Google DeepMind's AI Makes Hurricane Forecasts More Accurate

AI has been criticized for the huge energy use of data centers. But it is also making weather forecasts more accurate. The US National Hurricane Center is already using AI models in its daily work, and is working with Google DeepMind to build next-generation tools.

This year saw a record number of hurricanes in the Pacific, and Hurricane Isaias hit the southeastern US. DeepMind's model spotted Isaias as a possible tropical cyclone a full week before the storm formed. It can also create 1,000 possible scenarios for each storm and forecast its path, strength, and rainfall.

A DeepMind study published in Nature shows the model can match old models' accuracy one day earlier. The hurricane center says AI is especially good at tracking fast changes in steering winds, though a single model may still make sudden big changes to its forecasts.

Human experts still need to combine real-time data to make the final call. Meanwhile, the data center boom is raising carbon emissions, and climate change is making hurricanes stronger, slower, and more damaging.