New AI May Anticipate Earthquake Aftershocks


Scientists have tried to create models for aftershocks for decades. The can predict fairly accurately the size and time of aftershocks, but were they are weak is predicting the location of the aftershocks.

It seems that AI is here to solve that issue. A joint team of researchers from Google and Harvard has been training a neural network, similar to those that power Alexa’s voice recognition and Facebook’s picture tagging. More than 131,000 earthquakes and their aftershocks have been analyzed so the AI can efficiently recognize patterns.

Neural networks began as an attempt to recreate a working human brain in order to create autonomous devices and intelligent devices.  In order to work properly function it requires a set of training data from which it can learn relevant information. It can then use the information acquired in order to run complex cross-analysis.

In plain language, artificial intelligence is a master of pattern detection. You can offer any data, from music to aftershock hot zones and it will be able to find the pattern by using an advanced algorithm. In the case of facial recognition, the pattern is composed from the pixels that constitute the pictures. In the case of aftershocks, the AI uses a complex equation in order to identify the locations.

According to the paper which was published recently in the peer-reviewed Nature journal, the algorithm is able to be so accurate because it uses two metrics that have not been associated with aftershocks before. They are called maximum shear stress and the von-Mises yield criterion. The metrics have been used before in the bendable materials industry, for copper and aluminum among others, but this is the first time when they are used in aftershock analysis.

While this is an exciting first step, it will take a while until the AI will be fully operational and usable in actual situations according to one of the researchers.


Recommended For You

Leave a Reply

Your email address will not be published. Required fields are marked *