When she was just 10 years old, Rose Yu received a life-changing birthday gift: a computer, an unusual luxury in China 25 years ago. Initially, she focused on games, but by middle school, her talent for web design emerged, earning her accolades that paved the way for her future in computer science.
Yu pursued her undergraduate studies at Zhejiang University, achieving recognition for her innovative research. Her academic journey continued at the University of Southern California (USC), where her work culminated in a doctoral degree in 2017, recognized with an award for the best dissertation. Notably, she was honored with a Presidential Early Career Award from President Joe Biden in January 2021.
As an associate professor at the University of California, San Diego (UCSD), Yu specializes in "physics-guided deep learning," integrating physics principles with artificial neural networks. This novel approach has allowed her to enhance various applications, including traffic prediction, climate modeling, and the spread of COVID-19.
Yu’s vision is ambitious: to create a suite of digital lab assistants she calls AI Scientist, facilitating collaboration between human researchers and AI. This concept leverages the laws of physics to generate new scientific insights, potentially transforming the research landscape.
Her journey began with traffic modeling during her graduate studies at USC, inspired by the notoriously congested roadways around campus. She conceptualized traffic flow through the lens of diffusion and graph theory, utilizing extensive data on Los Angeles traffic to develop models that could predict traffic conditions for up to an hour ahead—an advance over previous forecasts limited to just 15 minutes. Google Maps adopted her code in 2018.
Building on this success, Yu expanded her focus to climate modeling, particularly in understanding turbulence—a significant factor in climate prediction. Collaborating with scientists at the Lawrence Berkeley National Laboratory, Yu sought to emulate turbulent flow patterns using deep learning. This approach sped up computations significantly, providing more efficient models for predicting events like hurricanes.
Turbulence is a recurring theme in Yu’s work, affecting various fields, including health and drone stability. Her research has examined blood flow mechanics and developed models for stabilizing drone flight amid turbulent airflow. Currently, she’s working on deep learning algorithms to predict plasma behavior in fusion power generation, a critical area for the development of sustainable energy sources.
Yu’s idea of an AI Scientist arose from algorithms her team developed to automatically discover symmetry principles from data. These discoveries revealed fundamental concepts in physics and hinted at the potential for AI to generate novel research ideas, marking the ambition behind the AI Scientist initiative.
Ultimately, the AI Scientist isn’t merely a sophisticated neural network; it’s a collective of computer programs designed to aid scientific discovery. This foundation model aims to integrate various data types—numbers, text, images, and videos—supporting scientists across diverse domains. Yu envisions it assuming tasks like literature reviews, hypothesis generation, and data analysis, empowering researchers to focus on the more creative aspects of their work.
In Yu’s vision, the AI Scientist will not replace human ingenuity; rather, it will alleviate mundane tasks, allowing researchers to explore the creative landscape of science more freely. The goal is to create an innovative partnership between humans and AI—enhancing the scientific process without overshadowing the invaluable role of human creativity.
