Weather forecasts are only 50% accurate if forecasted up to 10 days. It increases to 90% if the forecast is made for 5 days. It is certain that higher forecasting accuracy is needed to provide enough time to avoid disasters. This can now be done using artificial intelligence (AI) GraphCast AI by Google DeepMind.
When compared to weather forecasts made by the European Center for Medium-range Weather Forecast (ECMWF), GraphCast AI is 90% more accurate for forecasts between 3-10 days. What you can predict is air pressure, wind speed, wind direction and temperature.
What makes GraphCast AI interesting is that it has been trained using 40 years of ECMWF data and the forecast for these 10 days is done by GraphCast AI in just a few minutes using a single Google TPU v4 computer in the cloud. Previously, weather forecasting by ECMWF took several hours using a supercomputer
GraphCast AI is not only more accurate but faster and reduces electricity consumption by up to 1000 times mace to forecast the weather. The results of this study conducted by Google DeepMind have been published in the scientific journal Science.