Machine learning applied to the search for extraterrestrial signals

Scientists suggest using machine learning algorithms to search for signals from extraterrestrial civilizations. We have never seen such a signal, but artificial intelligence can tell us how to find it.

Machine learning will help to identify signals of artificial origin

Scientists continue to search for extraterrestrial signals

A team of scientists from the University of Berkeley recently proposed using machine learning to search for signals from aliens. Despite the fact that we have never seen any such signal, and we have nothing to compare it with, computer algorithms are able to recognize it. 

Humanity has been looking for signals from extraterrestrial intelligence for 60 years. According to scientists, it should be a signal in the radio range that can be picked up by a large “dish”. The only problem is that the radio range is extremely wide, and at different frequencies it constantly receives some noise. At the same time, it can be very difficult to distinguish the message of an extraterrestrial civilization from the radiation of a natural object or traces of human activity.

After all, we only see current fluctuations in the electrical circuit connected to the antenna. In the sixties of the last century, scientists came to the conclusion that it was easiest to send and receive messages at the frequency of neutral hydrogen radiation — 1.42 GHz. It was believed that it was enough to study all natural phenomena occurring at this wavelength and find an artificial signal among them.

Broadband Transmissions

But the idea that it is easy to understand which of the noises is actually a signal of artificial origin turned out to be false. A lot of “sensational” signals over the past decades, after a long check, turned out to be something simple that we had not taken into account before. And it became increasingly difficult to navigate these “false messages”.

Our understanding of signal transmission has also changed. From the use of some radio frequencies, we are increasingly moving to broadband digital communications. Therefore, we now expect the same from the aliens.

We can analyze wide bands of the radio band quite well. In this case, the equipment of space telescopes receives not a graph of the signal, but its spectrogram, but the possible errors become several times greater.

Machine Learning

Scientists are increasingly realizing that detecting a signal of artificial origin is increasingly turning into a search for a suspicious thing that we have never seen, among a bunch of other suspicious things.

This is where it is important to apply machine learning. This is one of the areas of artificial intelligence. It consists in the fact that an artificial neural network passes through an abundance of already known information. Then it forms a certain set of reactions based on it. At the same time, it does not care at all how diverse and contradictory the information with which it will work will be later.

We don’t have any genuine signal from another civilization. Therefore, we have nothing to compare the huge volume of graphs and spectrograms with. But it is quite possible that we do not need it.

According to Р

Follow us on Twitter to get the most interesting space news in time