AI & Astronomy Imagery

About Us

Katie Frey, the Data Curator for Programmatic Access at Stanford University Libraries' Research Data Services, develops best practices for API usage to support on-campus researchers. Before joining Stanford, Katie served as the Assistant Head & Digital Technology Development Librarian at the Center for Astrophysics | Harvard & Smithsonian.

Jooyeon Hahm, the Head of Data Science Training and Consultation, supports quantitative, computational, and algorithmic analysis of research data, including data management, analysis methods, workflow reproducibility, and ethical considerations.

This project emerged from Katie's passion for astronomy and Jooyeon's cautious interest in AI, sparked by their friendly office conversations. They aimed to explore how AI interprets astronomy images.

While astronomy often captivates many, astrophotography, which captures complex celestial phenomena, remains a specialized field requiring specific equipment, methods, and skills. Existing generative models excel at mimicking dreamy, cosmic feelings but lack understanding of specific celestial objects and phenomena (e.g., Rigel, Polaris, galaxies, and nebulae). AI-generated descriptions of astrophotographs tend to emphasize aesthetic qualities, using adjectives like "beautiful," "breathtaking," and "dreamlike" rather than providing accurate representations of extraterrestrial and extrasolar features. This project aims to show the strengths and weaknesses of generative AI, and that it will often create content to fill in the gaps of its training set making it unable to represent scientific concepts.

We encourage everyone interested in the intersection of AI and science to explore this online exhibit and the included peer-reviewed articles under Further Reading.  If you would like to learn more about the local image generation models used in this exhibit and available in Green Library, please contact Katie or Jooyeon!