AI & Hubble: 1,400 New Space Objects Discovered

Two astronomers from the European Space Agency (ESA) have discovered more than 800 previously undocumented “astrophysical anomalies” hidden in the Hubble archives. To do this, researchers David O’Ryan and Pablo Gomez trained an AI model to comb through 35 years of Hubble data sets, looking for UFOs and flagging them for human review. “This is a treasure trove of data that can uncover astrophysical anomalies,” Orion said in a statement.

Studying space is difficult. They are numerous, the environment is noisy, and the vast amounts of data produced by instruments like the Hubble Space Telescope can overwhelm even large research teams. Space is weird sometimes. Very strange. AI is very good at sifting through large amounts of information to identify patterns, pointing out anomalies that astronomers might have missed.

The model used by the astronomers, called AnomalyMatch, scanned nearly 100 million cropped images from the Hubble Legacy Archive, marking the first time the data set has been systematically searched for anomalies. Think of oddly shaped galaxies, light twisted by the gravity of massive objects, or disks forming planets when viewed from the edge. AnomalyMatch took just two and a half days to browse the dataset, which is much faster than a human research team would attempt to perform this task.

Results published in journal Astronomy and AstrophysicsIt revealed nearly 1,400 “anomalous objects,” most of which were merging or interacting galaxies. Other anomalies include gravitational lensing (where light is twisted into circles or arcs by massive objects in front), jellyfish galaxies (with dangling “tentacles” of gas), and galaxies containing large chunks of stars. “Perhaps most interestingly, there are dozens of objects that completely defy classification,” the European Space Agency said in a blog post.

“This is a wonderful use of artificial intelligence to maximize the scientific output of the Hubble archive,” Gomez said. “Finding so many anomalous objects in the Hubble data, as you might expect to find a lot of anomalous objects, is a great result. It also shows how useful this tool can be for other large data sets.”