Google’s really smart AI that was used a couple of months ago to discover two new exoplanets, was just open sourced by the company.

According to a blog signed by Chris Shallue, Senior Software Engineer with the Google Brain Team, the code for processing the data coming from the Kepler space telescope is now open for anyone to check out. They hope that this will be quite helpful for all astronomy enthusiasts and that it will be a useful starting point for developing similar models for other NASA missions.

The Kepler space telescope was launched nearly a decade ago to look for exoplanets. They managed to capture massive amounts of data, but processing everything is a slow job and it takes a lot of time. Astronomers normally figure out whether a certain star has any planets by judging the changes in the observed brightness of a star. More specifically, when a planet passes in front of its star, the brightness dims temporarily. By doing the math, they can tell how many stars pass in front of the star, how fast they spin around it and how big they are, to name a few.

The AI Google has developed can process all this data much faster and figure out which stars have planets and which don’t. Google used about 15,000 of 30,000 Kepler signals that had already been manually examined and classified by humans fed into the neural network. About 3,500 had already been verified as planets or strong planet candidates. All this data helped train the neural network to distinguish planets from false positives. Even so, the company admits that they haven’t yet managed to train the network properly to reject binary stars and other false positives, so more work needs to be done.

Up until now, Google has managed to go through 670 stars out of 200,000 observed by Kepler, so they have a long road ahead of them.

Sharing the AI

By making the code public, the door is open for other enthusiasts to help out on their own, bring improvements or discoveries. Hopefully, something will come of this. For the past couple of decades, as AI has evolved, we’ve seen numerous instances when humans and Artificial Intelligence have cooperated efficiently. One of the first cases was in 2005 when a couple of amateur chess players properly trained their systems to fight against the chess grandmasters, or even the other supercomputers. In their success, it wasn’t the performance of the computer they were using that was vital, but rather the way they managed to train their AI.

This is the perfect time in history to hope for more astronomic discoveries. Our satellites, telescopes, and space projects have accumulated massive amounts of data; so much, in fact, that it would take a long time to go through it all without some kind of help. Artificial Intelligence and neural networks are here to help – they’re the solution we were waiting for to process all this data. In fact, Artificial Intelligence is probably our best bet to find any kind of alien life form in outer space, by making assumptions about the new worlds we discover, by judging the temperatures, the position of the planets, the atmospheric composition, and so on.