AI Where it is Needed

A reasonable explanation ..... though not enough about security concerns,  that need to be better addressed when we move most AI to the Edge.

AI on Edge    By Samuel Greengard
Communications of the ACM, September 2020, Vol. 63 No. 9, Pages 18-20  10.1145/3409977

A remarkable thing about artificial intelligence (AI) is how rapidly and dramatically it has crept into the mainstream of society. Automobiles, robots, smartphones, televisions, smart speakers, wearables, buildings, and industrial systems have all gained features and capabilities that would have once seemed futuristic. Today, they can see, they can listen, and they can sense. They can make decisions that approximate—and sometimes exceed—human thought, behavior, and actions.

Yet, for all the remarkable advancements, there's a pesky reality: smart devices could still be a whole lot more intelligent—and tackle far more difficult tasks. What's more, as the Internet of Things (IoT) takes shape, the need for low latency and ultra-low energy sensors with on-board processing is vital. Without this framework, "Systems must depend on distant clouds and data centers to process data. The full value of AI cannot be realized," says Mahadev Satyanarayanan, Carnegie Group Professor of Computer Science at Carnegie Mellon University.

Edge AI takes direct aim at these issues. "To truly and pervasively engage AI in the processes within our lives, there's a need to push AI computation away from the data center and toward the edge," says Naveen Verma, a professor of electrical engineering at Princeton University. This approach reduces latency by minimizing—and sometimes complexly bypassing—the need for a distant datacenter. In many cases, computation takes place on the device itself. "Edge AI will enable new types of systems that can operate all around us at the beat of life and with data that is intimate and important to us," Verma explains.

The power of this framework lies in processing data exactly when and where it is needed. "Edge AI introduces new computational layers between the cloud and the user devices. It distributes application computations between these layers," says Lauri Lovén, a doctoral researcher and data scientist at the University of Oulu in Finland.

Pushing intelligence to the edge could also fundamentally alter data privacy. Specialized chips and cloudlets—essentially micro-clouds or ad hoc clouds that would function in a home, business, or vehicle—could control what information is sent from smart devices, such as TVs and digital speakers.   .... "      (Much More)