Industry executives and experts share their predictions for 2022. Read them in this 14th annual VMblog.com series exclusive.
Rapid Adoption of Hyperautomation at the Edge
By Bill Conrades, MBX Systems
Until recently, hyperautomation was used almost exclusively
in enterprise environments to turbocharge business process automation and
thereby streamline operations. The complexity of combining technologies such as
artificial intelligence (AI), machine learning (ML) and robotic process
automation (RPA) initially limited use of this "Automation 2.0" capability to
large organizations, but that promises to change in 2022.
As workplace shortages and operational costs continue to
increase - and as hyperautomation enters its third year on Gartner's list of
top technology trends - the intelligent digital workforce will move beyond the
data center to enable new efficiencies at the edge. Inferencing applications in
this category will span intelligent patient safety monitoring for hospitals,
advanced analytics in video surveillance, automated food preparation and
transferring, autonomous forklifts in factories, and more. Hardware appliances
will be purpose-built for operating machinery, and more applications will rely
on platforms designed to be mobile.
Four key factors have converged to pave the way for this
transition.
1 - Lower-cost AI
acceleration: Improved AI/ML price-to-performance is making edge-based
inferencing more affordable. Core components are getting smaller, cheaper and faster. Solutions
like NVIDIA Jetson, Google Coral, and Hailo are leading the charge with
high-efficiency, small form-factor embedded computing boards and acceleration
modules designed to run at the edge.
And the improvements keep
coming. Much more powerful AI accelerator cards as well as SoC AI solutions are
coming to market. As a result, solutions like NVIDIA's Jetson will soon have
enough power to replace some GPU-based edge solutions. This will slash the cost
of ownership for an inference machine from a range of $2,500 to under $1,000.
Advances like these are removing barriers to development of new products that
would have been impractical to bring to market even 12 months ago.
2 - Onboard AI
processing abilities: Thanks to these upgrades in chip technology, quality
onboard AI acceleration is being designed into the CPU. As an example, Intel's
new Alder Lake platform provides AI acceleration through DL Boost, a
combination of VNNI instructions on the CPU and GPU acceleration for AI
inferencing of high resolution image workloads with the OpenVINO toolkit. That
means that inferencing one or two models at the edge is now possible without an
additional accelerator card, keeping costs low and eliminating latency and
bandwidth constraints because the cloud is not needed for inferencing. This
will enable edge-based AI deployments in any environment and even at scale.
In a hospital setting, for example, every room can now be
furnished with a mobile cart outfitted with a camera, embedded workstation and
AI software that can detect activity such as an imminent fall without streaming
data to the hospital network. Visitors' body temperature can be screened at
hospital entrances without physical contact. Virtual patient interactions can
be conducted with speech recognition. Surgeons can control and navigate medical
images with speech and hand gestures. These kinds of applications can help
combat labor shortages in healthcare as well as other industries, and onboard
AI acceleration improvements makes them even more affordable.
3 - Fast-tracked
machine learning: Equally important in fueling the adoption of edge-based
hyperautomation is the emergence of machine learning frameworks like NVIDIA
Clara Guardian, Google TensorFlow and Intel OpenVINO that can dramatically
reduce development time. These building blocks make it faster and easier to
implement machine learning by providing libraries of tools such as pre-built
training models that can eliminate months of work.
Again, consider smart hospital applications designed for
deployment at the edge. NVIDIA Clara Guardian includes pre-trained machine
learning models in areas such as body pose, gesture recognition, heart rate
estimation, mask detection, and speech recognition for common patient requests
in a healthcare setting, as well as tools that make it possible to securely
manage AI deployments across multiple servers and edge devices. These resources
slash AI development time by a factor of 10, making the effort more financially
attractive to ISVs.
4 - Hardware building
blocks: Complementing these software tools are new mobile device and
reference platform options that further shrink both cost and time to market by
eliminating the need for custom hardware design that can require months of
prototyping and cost tens of thousands of dollars.
MBX Systems, for example, offers an exclusive mobile cart
platform enabling a single housing to be quickly customized to meet different
ISV needs, using workstations that are pre-certified for medical and global
use, and consolidating all development and assembly services under one roof. The company also offers a portfolio of
deployment-ready hardware reference platforms for AI applications being
delivered as embedded systems, edge devices or edge servers.
Taken together, these factors will help turn 2022 into a
banner year for edge-based hyperautomation solutions. As a result,
hyperautomation will become more than an efficiency enabler for enterprise IT
operations and instead provide a whole new range of opportunities for bringing
cutting-edge AI into the mainstream.
##
ABOUT THE AUTHOR
Bill Conrades is Sales
Engineering Manager at MBX Systems (www.mbx.com), a provider of custom computing hardware
engineering, manufacturing and support services for ISVs, OEMs and other
technology companies that deliver complex products on turnkey hardware.