The universe, decoded
A new Stanford center is uniting astronomers and data scientists to decipher the cosmos—and reinvent the future of discovery.
)
Every night, the universe sends a message.
Thousands of them, in fact—flashes of light from distant stars, the quiet drift of asteroids, the flare of a dying sun
Telescopes like the Vera C. Rubin Observatory are capturing it all, creating a wave of cosmic data so vast, no human alone could ever make sense of it.
At Stanford, a new kind of team is rising to meet that challenge.
By bringing together astrophysicists, data scientists, researchers in AI and machine learning, and more, the newly launched Center for Decoding the Universe aims to transform this torrent of information into insight—and unravel some of the deepest mysteries of space and time.
This image was made from more than 1,100 images captured by the Vera C. Rubin Observatory. Photo: NSF–DOE Vera C. Rubin Observatory
Telescopes have always been instruments of awe and discovery. But we’ve entered a radically new era—one defined less by what we can see, and more by what we can process. Space observatories like the James Webb are already delivering dazzling views of the distant universe. But the next generation of instruments—like the NSF-DOE Vera C. Rubin Observatory in Chile and NASA’s upcoming Nancy Grace Roman Space Telescope—promise to shift astronomy into overdrive, generating petabytes of data and revealing cosmic phenomena in real time.
In June, the Rubin Observatory released its first images of the southern sky, marking the dawn of a new era in astronomy. Designed and constructed at SLAC National Accelerator Laboratory, Rubin’s LSST digital camera, with a 3.2 billion-pixel CCD, is the largest ever built. The observatory’s ability to continuously scan the sky and map the cosmos at this scale underscores why Stanford researchers are racing to prepare. The data firehose has already been turned on.
“The Rubin Observatory will essentially be making a very deep map and a movie of the entire southern sky,” says Risa Wechsler, the Humanities and Sciences Professor II and director of the Kavli Institute for Particle Astrophysics and Cosmology at Stanford. “The telescope is designed to survey wide areas of sky both quickly and deeply. Over 10 years, it will be making an ever-deeper map, looking at every single thing that changes. We expect at least a million things to change every night, from asteroids in the solar system to supernovae 13 billion light years away, as well as things we’ve never seen—or even thought of—before.”
The Nancy Grace Roman Space Telescope, likewise, will dwarf the Hubble, observing 100 times as large an area, and collecting a thousand times as much information across a range of wavelengths, along with spectroscopy (to characterize the distance and composition of objects) and time-series measurements, each of which takes a huge amount of computing power to interpret.
This profusion of data is an extraordinary opportunity—and an enormous challenge. While astronomers have long used data science and machine learning to analyze images, the new telescopes mean that astronomers need to harness the rapidly advancing fields of data analysis and artificial intelligence to manage the massive amounts of data. (Rubin Observatory data will be hosted and processed at SLAC, as well as at various data centers around the world, and are expected to be used by many thousands of scientists.)
That demand is what drove Wechsler and Chris Mentzel, executive director of Stanford Data Science, to create the Center for Decoding the Universe. As a flagship program of Stanford Data Science, the center was designed to bring the university’s best data experts together with astronomers tackling some of the biggest problems in cosmology—including the nature of dark energy and dark matter, and the mysterious forces driving the universe’s expansion.
)
The observatory’s ability to continuously scan the sky and map the cosmos at this scale underscores why Stanford researchers are racing to prepare. The data firehose has already been turned on.
Two spiral galaxies that are actively merging. Photo: SALSA: Lopez-Rodriguez et al 2023, The Astrophysical Journal Letters, Volume 942, L13
For instance, to measure the Hubble constant (the rate at which the universe is expanding), doctoral candidate and Data Science Scholar Sydney Erickson, PhD ’27, models the gravitational warping, or lensing, around massive objects like galaxies.
“Right now, we have hundreds of these objects to work with,” Erickson says. “And that’s already challenging. With the new telescopes, we’ll have on the order of 10,000. It would be impossible for a person using an existing model to incorporate all those sources.”
One new tool Erickson and others are using, simulation-based inference, uses neural networks to detect patterns hidden in data sets that are too large for ordinary analytic software to plumb. While simulation-based inference was developed by physicists, Wechsler and Mentzel believe that the best way to supercharge data analysis in astronomy is to bring together experts from a wide range of specialties.
)
A visual representation of data being processed over the Rosette Nebula. Photo: RubinObs/NOIRLab/NSF/AURA/T.A. Rector (University of Alaska Anchorage)/H. Schweiker/WIYN
“We really want to spark collaboration,” Wechsler says. “Between astrophysicists and computer scientists as one example, but also by bringing people from many different disciplines together to solve really hard problems.”
Indeed, the Center for Decoding the Universe was designed with precisely this sort of cross-pollination in mind: “That’s the environment we’re building, at the intersection of AI, machine learning, and physics,” Mentzel says. “We’re trying to engineer serendipity.”
One of the great things about this approach, Mentzel notes, is that the benefits run both ways. In the near term, developing these tools and techniques will improve our understanding of the universe. But those same improvements will also translate to other spheres, like medical diagnostics and computer vision.
“The things data scientists and computer scientists learn from tackling our problems will advance work in a wide range of fields,” Wechsler agrees, adding that many have jumped at the chance to contribute. “People love astronomy,” Wechsler says. “I mean, who doesn’t want to figure out the secrets of the universe?”
##
Wechsler is the Humanities and Sciences Professor II and professor of physics in the School of Humanities and Sciences and professor of particle physics and astrophysics at SLAC National Accelerator Laboratory, as well as director of the Center for Decoding the Universe.
Catalyzing Discovery: The impact
Why it matters
Astronomy has entered a new era—one defined not only by what we can see, but by what we can compute. As next-generation telescopes like the Vera C. Rubin Observatory and the Roman Space Telescope unleash an unprecedented deluge of cosmic data, scientists face an urgent question: How do we make sense of it all? Stanford’s new Center for Decoding the Universe is answering that call, bringing together astrophysicists, data scientists, and AI experts to uncover insights hidden deep in the data. Their work promises to reshape our understanding of the universe—and push the boundaries of discovery across science and technology.
The opportunity
Large and complex data sets now drive nearly every aspect of science and discovery. In the decades to come, our progress will increasingly rely on our ability to learn from that data. Stanford Data Science is a collaborative effort to weave data science into the core fabric of Stanford’s research and teaching enterprises, equipping students and faculty to harness the data revolution in solving the world’s most consequential problems. With support, Stanford can scale bold interdisciplinary models like the Center for Decoding the Universe, accelerate innovation in AI and scientific discovery, and drive progress in fields as far-reaching as health care, climate science, and beyond. Learn more about Stanford Data Science.
)
:focal(1216x1354:1217x1355))
:focal(1216x1354:1217x1355))
)
)