A new artificial intelligence weather forecasting model developed by researchers at Google DeepMind has demonstrated superior performance compared to traditional physics-based models in global medium-range weather predictions. The AI system, named GraphCast, uses machine learning to analyze decades of historical weather data and identify complex atmospheric patterns, enabling it to generate accurate 10-day forecasts in under one minute on a single Google TPU. In a comprehensive evaluation against the European Centre for Medium-Range Weather Forecasts’ high-resolution forecast system, GraphCast provided more accurate predictions for over 90% of 1,380 test variables, including critical metrics like tropical cyclone tracking and extreme temperature forecasts. Researchers emphasize that the AI model is designed to complement rather than replace existing systems, offering a faster, more computationally efficient alternative that could enhance early warning systems for severe weather events. The model’s architecture processes weather data as a graph, allowing it to capture complex spatial relationships across the globe. While promising, experts note that further testing is required to fully assess the model’s reliability across all weather scenarios and its integration into operational forecasting workflows. Read the full article at https://technologyreview.com/2023/11/14/1084089/ai-weather-model-graphcast-outperforms-traditional-forecasts/.