A University of Oregon computer scientist working to make artificial intelligence even more useful by improving the way networks handle the large volumes of data needed by machine learning.
Through a new three-year National Science Foundation grant of $599,920, Lei Jiao, an assistant professor in the Department of Computer Science in the College of Arts and Sciences, is pursuing research on a new AI paradigm called “Edge AI” that aims to improve the common applications of existing AI as well as benefit network operators, service providers and other applications such as e-commerce. He is collaborating with Purdue University professor Xiaojun Lin.
“In contrast to today’s practices of hosting AI services in remote data centers or servers edge AI is a new paradigm that moves AI to the network edge,” he said.
Imagine that a remote server is like a supermarket. Customers will travel out of their way to gain access to the best selection and prices. Compare that with a corner store. It’s easier to shop at a neighborhood market, but you might pay a higher price for that convenience, and the selection isn’t as good.
With edge AI, Jiao wants to integrate microservers at easy-to-reach points in the wireless network to make the corner store just as good as the supermarket by equipping it to handle a variety of data just as efficiently.
The current challenge with existing AI is that it generates large streams of data that need to be sent to edge AI for accurate and fast responses. Unfortunately, edge AI is not suited for that. In order to be fast and accessible, it’s limited in the volume and kinds of data it can handle.
“It cannot handle data variations,” Jiao said. “This is like inventory changes in a store. As a customer’s shopping preference or habit changes, the AI that recommended goods for yesterday may not work well as needed for tomorrow.”
Jiao’s research project directly confronts edge limitations by designing and implementing new algorithms to control and guarantee the performance of the edge AI. These algorithms can continually adapt edge AI to handle varying streams of data accurately and in a timely manner, despite their limitations.
Jiao originally started as a college student majoring in mechanical engineering. However, he took some computer science courses and gradually started to become interested.
“I changed my major and have been pursuing my education and research career in the computer science area ever since,” Jiao said, “including my Ph.D education in Germany.”
Jiao said the importance of AI in the future digital economy and the effect of AI on people's daily lives inspired him to do this research.
However, he wishes people could become more cognizant of the importance of mathematics and theories for AI and not understand computer science as “coding” and “software development.” While not commonly thought of, theoretical work is important because it ultimately optimizes the performance of computer systems for the future. The work creates a new foundation and paves the way for new complex systems.
—By McKenzie Days, College of Arts and Sciences