NASA Kepler Googled an Alien Solar System and Found It Matches Our Own With a Record Eight Planets

An artist's portrayal of the Kepler Space Telescope, a major development in the hunt for extraterrestrial life, spotting exoplanets. NASA/JPL

Updated | Thursday afternoon, NASA announced that scientists used a Google neural network to identify an eighth planet in a foreign solar system. That makes the solar system, orbiting a star called Kepler-90, the first to tie our own in terms of number of planets. (Sorry, Pluto-lovers.) It's also the first exoplanet to be identified by a neural network.

That neural network was developed by Christopher Shallue, a software engineer focusing on artificial intelligence at Google, and Andrew Vanderburg, an astronomer at the University of Texas at Austin. The discovery is written up in a paper accepted for publication by The Astronomical Journal.

"A neural network is loosely inspired by the structure of the human brain," Shallue said during a press conference held to announce the discovery. Shallue and Vanderburg fed previously analyzed Kepler data to that neural network to teach it how to identify planets. That process got the neural network to a point at which it could correctly recognize a plant 96 percent of the time, Shallue said. Shallue said the model takes about two hours to train and will be released to the public.

Once the network was primed, the scientists fed it more data and waited to see what it could do. So far, the neural network has identified two previously unidentified planets that produced signals so faint scientists hadn't yet been able to spot them. One of those is Kepler-90i; the second, Kepler-80g, is the sixth planet in its solar system and appears to be precisely aligned orbitally with one of its neighbors.

Read more: Kepler Mission: 5 of the Most Incredible Discoveries From NASA's Alien-Hunting Spacecraft

The Kepler mission first launched in 2009 and was designed to spot exoplanets using what astronomers call the transit technique. That process works by staring at individual stars and measuring the amount of light they give off. As a planet passes between its star and the Kepler telescope, it blocks a tiny portion of the star's light. In the Kepler data, scientists can recognize the small dip in brightness. Once they see the same dip three times, they become suspicious that it's a real planet. Currently, the telescope's tally of confirmed exoplanets stands at 2,525.

The solar system where the new planet was discovered isn't new. "We already knew that Kepler-90 hosted seven planets," Vanderburg said during the press conference. Its star is a little larger and hotter than our own. At that point, this alien solar system tied the renowned TRAPPIST solar system, which at its discovery had the most planets capable of holding liquid water on their surface. But now the neural network has identified an eighth, known as Kepler-90i. Scientists have also been able to determine that the surface of the new planet is likely about 800 degrees Fahrenheit.

A depiction of the Kepler-90i solar system, the first known foreign solar system with eight planets. NASA/Wendy Stenzel

While the Kepler-90 solar system matches ours in terms of the number of planets, it doesn't quite mirror our home turf: All eight planets orbit their star within the distance between Earth and our sun. That might mean that planets that formed farther out somehow migrated in toward the star, Vanderburg said.

There's still plenty more Kepler data to be studied, Shallue and Vanderburg said. They intend to feed their neural network the complete data set, which includes 150,000 stars. There is also even more data from the spacecraft's new incarnation, K2, which gathers similar data on a larger swath of the sky. As Jessie Dotson, a NASA astrophysicist on the K2 mission, said, "I'm on the edge of my seat," to see what that analysis yields.

This story has been updated to more accurately reflect past use of machine learning and neural networks in exoplanet searches.