Artificial intelligence is the talk of the hour. Little by Little, AI machines are making their way through our daily lives, through the internet and its services as well as in the electronic devices we use every day, including our mobile phones.
But AI applications are not limited to the traditional concept of technology. The development of Artificial Intelligence can now help scientists predict the likelihood of life on other planets. A study recently presented at the European Week of Astronomy and Space Sciences (held in Liverpool) by a group of researchers from the University of Plymouth demonstrates how it is possible to use neural networks to classify planets in five different categories, estimating in each case, the probability that there is life on them. Such a system could be used perfectly in future space exploration missions.
Artificial neural networks are systems that try to replicate the way of learning of the human brain and are especially useful for identifying patterns that are too complicated to process by a biological brain.
Under the direction of Christopher Bishop, the team of scientists trained a neural network to classify exoplanets into five different categories, depending on whether they resemble the current Earth, the primitive Earth, Mars, Venus or the moon of Saturn Titan. The five reference worlds are rocky and possess an atmosphere, as well as being among the “most habitable” objects of our Solar System.
In the words of Bishop, “we are interested in these neural networks so that they establish the priorities of exploration of a hypothetical, intelligent and interstellar spacecraft that is capable of scanning a planetary system at a distance”.
According to the researcher, “we are also studying the use of large area deployable antennas to collect data from an interstellar probe that would be located far away from Earth. Something that would be very necessary if this technology comes to be used in robotic spacecraft in the future.”
The atmospheric observations, or spectra, of the five chosen bodies of the Solar System were supplied to the neural network, which was then asked to classify them into five planetary types. As we know that, for now, the existence of life is only confirmed on Earth, the AI classification used a “life-chance” metric based on the well-known atmospheric and orbital properties of the five reference worlds.
With these premises, Bishop trained the neural network with more than a hundred different spectral profiles, each with hundreds of parameters that contribute in one way or another to habitability. And the results, for the moment, have been satisfactory since the network worked correctly by presenting an atmospheric spectral profile of evidence, which has not been seen before.
“Given the results obtained so far,” says Angelo Cangelosi, project supervisor, “the method can be extremely useful when categorizing different types of exoplanets.”
The technique may also be ideal for selecting targets for future observations, given the increased detail and improved capabilities of the new space observation missions, such as ESA’s Ariel, and NASA’s James Webb Telescope.