How AI is changing weather forecasting

What this episode covers

Peter Battaglia, senior director of research at Google DeepMind’s sustainability program, explains how AI models like transformers and graph neural networks are improving weather forecasting by capturing complex spatial dependencies that traditional numerical models miss. He details the shift from coarse global forecasts to high-resolution, application-specific predictions, highlighting the challenge of predicting precipitation and the potential of novel data sources like cell phones and social media. Battaglia discusses the historical context of government weather bureaus like NOAA and ECMWF, and outlines future applications in energy grid management, supply chain planning, and disaster response.