PNNL

Artificial intelligence is all around us; it's in our selfies, it's in mapping apps that take us to the closest place for food, and it’s in the apps that help us translate other languages.

Data scientist Court Corley leads research to make artificial intelligence, almost always abbreviated as “AI,” work better for us and make sure AI stays safe and secure. Types of AI called machine learning, and especially deep learning, allow computers to learn in a way inspired by our brains, organizing and analyzing new information to make sense of it. AI can then base decisions on what it's learned, as it does when detecting breast cancer.

"AI is really about enabling machines to do things that humans can't, or sensitive things that humans can," Corley said. "Think about looking at a cell under a microscope that has been exposed to a toxin and it's abnormal. How do you tell a computer what that looks like? How can you build a computer model to be able to detect that in future images?"

Corley co-leads, with PNNL colleague Nathan Hodas, the development of software called Sharkzor that can build a computer model with smaller than usual amounts of data. Instead of AI doing all the work to understand, Sharkzor starts with a user who teaches it how to categorize images. Then Sharkzor tries a few itself and the user fixes its mistakes until Sharkzor gets the hang of it. This cycle of learning lets AI learn faster, requiring a lot fewer examples at the start.

But all this AI in our daily lives means we have to figure out ways to keep it safe, especially if it's running an autonomous vehicle or the electric grid. The key to security is protecting the data. “Design it right,” Corley said, “and AI itself will then learn the best ways to go about protecting applications.”

Corley teaches at Washington State University, Tri-Cities, and has organized many conference events for professional societies such as the Association of Computing Machinery, the Institute of Electrical and Electronics Engineers, and the International Security and Defence (sic) Systems meetings.

More Information

PNNL Staff Biography

LinkedIn profile

Twitter: @courtcorley

Listen to Court Corley talk about the "Golden Age" of artificial intelligence, AI-human teams, and how AI is like the early days of penicillin on Direct Current - An Energy.gov Podcast (22:39)