Short Stories of AI in, and for, Astronomy
Presenter: Dr. John Wu, Associate Astronomer and Applied AI Scientist, the Space Telescope Science Institute
Abstract:
(How) Do Astronomers Trust AI for Research?
Navigating the Cosmic Library
Whose Telescope Is It Anyway?
Hidden Mechanisms of Language Models in Astronomy Literature
Hidden Mechanisms of Vision Models in Galaxy Images
The Platonic Universe
Bio: Dr. John Wu is an Associate Astronomer and the Applied AI Scientist at the Space Telescope Science Institute, and he holds visiting affiliations at the Johns Hopkins University Department of Physics & Astronomy and the Department of Computer Science. His research focuses on studying galaxies with multi-wavelength observations, machine learning, and interpretable artificial intelligence (AI) for scientific discovery. He was previously a postdoc at Johns Hopkins University, received his PhD in Physics from Rutgers University, and received a BSc in Physics from Carnegie Mellon University.
Website | GitHub | Bluesky | Google Scholar
Cosmic Co-Discovery with AI
Presenter: Dr. Ioana Ciuca, Interdisciplinary Researcher, Department of Astrophysics and Computer Science, Stanford, and co-founder, UniverseTBD.
Abstract: What does it take for AI to help us with our science? We explore this question through two lenses. First, I introduce ReplicationBench, a benchmark developed at the Center for Decoding the Universe at Stanford that tests whether AI agents can replicate peer-reviewed astrophysics papers end-to-end. Benchmarking frontier models on author-written tasks of varying difficulty, we find that even the best agents succeed less than 20% of the time, revealing systematic failure modes in how AI reasons about science. Second, through UniverseTBD, our interdisciplinary community working to democratize science through AI, I present our latest results on astronomical foundation models, showing that larger models converge toward shared representations regardless of architecture, suggesting the community could build on existing models rather than train from scratch. I close with an open invitation to help shape how AI transforms research.
Bio: Dr. Ioana Ciuca is an interdisciplinary researcher in Astrophysics and Computer Science at Stanford, coordinating the Frontier AI projects of the Center for Decoding the Universe to advance scientific discovery in Astrophysics. Together with Kartheik Iyer, she co-founded UniverseTBD, a global movement of explorers working to democratize science and help us understand the Universe.