Digital Desk: An AI earthquake predictor has achieved 70% accuracy in a groundbreaking trial in China. The AI tool, developed by researchers at the University of Texas at Austin, was trained on a massive dataset of seismic data and earthquake records. It was then able to successfully predict 14 out of 20 earthquakes that occurred during a seven-month trial in China.
This is the first time that an AI tool has been able to predict earthquakes with such a high degree of accuracy. It is a significant breakthrough in the field of earthquake prediction, and it has the potential to save lives and reduce the damage caused by earthquakes.
During the trial, the AI successfully predicted 14 earthquakes within a 200-mile radius of their anticipated locations, matching their expected strength with remarkable precision. However, it did miss one earthquake and issued eight false alarms, according to scientists from The University of Texas (UT) at Austin in the United States.
The AI tool’s training involved identifying statistical anomalies in real-time seismic data, closely correlated with historical earthquake occurrences. The approach utilised a relatively straightforward machine learning technique. The research findings were published in the Bulletin of the Seismological Society of America.
Researchers provided the AI with statistical features based on their understanding of earthquake physics, instructing it to learn from a five-year database of seismic recordings. Once trained, the AI made forecasts by detecting subtle indications of impending earthquakes amidst the Earth’s background seismic activity.
Although the applicability of this approach in other regions remains uncertain, the researchers express confidence that, in areas with robust seismic monitoring networks like California, Italy, Japan, Greece, Turkey, and Texas, the AI’s success rate could be further enhanced, enabling predictions accurate within a few tens of miles.
This achievement signifies a significant milestone in the field of AI-driven earthquake forecasting, as noted by the research team. Sergey Fomel, a professor at UT’s Bureau of Economic Geology, described earthquake prediction as the “holy grail” and believes that this advancement demonstrates the potential for solving what was previously considered an insurmountable problem.
Alexandros Savvaidis, a senior research scientist leading the bureau’s Texas Seismological Network Program (TexNet), emphasised the value of even a 70 percent accuracy rate, stating that it has the potential to minimize economic and human losses and significantly enhance global earthquake preparedness.
The trial, in which UT’s AI outperformed 600 other designs, was part of an international competition hosted in China. The project was led by Yangkang Chen, a seismologist from the bureau and the AI’s primary developer. Chen expressed the team’s future goal of combining physics and data-driven methods to create a generalized tool applicable worldwide, similar to ChatGPT.