Google Unveils WeatherNext 3: A Breakthrough in AI-Powered Weather Forecasting

Google's latest AI weather model, WeatherNext 3, is poised to revolutionize the way we forecast the weather. Developed by Google DeepMind and Google Research, this new model has proven to be the most accurate among leading contenders in a recent test on Operational WeatherBench.

The model uses deep learning techniques to analyze data from government-owned supercomputers, which have become remarkably accurate but are expensive and slow. By leveraging raw empirical observations, such as weather satellite data collected in real-time, the model can provide more accurate forecasts with comparable speed.

WeatherNext 3 tackles three key weaknesses of AI forecasting models: predicting over a wide area, struggles with rain forecasting, and dependence on formatted datasets from government agencies. The new model can predict down to a resolution of 5 km, improves rain forecasting by 60%, and produces hourly forecasts instead of every six hours.

Google plans to feed WeatherNext 3 into its weather information products, including search results, Google Maps, and Gemini, as well as making it available on its cloud platforms. This will enable users and researchers to access the model's predictions more easily.