Researchers at the Max Planck Institute for the Science of Light (MPL) have been investigating how an artificial intelligence system can scan an infinitely enormous area of possible designs to discover entirely new solutions. The results were just published in the journal Physical Review X.
Illustration of the first gravitational wave event observed by LIGO. The detected wave forms from LIGO Hanford (orange) and LIGO Livingston (blue) are superimposed beneath illustrations of the merging black holes. Image Credit: Aurore Simmonet (Sonoma State University), Courtesy Caltech/MIT/LIGO Laboratory
Extreme cosmic events, such as colliding black holes or star explosions, can create spacetime ripples known as gravitational waves. Their discovery provided new insights into the universe. They must be observed using ultra-precise detectors that are extremely challenging to design.
Einstein made the theoretical prediction of gravitational waves more than a century ago. Only in 2016 were they directly identified due to the complexity of developing the required detectors. Together with the LIGO (“Laser Interferometer Gravitational-Wave Observatory”) team, which successfully constructed those detectors, Dr. Mario Krenn, head of MPL's Artificial Scientist Lab research group, has developed an AI-based algorithm called Urania to create new interferometric gravitational wave detectors.
The measurement technique known as interferometry makes use of the interference of waves, or their superposition upon meeting. The detector's layout and characteristics needed to be optimized, so the scientists approached the problem using methods inspired by modern machine learning. They framed the task as a continuous optimization problem to find the most effective solution.
They've developed several innovative experimental designs that outperform the most widely known next-generation detectors. These findings have the potential to increase the detectable signal range by more than a factor of 10.
Nonconformist and Creative: That’s What Urania Discovered
The researchers unearthed a number of well-known approaches in the algorithm solutions. Urania presented unconventional ideas that could transform the comprehension of detector technology.
After roughly two years of developing and running our AI algorithms, we discovered dozens of new solutions that seem to be better than experimental blueprints by human scientists. We asked ourselves what humans overlooked in comparison to the machine.
Dr. Mario Krenn, Leader, Artificial Scientist Lab, Max Planck Institute for the Science of Light
The researchers broadened their scientific methodology to better grasp the AI-discovered tactics, concepts, and procedures. Many of these remain absolutely alien to them. They assembled 50 top-performing prototypes in a public Detector Zoo and made them available to the scientific community for future investigation.
The recently released research demonstrates that AI could uncover unique detector designs and inspire human researchers to pursue new practical and theoretical concepts. More broadly, it implies that AI could play an important role in developing future tools for studying the universe.
Krenn concluded, “We are in an era where machines can discover new super-human solutions in science, and the task of humans is to understand what the machine has done. This will certainly become a very prominent part of the future of science.”
Journal Reference:
Krenn, M., et al. (2025) Digital Discovery of Interferometric Gravitational Wave Detectors. Physical Review X. doi.org/10.1103/PhysRevX.15.021012