CSE Colloquium: The scientific journey with deep learning and large language models

Presenter: J. Xavier Prochaska, UCSC Astronomy and Astrophysics

Abstract:
I will review the rise and roles of deep learning models in the scientific domains of astronomy and oceanography. Initially met with healthy skepticism, deep learning methods now pervade each field to enhance and automate classification, improve data processing, and enable new inferences from complex datasets. I will introduce the major datasets in each field (e.g. Pb of satellite imagery and numerical simulation outputs) and then discuss the scientific frontiers advanced with deep learning techniques. Conversely, I will also emphasize the limitations: discoveries in these data-driven fields continue to be achieved with new observations and new observing systems. As time permits, I will discuss the widespread impacts of large language models on the scientific pursuit and the risks that I perceive for the next generation of scientists.

Bio:
J. Xavier Prochaska is a Distinguished Professor at the University of California, Santa Cruz from the Departments of Astronomy and Astrophysics and Ocean Sciences. After a career focused on gas in, around and in between galaxies, he pivoted to oceanography. His research interests include applying artificial intelligence to remote sensing datasets, exploring the Earth’s biomass with hyperspectral ocean color observations, assessing dissolved oxygen and the transfer of energy at the submesoscale with glider data, and studying the enigmatic signals known as fast radio bursts.

Hosted by: Professor Daniel Fremont and Professor Leilani Gilpin

Zoom: https://ucsc.zoom.us/j/95300285113?pwd=62tnEPUtRrsmbW7SHTavaRbaF2afSr.1