A new artificial intelligence weather forecasting model developed by Google DeepMind has demonstrated superior accuracy and speed compared to conventional physics-based models in a comprehensive global evaluation. The GraphCast system, which uses machine learning trained on decades of historical weather data, significantly outperformed the European Centre for Medium-Range Weather Forecasts (ECMWF) high-resolution forecast (HRES) system on over 90% of 1,380 test variables. Key improvements include more accurate predictions of extreme weather events like tropical cyclones and atmospheric rivers, while generating a 10-day forecast in under a minute on a single Google TPU v4 machine—a process that takes traditional systems hours on supercomputers. Researchers emphasize that AI models like GraphCast should complement rather than replace existing numerical weather prediction systems, creating a new “hybrid” era of forecasting that leverages the strengths of both approaches. The findings were published in the journal Science. Read the full article at https://sciencedaily.com/releases/2023/11/231114143522.htm