A researcher at Harvard’s astronomy department, Phoebe DeVries, proposed to partner with Google to use artificial intelligence for earthquake prediction and an article with the results has already been published.
According to the study reported on nature.com, earthquakes usually occur sequentially, starting with an initial tremor and ending with several aftershocks of lesser magnitude (usually) compared to the main quake.
While predicting these aftershocks is not an easy task, with Google’s machine learning experts they have used in-depth engineering to determine where the aftershocks could occur and trained the system with information from more than 118 major earthquakes around the world.
Using the information, they have created a neural network that analyses the relationship between the initial tremor and the aftershocks, creating a much more complete model of the aftershock locations that could help deploy emergency services and evacuate areas threatened by aftershocks.
The article contains a map of the replication probability distribution of the Landers earthquake, which we show in the upper screenshot where the dark red regions are most likely to feel replicated. While the black areas are real replicas, the yellow line shows the faults that were breached during the main tremor.
In this case, artificial intelligence can help a lot and save many lives.