DeepMind's WeatherNext model achieves breakthrough forecasting cyclones

	DeepMind's WeatherNext model achieves breakthrough forecasting cyclones(deepmind.google)
	445 points by bhavansig 13 days ago \| 130 comments
	tl;dr: DeepMind's WeatherNext model predicts tropical cyclone track, intensity, and wind structure with state-of-the-art accuracy, gaining roughly 24 hours of lead time over prior models—equivalent to a decade of meteorological progress. Notably, it operates at a coarse 28x28km resolution (100x coarser than traditional models), yet outperforms them, using Functional Generative Networks to run 1,000-member ensembles in under a minute on a TPU. Google is open-sourcing the code and weights, including a "mini" version runnable in a free Colab notebook.
	HN Discussion: ↑Problem-specific AI models like this are more interesting than the current LLM hype ↑Excitement about impactful AI applications beyond coding agents ~Noting limitations around uncertainty handling and deterministic vs ensemble forecasting ↓This methodology is not novel; ECMWF already deployed similar approach ↓Predicting major weather events is easy; the real challenges are in fine-grained details