In the AI of the hurricane

Artificial intelligence models are getting better at predicting the paths of hurricanes.

Earlier this year, a model called GraphCast from Google’s DeepMind AI lab was able to accurately predict the path of Hurricane Beryl, forecasting that it’d take a turn away from Mexico and hit Texas. That prediction was correct — and came a week earlier than conventional meteorological models. Meanwhile, the European Centre for Medium-Range Weather Forecasts’ own AI model correctly predicted that Hurricane Francine would hit Louisiana before traditional models. These models can process large amounts of historical data to make their forecasts, while other models such as Nvidia’s FourCastNet work by modeling the entire Earth.

Still, while AI weather models appear good at tracking the direction of hurricanes, they struggle with determining other factors such as the intensity of the storms. So, at least in the near term, these models will work alongside human modelers rather than replace them.