Google DeepMind’s AI Breakthrough at LIGO Promises Clearer Detection of Black Holes and Cosmic Collisions

Researchers at Google DeepMind have introduced Deep Loop Shaping, an AI-powered control system designed to stabilise the world’s most sensitive observatories, including the Laser Interferometer Gravitational-Wave Observatory (LIGO).

The breakthrough, published in Science, promises to unlock more precise measurements of cosmic events such as black hole mergers and neutron star collisions.

"We developed Deep Loop Shaping in collaboration with LIGO (Laser Interferometer Gravitational-Wave Observatory) operated by Caltech, and GSSI (Gran Sasso Science Institute), and proved our method at the observatory in Livingston, Louisiana," Google DeepMind said in a blog post.

Deep Loop Shaping reduces noise in LIGO’s feedback systems by 30 to 100 times, stabilising the mirrors critical to detecting gravitational waves. These waves, ripples in space-time first predicted by Albert Einstein, require measurements at scales smaller than a proton — making vibration control essential.

“Studying the universe using gravity instead of light, is like listening instead of looking. This work allows us to tune in to the bass,” said Rana Adhikari, Professor of Physics at Caltech.

LIGO, located in Louisiana, relies on 4-kilometer laser arms and ultra-sensitive mirrors to capture signals. Environmental disturbances — even ocean waves 100 miles away — can disrupt readings. Deep Loop Shaping uses reinforcement learning to suppress this “control noise,” significantly improving detection capabilities.

Researchers say applying the method across LIGO could help astronomers record hundreds more cosmic events annually, providing insights into galaxy formation and elusive intermediate-mass black holes. Beyond astrophysics, the technique could benefit fields like aerospace, robotics, and structural engineering.