AI Learns to Listen to Black Holes: A Simple Recipe for Gravitational Waves ⚡ экспресс
The GWAgent system, based on a large language model, has been developed to create analytical (formulaic) surrogates that replace expensive simulations. Using the example of gravitational waves from the merger of eccentric black holes, it was shown that a physically motivated hint significantly improves accuracy. The resulting model achieves a mismatch error of 6.9×10⁻⁴ and speeds up calculations by 8.4 times, outperforming both symbolic regression and conventional machine learning. The agent also revealed hidden physical structure, allowing the eccentricity of the real event GW200129 to be measured. This demonstrates that agentive AI systems with data validation can create accurate, fast, and understandable models.
When black holes collide, the very fabric of spacetime trembles. That rumble—gravitational waves—gets picked up by detectors, but decoding the signal used to take hours of computer simulation. The GWAgent algorithm flips the script: it acts like a master chef who tastes a finished dish and is asked to name the recipe. The chef isn't working blind—there's a hint like "base is puff pastry." Similarly, the agent leans on basic physics, tries out simple formulas, tests them against real simulations, and within minutes lands on a nearly exact answer—eight times faster than complex calculations.
Applying the method to the real signal GW200129, recorded by LIGO, scientists were surprised to find that the two black holes were not spiraling in a neat circle but in a highly flattened oval. That's a telltale sign that they were brought together by the hustle and bustle of a dense star cluster—a scenario for which there was previously no direct evidence.
🎯 The first gravitational wave ever detected stretched the four-kilometer arm of LIGO by just a thousandth of the width of a proton—a minuteness that humans managed to measure.
🎬 Sci-fi writers often turn mysterious cosmic signals into alien messages. But even the silent whisper of black holes needs a skilled decoder—and now AI is stepping into that role.