Google’s DeepMind tackles weather forecasting, with great performance
Google DeepMind claims that its GenCast system outperforms traditional computing methods of weather forecasting out to at least a week and often beyond. GenCast merges some computational approaches used by atmospheric scientists with a diffusion model, resulting in a system that maintains high resolution while cutting the computational costs significantly. It has relatively low computational requirements, so individual academic research teams will likely be able to experiment with it.