It was only a matter of time until scientists discovered a star system that had the exact same number of planets as Earth’s solar system. NASA’s Kepler Space Telescope discovered an eighth planet orbiting a star similar to our own sun which is about 2,500 light-years from Earth. The discovery itself can stand on its own as a notable breakthrough, however, what’s more compelling is how this find was possible.
Thanks to technology provided by Google, the Kepler Space Telescope was able to find the elusive eighth planet known as Kepler-90i, a sizzling hot and rocky planet that orbits its star once every 14.4 days. Artificial intelligence has been slowly pouring into NASA’s technological upgrades, and the people at Google have been making sure that those upgrades are effective in making more worthwhile discoveries.
As with all machines, they require data to process and generate an algorithm. Planets detected by the Kepler telescope appear as silhouettes that block the light coming from a star, which is the result of a celestial body passing in front of its parent star and blocking some of the light. Google’s A.I. was tasked with sifting through all previous exoplanet discoveries and comparing false alerts with true finds. The algorithm was tailored specifically for finding exoplanets.
Researchers, Christopher Shallue, and Andrew Vanderburg, first came up with the idea of instructing a computer to recognize exoplanets in the light readings recorded by the Kepler Space Telescope. The two researchers were inspired by the way neurons connect in the human brain and this became the basis for the new sorting technology.
“The Kepler-90 star system is similar to a mini-version of our solar system. You have small planetoids inside and big planets outside, but everything is scrunched much closer,” said Vanderburg, a NASA Sagan Postdoctoral Fellow and cosmologist at the University of Texas at Austin.
Shallue, who is a senior software engineer in Google’s AI team, became interested in exoplanet discoveries and wanted to apply the “neural network” tech to existing Kepler data and see if there were differences in its detection pattern.
Shallue and Vanderburg first fed their neural network, 15 thousand previously-vetted signals from the Kepler exoplanet list. It then sifted through all of the data and determine which signals were of exoplanets and which were false-positives. The neural-network had a 96 percent success rate, according to the two researchers. Their findings have been published in The Astronomical Journal.
Image Source: Nasa.gov