Biography
Tuan Do is an Associate Professor at UCLA in the Physics and Astronomy Department. He is the PI of the UCLA Astrophysics Data Lab. He got his undergraduate degrees in Physics and Astrophysics from UC Berkeley and his Astrophysics Ph.D. at UCLA. His research focuses on translating ML/AI models for Astrophysics applications. He created the first Machine Learning for Physical Science course at UCLA.
SRP Project Title
Integrating LLMs into Machine Learning for Physics and Astronomy Education
NAIRR Project
Improving Access to Computation for Machine Learning for Physical Sciences Course
Topical Areas
Artificial Intelligence and Intelligent Systems; Astronomy and Planetary Sciences; Particle and High-Energy Physics
Abstract
This project is to study how to use large language models (LLMs) and related technologies into a course on Machine Learning for Physical Sciences at the upper division undergraduate level. LLMs are now a prominent technology in computer science and the industry, and many students have experience using them. However, students do not typically get the opportunity to setup their own LLM and do experiments on them. Specifically, the use of LLMs in scientific education is not well explored. This project aims to develop a course unit on both using existing LLMs as well as the underlying LLM architectures (e.g. transformers) for science. Potential projects include setting up and deploying local LLMs and examining their abilities to do physics research. Another is to combine images and text from astrophysics into a transformer to use data fusion for classification. By giving students experience in underlying technologies behind these tools, they will become more knowledgeable and responsible users of AI and develop skills useful for their careers.
Desired Skills
Background in teaching, designing lab courses, and/or deploying large language models.
Lightning Talk Title
Developing an LLM curriculum for AI/ML in Physical Sciences courses
Keywords
LLMs; transformers; science education; Physical Sciences; Physics; Astrophysics; Data Science; Lab Classes