Astronomers discover over 800 cosmic anomalies using new AI tool

2 hours ago7 min read

In a move that feels ripped from the pages of a sci-fi novel, astronomers at the European Space Agency have unleashed an AI tool on the Hubble Space Telescope’s vast archive, and the results are cosmic. The neural network, dubbed AnomalyMatch by its creators David O’Ryan and Pablo Gómez, performed a task of staggering scale in mere days, sifting through nearly 100 million image snippets from Hubble’s 35-year history.Where human experts would be overwhelmed by the sheer volume of data, the AI acted as a tireless scout, flagging 1,400 potential oddities. After human review, over 800 of these were confirmed as previously undocumented cosmic anomalies—a treasure trove of the universe’s weird and wonderful.This isn't just about finding more merging galaxies, though it did plenty of that, revealing spectacular interactions that twist galaxies into odd shapes and long, starry tails. It also spotted gravitational lenses, those natural telescopes where a foreground galaxy’s gravity warps the light from a background object into rings and arcs.The haul included edge-on planet-forming disks, clumpy star-forming galaxies, and even elusive 'jellyfish' galaxies with streaming tendrils of gas. Most intriguingly, the AI pointed to dozens of objects that defy current classification altogether, hinting at phenomena we don’t yet understand.As Gómez noted, this is a fantastic way to maximize the scientific return from a legendary archive like Hubble’s, proving that even in well-studied data, secrets remain. The real promise, however, lies in the future.This tool is a prototype for the coming data deluge from observatories like the James Webb Space Telescope and the Vera C. Rubin Observatory, where AI will be an indispensable partner in navigating the cosmos, transforming petabytes of raw data into profound discoveries about the fabric of our universe.

#featured

#AI

#astronomy

#Hubble

#anomalies

#galaxies

#gravitational lenses

#AnomalyMatch