In this episode, we dive into Gencast—the diffusion-model weather predictor from Google DeepMind trained on 40 years of ERA5 data. We'll explain how it generates ensembles (not images), outperforms current systems up to 15 days, and forecasts extremes like heat waves and cyclones. Plus, the code and weights are open source, lowering barriers for researchers worldwide and speeding up innovation in meteorology and climate resilience.
Note: This podcast was AI-generated, and sometimes AI can make mistakes. Please double-check any critical information.
Sponsored by Embersilk LLC