The Use of Artificial Intelligence in Astronomical Discovery


Photo via Maxpixel


Artificial Intelligence, or AI for short, is the mimicking of human cognitive functions associated with learning and problem solving using some specialized software programs.

With the dawn of the computer age in the 20th century, AI has now come to fruition as computer and programming technologies continue to advance. And so it should not come as a surprise that it is now also being used in astronomy, and especially in the discovery of new stars and planetary systems. An example of this is the newly-discovered planet Kepler-901 which is a hot and rocky planet orbiting a star every 14.4 days The said planet was discovered using artificial intelligence.

Machine learning is one approach to artificial intelligence and in this case, it was able to learn how to find distant planets and other systems. And in this case, machine learning from Google was used.

As written in Astronomy Now, an astronomy magazine in the UK, Christopher Shallue and Andrew Vanderburg trained a computer to learn to identify exoplanets in readings by Kepler that were done earlier. Miniscule light is captured when a planet passes in front of a star and using neuronal connections in the brain as inspiration, the neuron-like network sifted through Kepler data and found the eighth planet orbiting Kepler-90 from data that were previously missed.

They did it by first training the neural network with a set of 15,000 vetted signals from the Kepler exoplanet catalog. It learned to identify true planets and false positives some 96 percent of the time. Then after the AI had learned, they set it out searching in 670  star systems that had multiple planets as they assumed that it would be the best area to look for other exoplanets. Their assumptions were validated by the discovery of Kepler 90i using the artificial intelligence. 

There may be a lot more exciting discoveries waiting in the archived Kepler data which have waited until this time when artificial intelligence is ripe enough to be used on old data, according to NASA. This shows that artificial intelligence, especially artificial neural-like AI systems, can mimic the learning system of humans and will be effective in astronomy.