From pixel to universe and back

1 Dec 2021

Astrophysicist Daniel Gruen is using artificial intelligence to explore the cosmos. He accepted a professorship at LMU’s Faculty of Physics in summer 2021.

Daniel Gruen wants to make the invisible visible. To do so, he depends as much on the capabilities of telescopes as on the ability to crunch vast quantities of data. “My research ranges from the pixel data we record with telescopes – images and spectra of galaxies – to inferences about the fundamental physical laws and elements of the universe,” the astrophysicist says. "We take pictures of the universe and ask ourselves: How can we learn something from them about how the universe has evolved and how its story will unfold from here?”

Back to the front line

Daniel Gruen studied physics at LMU. Back when he was preparing his master’s thesis, the cosmologist moved to the USA, like so many of his colleagues who were collaborating in the States on tremendous experiments to explore space. Under Professor Gary Bernstein at the University of Pennsylvania, Gruen investigated gravitational lensing – a topic that shapes his research to this day.

“Gravitational lensing is pivotal to cosmology, because it allows us to see what structures of matter exist in the universe,” he explains. “When light in the universe passes matter on its way to us, it is deflected. That distorts our images of distant galaxies. And from this distortion we can learn something about the distribution of matter that would otherwise be invisible to us, because the majority of matter in space is what is known as dark matter.”

After completing his studies at LMU and the University of Pennsylvania, Gruen engaged in research as a NASA Einstein Fellow and postdoctoral researcher at Stanford University and, subsequently, as a Panofsky Fellow in the SLAC National Accelerator Laboratory at Stanford University’s Kavli Institute. Here he led a working group before deciding to return to LMU.

Since summer 2021, he has held the Chair of Astrophysics, Cosmology and Artificial Intelligence here in Munich. “The conditions in place at LMU to build a large-scale, long-term and broad-based research project are very good and stand up excellently to international comparison,” Gruen says. “The same goes for opportunities to engage in collaboration on major experiments.” This, he adds, is in part thanks to the AI initiative in Bavaria that has facilitated the creation of professorial chairs which “can implement new concepts”.

“If we as academics at LMU make good use of these possibilities, then the sky is the limit,” the astrophysicist concludes – consciously choosing an American idiom that effectively affirms: “Nothing is impossible.”

Mittelalter schwarzhaariger Mann mit Bart steht zwischen einen Sternenbild und einem Rechenserverschrank

Daniel Gruen


Ahead of the AI curve

Humankind uses huge telescopes whose ever larger mirrors look to the heavens to try to paint a picture of the vast expanses of space. These telescopes are far removed from civilization: in Chile’s Atacama Desert, for example, and on high mountains such as La Silla. Today, the sheer volume of data collected by these gigantic devices gives rise to new research questions. “In astrophysics, ever bigger mirrors, bigger cameras and better optics are now giving us incredibly higher volumes of data very fast,” Gruen says. “Whereas I was still working with 10,000 galaxies for my doctoral thesis, we are working with 100 million of them today and will work with billions of galaxies in a few years.”

Merely being able to process such immense quantities of data is a challenge for research. For astrophysicists such as Daniel Gruen, artificial intelligence (AI) applications have therefore become indispensable tools. “We have to think up ever better algorithms,” he says, admitting that he first has to ponder what exactly he wants to calculate. “There is so much information in the data. But it is not obvious where that information is hidden and how to extract it.” AI could be a key to doing so.

The cosmologist is certainly benefiting from the rapid development of artificial intelligence: “Exciting new concepts that open up new opportunities are being discovered all the time.” On the other hand, he first has to adapt it to the very specific research questions he addresses in astrophysics. In effect, he always needs to be ahead of the AI curve: “The problems we are trying to solve are different to those where AI has already become a part of everyday life. We are learning to change things about the way machines learn.”

Open to new ideas

Gruen also sees this as a wonderful space for students and junior researchers. “We need young people with fresh ideas, people who think creatively about how artificial intelligence can be used. If you have a good grasp of AI, there is an incredible amount you can achieve in astrophysics.” He nevertheless stresses that this does not apply only to his discipline: “It is an area of research that is tremendously useful to students even if they don’t necessarily want to become astrophysicists.”

At LMU, he now relishes the chance to interact with colleagues, including those from other disciplines. Gruen is, for example, part of the team of speakers who will deliver the KI Lectures, a series of interdisciplinary talks on artificial intelligence. “Forums for dialogue are beginning to emerge, and that is something I am very much looking forward to. I am sure that this will yield new insights for us, too.” The astrophysicist emphasizes that he is “open to new ideas”.

Dialogue also takes place within his discipline at LMU, as well as within the framework of the cluster of excellence Origins and with scientists at both the Max Planck Institutes and the European Southern Observatory in Garching. “This concentration of astrophysicists in a single location is without equal,” Gruen says. “Munich is the place to be for astrophysics in the whole of Europe.” For a researcher seeking to solve the mysteries of the universe, Daniel Gruen clearly seems to be in the right place at the right time.

What are you looking for?