The weather forecasting AI algorithm developed by Google’s DeepMind, known as GraphCast, has outperformed the European Centre for Medium-Range Weather Forecasts (ECMWF) model on more than 90 per cent of 1,380 metrics, including temperature, pressure, wind speed and direction, and humidity at different levels of the atmosphere.

The machine-learning model takes less than a minute to make 10-day weather forecasts worldwide on a desktop computer, and is more precise than other approaches.

“GraphCast currently is leading the race among the AI models,” said computer scientist Aditya Grover at the University of California, Los Angeles.

Traditional weather forecasts use numerical weather prediction (NWP), which uses mathematical models based on physical principles. These tools rely on data from thousands of weather stations at different levels of the atmosphere around the globe. For this reason, they require vast amounts of computing power. 

In contrast, GraphCast...