but we have a
Xilinx takes a similar position. “Edge AI is basically a self-sufficient wide range of customers doing all kinds of
intelligence deployed in the field without reliance on a data center,” said machine learning at the edge. However, if you
Nick Ni, the company’s director of product marketing for AI, software Markus Levy, look at the technology adoption cycle, I still
and ecosystem. “It is essential for applications that require real-time NXP believe the majority of the industry is not
response, security — for example, not sending confidential data to the even at the ‘early adopter’ stage, and this will
data center — and low power consumption, which is most of the devices really begin to unfold toward middle to late 2020.
out there. Just as humans don’t rely on a data center to make countless “Customers are still comprehending the cool things that are possible
decisions daily, edge AI will dominate the market in applications like with machine learning. But I typically give a few guidelines. [First,] can it
semi-autonomous cars and smart-retail systems in coming years.” save money — for example, by making a factory assembly line run faster
Andrew Grant, senior director for artificial or more efficiently by replacing headcount that was previously doing
intelligence at Imagination Technologies, visual inspection? [Second,] can it make money — for example, by adding
affirmed that idea. “It’s all edge as far as a cool feature to a product that makes it more useful? Maybe this is a
we are concerned; it’s the customer who barcode scanner that, by using machine learning, can remove wrinkles in
decides where it goes,” he said. “We’ll see a package that were previously making it impossible to scan accurately.”
very much a hybrid approach, and there’s Infineon’s Furtner asked essentially the same question in a different
absolutely a role for the cloud and data way: “What is the benefit of edge AI?”
centers in this, too.” He added, “The great thing about the edge is that we can turn its
Grant added that “the speed with which ‘weaknesses’ with respect to constraints into strengths. People do
Nick Ni, the market is moving [to the edge] is phe- care about things like ease of use, functionality, privacy, security,
Xilinx nomenal. There’s been a wave of movement cost, climate, or sustainable use of resources. These are all benefits
to the edge, but for many applications, it we can make possible with edge AI. We are convinced that AI at the
takes time for the silicon to materialize. We were talking to a traffic right places enhances our life and that there are many use cases for AI
management company in China; they are moving data back and forth in endpoints. Edge AI is used for predictive maintenance and further
from the cloud. When I explained to them what we do, they immedi- automation or robotics, home automation, or smart farming, to name
ately saw the benefit of not having to take the data to the cloud if the a few applications. With our work on low-power AI-enabled sensors,
traffic lights themselves can determine whether a car is moving or not.” we make intuitive sensing more ubiquitous, spurring new applications
Embedded systems provider Adesto Technologies doesn’t necessarily in the home or city that can make lives easier, safer, and greener. Being
differentiate between edge and endpoint, given that the company pro- non-dependent on the cloud enables fully new usage models in indus-
vides devices for IoT edge servers as well as IoT edge devices. “While we try or home applications that cater for privacy and security.”

www.eetimes.eu | APRIL 2020