AI Climate Modeling: Promise, Power, and Proof Gaps

Compute trade-offs span training, inference hardware, energy, data, and operations.

The research brief names GraphCast, GenCast, Pangu-Weather, FourCastNet, and NeuralGCM as notable systems.

The findings identify hybrid forecasting as a way to blend climate predictions with AI models.

The supplied findings name the projects but do not provide enough detail to explain each system’s specific function.