ECMWF makes a machine-learned forecast model operational

node

The European Centre for Medium-Range Weather Forecasts brought its Artificial Intelligence Forecasting System into operational service, reporting gains of up to 20% over the physics-based system on several measures including tropical cyclone track error. An intergovernmental agency whose forecasts are used for aviation, shipping and emergency planning thereby accepted operational accountability for learned-model output. It did not switch the physics-based system off: the numerical model still produces the reanalysis the learned model is trained on, so the two are complements and the learned model cannot outrun its own training source.

A node is not a place. Drawing it on a map would assert something about the world that no stored fact supports.