New AI Trick Can Pinpoint Gravitational Wave Sources in Just One Second – Gizmodo
A team of researchers just introduced a new way of detecting gravitational wave sources that they posit could vastly improve precision of detections and expedite the detection of such enigmatic events.The team’s research, published today in the journal Nature, outlines an algorithm to study neutron star mergers’ gravitational wave emissions. Once identified, astronomers around the world could be notified of the event, allowing experts to collect as much information as possible about the fleeting, mysterious sources of gravitational waves.Let’s slow down for a moment. Gravitational waves are ripples in space-time, first predicted by Einstein over a century ago and only observed for the first time in 2015 by part of what is now the LIGO-Virgo-KAGRA Collaboration. Gravitational waves are generated by interactions of some of the universe’s densest objects: black holes and neutron stars.The team’s algorithm targeted neutron stars in a death spiral with one another, slowly encroaching on each other until they merge—ergo, a “neutron star merger.” Detecting the gravitational waves emitted by neutron stars and black holes helps astronomers understand the structure of neutron stars, the origin of some of the heavy elements, better tests the theory of general relativity and measure the rate of the universe’s expansion, and potentially shed light on the nature of dark matter.Artificial intelligence can speed up analysis of these gravitational wave events, and, based on the team’s results, improve accuracy in the prediction of the source merger’s location. According to the team, the method can asses the origin of gravitational saves in just one second, and that the method can serve as a blueprint for data analysis for next-generation gravitational wave detectors, such as LISA.“Once trained, when a new observation is made, the neural network can take the measurement as input and predicts the BNS [binary neutron star] properties (including localization) within a second,” said Maximilian Dax, a machine learning researcher and physicist at the University of Tübingen, and lead author of the study, in an email to Gizmodo. “This is so fast because we don’t need new GW [gravitational wave] simulations at inference.”“We hope that our method will help to observe more electromagnetic signals emitted by BNS mergers, and to observe them earlier (ie, closer to the merger),” Dax added. “Such multi-messenger observations are extremely exciting, and are relevant in a variety of fields, including cosmology, nuclear physics and gravity.”The team’s algorithm is 30% more accurate in its results than previous iterations, and can help astronomers determine what merger events require further, often time-sensitive, observations.“Machine learning has garnered a lot of attention recently within gravitational-wave research as a way to improve or even replace existing analysis techniques,” said Michael Williams, a cosmologist at the University of Portsmouth in the United Kingdom, in a News & Views article.“Several challenges remain, however,” Williams, who’s not affiliated with the new research, added. “The performance of machine-learning algorithms is, in general, highly dependent on their training. For this algorithm, one problem is that the properties of real noise in gravitational-wave detectors vary over time from the properties assumed when training the network. This introduces systematic errors that can bias results.”The “real trial by fire,” Williams concluded, is whether the team’s algorithm will be able to disseminate information about the next binary neutron-star merger when it occurs.Time will tell how effective the machine-learning-based approach is, but with state-of-the-art observatories coming online in the near future—perhaps most notably the Vera Rubin Observatory and its LSST Camera—detecting the cosmos’ transient events as soon as possible will be mission critical.
Artificial intelligenceBinary starsmachine learningNeutron stars
Get the best tech, science, and culture news in your inbox daily.
News from the future, delivered to your present.
Please select your desired newsletters and submit your email to upgrade your inbox.
The U.S. should not create its own Manhattan Project for AI, because such a project would invite retaliation from adversaries.
Some articles will also include a new section that automatically generates counterarguments using AI models from Perplexity.
Three people, including one Chinese national, were charged with conspiracy to commit fraud.
The Google co-founder says 60-hours a week is the productivity sweet spot for AI engineers.
Microsoft amended a lawsuit filed last year to name the four defendants whom it alleges misused its AI models to create celebrity deepfakes.
The platform has since reinstated the videos.
Best of CES 2025 Awards ➜We may earn a commission when you buy through links on our sites.
©2025 GIZMODO USA LLC. All rights reserved.Mode
Follow us
Mode
Follow us
