Predicting earthquakes has always been the most difficult, and until now impossible, challenge in all of Earth science.
Decades of significant international research have yet to uncover the signals that indicate how destructive an earthquake is and when it occurs.
Currently, it is simply impossible to predict an earthquake.
Essential help could come from Artificial Intelligence, which can automatically recognize the signals that precede an earthquake by analyzing data related to the movement of Earth's cracks.
This advancement comes from "Machine Learning", an unsupervised, automatic method, without predefined instructions, that identifies similarities or structures in data that precede an earthquake.
A team of scientists has extracted all the data from the five most significant earthquakes around the world, earthquakes that were not preceded by any seismic waves.
All this information was processed by the algorithm, which cataloged and classified the data, grouping it into different families.
The system was able to understand which data to work with, which to look for, in short, the data whose changes preceded the oscillations.
This still does not help in predicting earthquake occurrence, but it highlighted which indicators a systematic study of their possible prediction should focus on.
According to the researchers, the study showed that Artificial Intelligence analysis, using highly advanced processing techniques, can improve the understanding of the physical phenomena that drive the evolution of cracks and, most importantly, precede earthquakes.