As per reports, a group of scientists in China found that an AI-powered tool was 70% accurate in predicting earthquakes a week before the calamity during a seven-month trial. The researchers from the University of Texas at Austin, US, said, “The result was a weekly forecast in which the AI tool successfully predicted 14 earthquakes within approximately 200 miles (320 kilometres) where it calculated they would happen and at almost exact strength”. The AI-based tool, however, missed one earthquake and predicted eight wrong warnings, they added.  

The AI model was trained to spot statistical bumps in real-time seismic data that the scientists had compared with past earthquake experiences. They mentioned adding that the method follows a relatively simple machine learning approach. The study was published in the journal Bulletin of the Seismological Society of America. The AI model was given a few statistical features per the team’s knowledge of earthquake physics and told to train itself on a five-year database of seismic data. The researchers mentioned in their study that the model gave its forecast by analysing the signs of incoming earthquakes among the background rumblings on the earth’s surface.  

Even though the researchers are unsure if the same approach can be applied to other locations, they are confident about places like California, Italy, Japan, Greece, Turkey and Texas, as these places have robust seismic tracking networks. Hence, AI could enhance its success rate and provide precise predictions within a few tens of miles. The researchers added that this study’s findings are considered a milestone in research for AI-driven earthquake forecasting.  

Sergey Fomel, a professor at UT’s Bureau of Economic Geology and a research team member, mentioned, “Predicting earthquakes is the holy grail”. “We are not yet close to making predictions for anywhere in the world, but what we achieved tells us that what we thought was an impossible problem is solvable in principle,” he added.  

Alexandros Savvaidis, a senior research scientist leading the bureau’s Texas Seismological Network Program (TexNet) - the state’s seismic network, said, “You don’t see earthquakes coming.” 

“It’s a matter of milliseconds, and the only thing you can control is how prepared you are. Even with 70 per cent, that’s a huge result and could help minimise economic and human losses and has the potential to improve earthquake preparedness worldwide dramatically,” Savvaidis added. 

The trial was conducted as a part of an international competition which was held in China in which the UT-developed AI came first out of 600 other designs. Yangkang Chen, bureau seismologist and the AI's lead developer led UT's entry. Chen said, “Our future goal is to combine physics and data-driven methods to give us something generalised, like chatGPT, that we can apply to anywhere in the world”.