A new study published in Science demonstrates that an artificial intelligence-based weather forecasting model has achieved superior accuracy compared to conventional physics-based models. The AI system, developed by researchers at a leading tech institute, was trained on decades of historical weather data and can generate high-resolution global forecasts in minutes, a task that typically takes hours for traditional supercomputers. In a head-to-head comparison over a year, the AI model more accurately predicted key atmospheric variables like temperature, pressure, and humidity at various altitudes. While experts acknowledge the breakthrough, they note that integrating AI’s probabilistic strengths with the deterministic understanding of physical models will be crucial for future operational use. The findings suggest a significant shift toward hybrid forecasting systems that could improve early warnings for extreme weather events. Read the full article at https://sciencedaily.com/releases/2024/05/240508113104.htm.