The data to be translated: Google DeepMind's team has developed the machine learning weather forecasting model GraphCast, which, it is claimed, outperforms existing traditional numerical prediction systems in over 90% of cases, being faster, more effective, and more energy-efficient. Unlike running meteorological simulations, GraphCast relies on historical data for predictions. It forecasts the weather every six hours and can predict conditions up to ten days in advance, demonstrating strong performance in forecasting extreme weather events. The study's authors believe that GraphCast can serve as a complementary tool to meteorologists' existing arsenal and holds revolutionary implications for the field of weather forecasting.