Artificial Intelligence System for Predicting Earthquakes Aftershocks Developed by Google and Harvard

By , in Sci/Tech on . Tagged width: , , ,

Researchers of earthquakes have been working on modeling aftershocks for years. By now, they are capable of forecasting quite precisely when the replicas will take place and how big they may be. But they’ve been a long way from flawless in tracking where these aftershocks will take place.

The obvious solution, however, appears to be the Artificial Intelligence. A team of collaborators from Google and Harvard informed that after coaching a neural network (the same kind of AI that permits Facebook photo tags and Alexa’s voice transcriptions) with a database of over 131,000 earthquakes and the subsequent aftershocks’ locations, they have discovered the most effective method of forecasting where future aftershocks will take place.

This Artificial Intelligence, at its core, is only a sophisticated pattern matching system. The researchers show it locations of earthquakes aftershocks and the AI tries to discover underlying patterns, using various algorithms.

The Artificial Intelligence system for predicting earthquakes aftershocks is based on two complex metrics used in the science of bendable materials

In the case of facial recognition, for example, that pattern would be the positions of the pixels that form the face of individuals. On the other hand, when it comes to earthquakes aftershocks prediction, the Artificial Intelligence developed by researchers from Google and Harvard uses specific equations and algorithms to find a pattern in the aftershocks locations.

The new AI system to predict aftershocks locations was presented yesterday in Nature journal. In the study’s report, the scientists wrote that the AI employed two complex metrics, the maximum shear stress change and von-Mises yield criterion, which both had not previously been considered to be linked with aftershocks. Both these metrics are usually used in the sciences of bendable materials, until now when they might also be used in aftershocks predictions.

“We’re quite far away from having this be useful in any operational sense at all. We view this as a very motivating first step,” said Harvard researcher Phoebe DeVries, the study’s co-author.