AI Explores Hubble’s Archive for New Discoveries

A duo of astronomers from the European Space Agency (ESA) has uncovered over 800 previously unrecorded “astrophysical anomalies” lurking in the archives of the Hubble Space Telescope. Researchers David O’Ryan and Pablo Gómez developed an AI model to sift through Hubble’s extensive 35-year dataset, searching for unusual objects and marking them for further examination. O’Ryan described it as “a treasure trove of data in which astrophysical anomalies might be found.”

Exploring the cosmos is no easy feat. The vastness of space, combined with the overwhelming data produced by instruments like Hubble, can challenge even the most robust research teams. Plus, space can be downright bizarre. This is where AI shines, adept at analyzing massive datasets to identify patterns and highlight the peculiarities that astronomers might overlook.

The AI model, named AnomalyMatch, meticulously scanned nearly 100 million image cutouts from the Hubble Legacy Archive, marking the first systematic search for anomalies within this dataset. It identified everything from oddly shaped galaxies to light distorted by the gravity of massive celestial bodies, as well as planet-forming discs viewed edge-on. Remarkably, AnomalyMatch completed its analysis in just two and a half days, a task that would have taken human researchers significantly longer.

The results, shared in the journal Astronomy & Astrophysics, revealed close to 1,400 “anomalous objects,” with the majority being galaxies that are merging or interacting. Other intriguing findings included gravitational lenses (where light is bent into circles or arcs by massive objects), jellyfish galaxies (characterized by trailing “tentacles” of gas), and galaxies with large star clusters. “Perhaps most intriguing of all, there were several dozen objects that defied classification altogether,” noted ESA in a blog post.

“This is a fantastic use of AI to maximize the scientific output of the Hubble archive,” Gómez remarked. “Finding so many anomalous objects in Hubble data, where you might expect many to have already been found, is a great result. It also demonstrates the potential of this tool for analyzing other large datasets.”