Part 4/10:
Approximately five to six years ago, a paradigm shift began when researchers proposed employing artificial intelligence to analyze seismic data. The idea was to leverage machine learning algorithms to sift through the complex noise and detect hidden signals that could foreshadow earthquakes.
Chris Marone, a leading earthquake physicist at Penn State University, and his colleagues started experimenting with laboratory earthquake machines—servohydraulic biaxial deformation devices mimicking fault slip. Traditionally, micro-earthquakes recorded during frictional sliding in these experiments were dismissed as mere noise. However, applying machine learning techniques revealed that what was once discarded actually contained rich, predictive information.