With edge computing becoming the next big thing, AI on-the-edge is quickly following suit. It unlocks a whole new world of possibilities, including predicting customer needs before they even know them.
To tap into this opportunity, organizations don’t need to choose a risky “all in” approach; a small iterative approach reduces the risk while ensuring edge AI projects aligns the overall business strategy.
Wolf Ruzicka, chairman, EastBanc Technologies and Polina Reshetova, data scientist, PhD, EastBanc Technologies, discussed this concep during their Data Summit 2019 presentation, “AI On the Edge.”
The Edge consists of devices that are small in size but still have computational power. They are located away from the main brain of the operation, Reshetova explained.
“The Internet of Things and these edge devices interact,” Reshetova said.
AI on the edge can minimize delay, improve privacy, conserve bandwidth in the IoT system, and deliver personalization/self-improvement.
An iPhone X FaceID is an example of AI on-the-edge. A data model made by Apple consisting of someone’s pictures can identify the owner of the phone.
“It’s the compilation of your pictures that runs the model,” Reshetova said.
Apple’s Siri is an application that operates on the edge. The platform runs on a Deep Neural Network (DNN), five hidden layers, with 32, 128, or 192 units. When someone activates the device by saying, “Hey Siri” the application taps into the always on processor which connects to the main processor.
AI on the edge powers security and Surveillance technology, she explained. In the future an officer may be able to identify someone by their pictures online.
“So the demand for this is there,” Reshetova said.
Before starting any project that utilizes AI on-the-edge, the smaller the project, the better it will be able to handle, Ruzicka said. This is important because smaller models can deliver more focused results and data.
This presentation is available for review at http://www.dbta.com/DataSummit/2019/Presentations.aspx.