UC Berkeley researchers use artificial intelligence to detect radio bursts in space

SETI Research Center/Courtesy

Related Posts

Researchers at Breakthrough Listen, a UC Berkeley-based research initiative, are using artificial intelligence, or AI, to discover new fast radio bursts in an ongoing effort to locate extraterrestrial intelligence.

In 2017, researchers at the Green Bank Telescope in West Virginia detected 21 bursts within the first hour of their six-hour investigation. After the employment of AI to detect these radio bursts, Breakthrough Listen found an additional 72 radio bursts emitted from an unknown source approximately 3 billion light years away from Earth. Unlike audio signals, fast radio bursts are short bursts of radio emission, ranging from a microsecond to a second long.

“That’s really cool,” said ASUC Senator and campus junior Anne Zepecki. “I think that really goes to show the diverse application of technology and artificial intelligence. … Science can do a lot of good and can help us learn a lot about the world.”

The AI technology allows machines to learn in a fashion similar to how Google trains its search engine, according to associate project astronomer Steve Croft. He added that by feeding many examples of familiar radio bursts from already-known sources, such as satellites and cell phones, the algorithm can learn to detect unfamiliar ones.

Learning algorithms, such as the one used by Breakthrough Listen researchers, have been proposed since the 1970s, according to Gerry Zhang, a campus graduate student in astronomy and researcher at Breakthrough Listen. He added that technological advances allowed “explosive growth” in these algorithms, especially after 2012.

“In this particular case, AI allowed us to have higher sensitivity and speed over traditional algorithm,” Zhang said in an email. “We detected 72 new pulses, which, among other features, allowed us to put a first constraint on the periodicity of these signals.”

Croft said the radio bursts are not audio signals but are pulses detected by the algorithm as a series of images and could be a sign of “extraterrestrial intelligence.”

Zhang notes that this technology has “extensive” applications in the defense and telecommunications industries. He added that improving the capabilities of signal recognition is an exciting effort for the radio frequency community.

“I think it’s a groundbreaking result that is an imaginative use of existing technology to shed some new light on an astrophysical source,” Croft said.

Contact Yao Huang and Bryanna Paz at [email protected].