A new study published in Science demonstrates that an artificial intelligence weather forecasting model has achieved superior accuracy compared to the world’s leading traditional physics-based systems. The model, developed by Google’s DeepMind and named GraphCast, significantly outperformed the European Centre for Medium-Range Weather Forecasts (ECMWF) high-resolution forecast (HRES) system—considered the gold standard—in over 90% of 1,380 test variables. GraphCast provides medium-range global weather predictions up to 10 days ahead in under one minute on a single Google TPU v4 machine, a process that takes conventional supercomputers hours. The AI system was trained on decades of historical weather data from ECMWF to learn cause-and-effect relationships in Earth’s weather patterns. While experts note that AI models like GraphCast still rely on traditional models for their initial conditions and are not a complete replacement, they represent a major shift towards faster, more efficient, and highly accurate forecasting. This advancement could improve severe weather warnings and climate change modeling. Read the full article: https://www.science.org/doi/10.1126/science.adn2198