Astronomers use AI tools to find new galaxies using existing data

In short: Astronomers are using focused AI tools to spot unusual space objects in telescope data, often without needing huge computers.

What's going on

Researchers at the European Space Agency, David O’Ryan and Pablo Gómez, used a neural network called AnomalyMatch to search the Hubble Space Telescope image archive. A neural network is a type of AI that learns from examples, a bit like training a sniffer dog to recognize certain scents. They report finding hundreds of unusual objects, including oddly shaped galaxies that had been missed when people looked through images by hand.

At Oxford University, physicist Héloïse Stevance helped build an AI-based Virtual Research Assistant to sort daily alerts from ATLAS, a group of five ground telescopes that scan the sky. The tool reduced by about 85 percent the time scientists spent “eyeballing” alerts to confirm real supernova signals. Supernovae are extremely powerful explosions from dying stars.

At the University of Birmingham, researchers are using AI “emulators” to speed up scientific modeling. An emulator here means an AI that can imitate a slow, complex simulation, like a quick sketch that matches a detailed painting closely enough to guide decisions. After training, one model can be checked in under a millisecond, according to physicist Guy Davies.

What to watch

More data is coming soon. The Vera Rubin Observatory in Chile is expected to take an image every 30 seconds for 10 years, creating what it calls the largest astronomical time-lapse ever. Researchers say these smaller, specialized AI tools can help find the interesting needles in that haystack, sometimes using as little as a laptop or a single graphics processor.

Source: Financial Times