Astronomers discover over 800 cosmic anomalies using a new AI tool

Here’s one use of AI that seems to do more good than harm. A pair of astronomers from the European Space Agency (ESA) have developed a neural network that searches for anomalies through space images. The results far exceeded the expectations of human experts. In two and a half days, it examined approximately 100 million image cutouts and discovered 1,400 unusual objects.

The creators of the AI ​​model, David O’Ryan and Pablo Gomez, call it AnomalyMatch. The pair trained it (and implemented it) on the Hubble Legacy Archive, which contains thousands of datasets from Hubble’s 35-year history. “While trained scientists excel at detecting cosmic anomalies, there is too much Hubble data for experts to sort through by hand at the required level of fine detail,” ESA writes in its press release.

After less than three days of scanning, AnomalyMatch returned a list of potential anomalies. It still required human eyes at the end: Gomez and O’Ryan reviewed the candidates to confirm which ones were indeed unusual. Of the 1,400 anomalous objects the pair confirmed, more than 800 were previously undocumented.

Most of the results showed galaxies merging or interacting, which could create strange shapes or long tails of stars and gas. Others were gravitational lenses. (This is where the foreground galaxy’s gravity bends space-time so that light from the background galaxy bends into a circle or arc.) Other discoveries include planet-forming disks, galaxies with giant clusters of stars, and jellyfish galaxies. Adding a bit of mystery, there were even “several dozen objects that completely defied classification.”

“This is a brilliant use of AI to maximize the scientific output of the Hubble collection,” Gomez is quoted as saying in ESA’s announcement. “Finding so many anomalous objects in the Hubble data, where you might expect many have already been found, is a fantastic result. It also shows how useful this tool will be for other large datasets.”