Image source:
From Google's AI division and Harvard University, Researchers have created an AI model which can predict the place of aftershocks till one year after a major earthquake. The model was trained at 199 different major earthquake programs, followed by 130,000 aftershocks. The model was found to be more accurate than a previously used device to predict aftershocks.

Phoebe DeVries of the Earth and Planetary Science Department at Harvard University in an interview said that, After changing the model stress to neural network, neural networks can predict the aftershock locations more accurately in test datasets than the type of baseline coulomb failure stress change criterion which is used very much in the study of aftershock locations.

The data used to train the model came from remarkable earthquakes of the past few decades, such as the 2011 Japan earthquake, 2004 Sumatra earthquake, and the 1994 Los Angeles Northridge Earthquake.

This study was written by DeVries with Google ML researcher Martin Wattenberg, Fernando Viegas, and Google AI chief Brendan Meade.
No real earthquake researcher participated in the research, although Devries and Meade consider themselves as computational earth scientists.

Meade said that this model is unable to factor major underground earthquake produced by other natural disasters like volcanic explosions.

At last Meade said that The model was trained and developed using data of past few years major earthquakes but to move forward, the future earthquake will be informed by the data.

For more Trending News on Artificial Intelligence, Stay tuned with us on FoxSplit.