AI Struggles To Predict Extreme Weather

New research published in Science Advances indicates that physics-based models are performing better than AI-driven models when forecasting record-breaking extreme weather events. The findings suggest AI may struggle with high-impact meteorological events. Artificial intelligence is not performing better than existing alternatives when it comes to forecasting extreme weather. The finding suggests AI models may not provide superior accuracy for predicting severe meteorological events. A Swiss study published in Science Advances found AI models are not more accurate than alternatives for extreme weather like wind storms and temperature extremes. The study shows that AI performance declines as weather conditions become more extreme. This occurs despite supercomputers performing increasingly complex weather simulations. The findings suggest AI cannot yet provide the millimeter and second-level precision expected for upcoming days. Supercomputers are currently used to run increasingly complex and precise weather simulations. While supercomputers are increasingly using complex simulations and AI to improve accuracy, the technology currently fails to provide millimeter or second-level precision for extreme conditions. Traditional meteorological models maintain an advantage because they incorporate known laws of atmospheric physics. In contrast, AI models rely on identifying patterns within training data.