.svg){kind=spacer}
Details
- Google DeepMind has introduced Deep Loop Shaping, an AI-driven method using reinforcement learning to cut control noise in gravitational wave observatories by 30 to 100 times compared to existing controllers, with findings published in the journal Science on September 4, 2025.
- The effort is a collaboration between Google DeepMind, LIGO (operated by Caltech), and the Gran Sasso Science Institute, with the technology successfully tested at LIGO’s Livingston, Louisiana facility.
- By employing frequency domain rewards, the AI trains controllers that precisely stabilize LIGO’s 40-kilogram suspended mirrors, reducing noise below quantum fluctuation levels in the observation band used for identifying black hole mergers of several hundred solar masses.
- The innovation directly addresses a core challenge facing LIGO since 2015: environmental noise—such as distant ocean waves—can interfere with measurements that must be accurate to within 1/10,000th the diameter of a proton.
- This advancement could enable detection of hundreds more gravitational wave events annually, sharpen observations of elusive intermediate-mass black holes, and has potential future uses in fields like aerospace, robotics, and structural engineering.
Impact
Deep Loop Shaping marks a turning point for AI-driven scientific instrumentation, potentially doubling LIGO’s detection capacity and deepening our understanding of the universe’s most energetic events. The result underscores AI’s expanding influence in high-precision physics, echoing similar successes in particle physics instrumentation. By overcoming a foundational noise barrier, this method could quickly broaden the scope and sensitivity of both ground-based and future space-based gravitational wave observatories.
