Breakthrough Listen

New AI Project Can Find More Light Flashes Than Scientists Can Do

A new project that is called “Breakthrough Listen” has discovered more fast radio bursts and mysterious flashes in deep space than astronomers can do.

Breakthrough Listen project has used artificial intelligence which may find “light pulses called fast radio bursts (FRBs) emanating from the dwarf galaxy FRB 121102 within the span of 1 hour”.

“On Aug. 26, 2017, astronomers with the Breakthrough Listen project — a $100 million effort to hunt for signs of intelligent alien life — spotted 21 repeating light pulses called fast radio bursts (FRBs) emanating from the dwarf galaxy FRB 121102 within the span of 1 hour”, informed

“Not all discoveries come from new observations,” Breakthrough Initiatives Executive Director Pete Worden said in the statement. (Breakthrough Listen is part of the larger Breakthrough Initiatives program, which also runs Breakthrough Starshot, Breakthrough Message and Breakthrough Watch.) “In this case, it was smart, original thinking applied to an existing dataset,” Worden added. “It has advanced our knowledge of one of the most tantalizing mysteries in astronomy.”

[That mystery remains, of course; we still don’t know what FRBs are. But the artificial-intelligence approach employed in the new study could lead to a variety of advances down the road], Zhang said. The team of new project applied machine-learning techniques on August 2017 data set, which was developed by the Green Bank Telescope in West Virginia and was originally analyzed using traditional methods.

The scientists, directed by UC Berkeley doctoral student Gerry Zhang, educated an algorithm called a “convolutional neural network” to spot FRBs among the 400 terabytes of data. The approach is similar to that employed by IT companies to optimize internet search results, Breakthrough Listen representatives said in a statement.

Zhang and his colleagues discovered an additional 72 light flashes, bringing the total number of FRBs spotted on that day.

“This work is only the beginning of using these powerful methods to find radio transients,” he said in the same statement. “We hope our success may inspire other serious endeavors in applying machine learning to radio astronomy.”