UT Astronomer Discovers New Planets – Using Artificial Intelligence

Machine learning could change the game for NASA.

By Alexandra HartDecember 15, 2017 7:09 am

Thursday afternoon, NASA announced a new discovery from its Kepler space telescope mission – the presence of two new exoplanets. While that alone isn’t that remarkable, since Kepler has discovered and confirmed upwards of 2,300 exoplanets in distant solar systems, how scientists found those planets is noteworthy.

Andrew Vanderburg, astronomer and NASA Sagan Postdoctoral Fellow at the University of Texas at Austin, who worked on this discovery, says the planet was discovered using artificial intelligence.

“Instead of using traditional methods to identify exoplanets, we’ve incorporated an advanced type of machine learning called a neural network to help identify planets and sort away the bad signals, the false positive signals from the real planets,” he says. “It’s a form of artificial intelligence, very loosely inspired by the structure of neurons in your brain.”

One of the newly discovered planets, Kepler-90i, is also remarkable because it’s the eighth planet in its system.

“This makes Kepler 90 the only star in our galaxy that we know of, besides our own sun, to host that many planets,” Vanderburg says.

He says the planet is only about 30 percent larger than Earth, and it’s much hotter than our planet.

Even with his potentially powerful new tool, developed in collaboration with Google software engineer Christopher Shallue, Vanderburg says he’s not expecting machines to take over human astronomy jobs any time soon.

“I don’t think it’s ever going to replace astronomers. It only will help astronomers do things that otherwise are difficult or tedious,” he says. “I’ve spent many hours of my life staring at signals for planets and deciding that’s not a planet. So if I had a tool like this machine learning technology that would do it for me, that would give me more free time to do the creative things that humans do best.”


Written by Jen Rice.