Earthquakes are a huge nuisance, even in today's age full of technology that can record and warn in advance of future tremors. The epicenter, size and strength of an earthquake can be estimated a long time ago, but what still represents a real problem today are the aftershocks. Although they are not as powerful as earthquakes themselves, they are very unpredictable and often cause even more damage than earthquakes themselves. Scientists are able to detect almost all the attributes that aftershocks have, except for their location. In solving this problem, the company Google lends a helping hand, which in cooperation with Harvard University is developing artificial intelligence for estimating locations associated with aftershocks.
You could be interested in
According to a scientific magazine Nature, in which this method was published, is research using machine learning, which can detect earthquake locations much more accurately than conventional models. The researchers tested this phenomenon on a neural network by comparing more than 131 events that included earthquakes with aftershocks. Compared to the existing Coulomb model, which is used in practice for similar events, the artificial intelligence was significantly better. When compared on a scale of 000 to 0, with 1 being a perfect and error-free estimate, the Coulomb model scored 1 and the AI up to 0,583. According to the research leaders, this result is more than gratifying and they believe that the method will take hold in real practice over time.
"When it comes to earthquakes, it is necessary to know three basic pieces of information. When the earthquake will occur, what its magnitude will be and where it will take place. The important data for our work is information about the size and time scale of individual earthquakes. Our task was to deal with the third unknown in the form of the location," says Professor Brendan Meade from Harvard, who participated in the research and subsequent publication. The excellent result of artificial intelligence ensures the detection of connections in seismological graphs. Due to the variety of relief, energy waves and various other variables, it is quite difficult for the technology in use to determine a given prediction. However, with artificial intelligenceance significantly increase, for which not only the clever heads from Harvard are to blame but also the company Google, which significantly assisted in the research.
Dalibor Tichý would like someone to tell China that it is not worth copying apple : )