In case you hadn’t noticed, space is pretty large, which makes it challenging to survey with mere human eyes. But for AI, that enormous scale presents an opportunity to further study faraway stars and celestial objects closer to home–including the moon. Astronomers made their first major AI breakthrough earlier this month, when a team of researchers announced that they had used a neural network to find 6,000 new moon craters.
The University of Toronto-Scarborough team, led by Mohamad Ali-Dib, a postdoctoral fellow in the Centre for Planetary Sciences (CPS), took data from elevation maps of the moon gathered from orbiting satellites and fed it into their “convolutional neural network,” which is the same kind of machine learning system popular with self-driving cars. After training the neural network on a data set that covered two-thirds of the moon, they tasked the network with counting craters on the remaining third. The strategy worked well, revealing 6,000 previously unidentified craters. While counting craters might sound like mind-numbingly dull work (even for AI), the payoff is substantial: craters offer important clues about the history of the solar system and the rocks that zip around it.
What’s next for this neural network? Counting craters on Mercury, and then maybe Mars, Ceres, and the icy moons of Jupiter and Saturn.
H/T The Verge