Researchers at the University of Washington and partners at Carnegie Mellon University are collaborating with the U.S. Department of Energy's SLAC National Accelerator Laboratory on a new project to help scientists use artificial intelligence to better understand the universe.
Modern astronomy is producing more data than ever before. Powerful telescopes - like the Simonyi Survey Telescope at the DOE-NSF Vera C. Rubin Observatory - and other instruments around the world are collecting detailed information about billions of stars, galaxies and other cosmic objects. These observations help researchers investigate some of the biggest mysteries in science, including the nature of dark matter and dark energy, how the universe evolved over time and what it is made of.
But there is a challenge: Much of the data from those projects is stored in different formats, housed at different institutions and difficult to combine. As a result, scientists often spend significant time preparing data before they can begin analyzing it.
"The scientific opportunities and the data analysis challenges are incredible," said Rachel Mandelbaum, head of physics at Carnegie Mellon University.
The new effort, led by Mandelbaum and funded by the U.S. Department of Energy's Genesis Mission, will create a shared data service to allow researchers to seamlessly access and combine information from multiple astronomy experiments. Rather than moving massive datasets from one location to another, the system will allow the data to remain where it is stored while making it available through a central platform.
"A new generation of telescopes and surveys will each change the way we understand our universe," said co-investigator Andrew Connolly, a UW professor of astronomy and director of the eScience Institute. "But it is when we bring these data together to look at the universe from a unified perspective that these discoveries will be truly transformative."
The data infrastructure developed as part of this project will be available to the astronomical community at the SLAC-hosted Rubin Observatory's U.S. Data Facility and via the American Science Cloud, which integrates the nation's most advanced high-performance computing systems, scientific facilities, data resources and production capabilities into a single, coordinated AI-driven system. The project extends the UW's investments in Rubin Observatory and the UW Institute for Data Intensive Research in Astrophysics and Cosmology, which are supported by Charles and Lisa Simonyi and led by James Davenport and Mario Jurić.
The infrastructure also will support a growing area of AI known as foundation models. These AI systems are trained on large and diverse datasets, enabling them to recognize patterns and connections that might otherwise go unnoticed.
In astronomy, foundation models could help researchers analyze many different types of observations at once, including images, measurements of light from distant objects and records of how those objects change over time. By bringing these data sources together, scientists hope to uncover new insights about the universe more quickly and efficiently.
Ultimately, the team hopes to transform the vast collections of astronomical data being gathered today into a long-lasting scientific resource, helping researchers answer some of humanity's most fundamental questions about the origin, evolution and makeup of the universe.
Other UW co-investigators include Neven Caplar, a research scientist and engineer in astronomy. Other co-investigators include Jeremy Kubica, director of engineering for the LINCC project and Adam Bolton, senior staff scientist at SLAC and at Stanford's/SLAC's Kavli Institute for Particle Astrophysics and Cosmology.
Additional modeling will be provided by Francois Lanusse of the French National Centre for Scientific Research. Their work is an extension of the Schmidt Sciences-supported LINCC Frameworks program, a partnership led jointly by CMU and the UW.
This story was adapted from a press release by Carnegie Mellon University.
