How to Make AI Faster and Smarter—With a Little Help from Physics
1 min read
Summary
Rose Yu, now an associate professor at the University of California, San Diego is a pioneer in the field of physics-guided deep learning, having spent years trying to improve AI’s application for the field of physics.
Yu came to the field after getting traffic congestion while commuting near the University of Southern California, where she was a student, and wondered if there was anything that could be done about it.
This led her to develop new traffic forecasting models which were more reliable andvalid for one hour compared to preceding forecasts which were only reliable for 15 minutes and were deployed by Google Maps in 2018.
Building on this success, Yu and her colleagues looked to apply physics-guided deep learning to predicting the evolution of turbulent flow, a major factor in climate models. This was also successful, speeding up predictions by a factor of 20 in 2D settings and 1,000 in 3D settings.
Yu’s ultimate goal is to create a suite of digital lab assistants which she dubs AI Scientists which can aid human researchers in the discovery process.
