As Tropical Storm 'Melissa' lingered south of Haiti, Philip Papin, the chief forecaster at the U.S. National Hurricane Center (NHC), relied on Google DeepMind’s hurricane-specific AI model to make a bold and precise forecast. He predicted that the storm would rapidly escalate into a Category 4 hurricane within 24 hours and make a direct path toward Jamaica’s coast. His prediction proved remarkably accurate, as 'Melissa' ultimately struck Jamaica as a Category 5 hurricane.

NHC forecasters are increasingly turning to this AI model—the first of its kind designed specifically for hurricane forecasting—which has outperformed traditional numerical weather prediction systems. This year, the model demonstrated superior accuracy in tracking the paths of 13 Atlantic tropical storms, offering rapid computation and cost efficiency. However, it’s important to note that the model is based on machine learning techniques rather than generative AI.

While it leads in global hurricane path predictions, the model occasionally shows limitations in forecasting extreme intensity levels, as seen in its projections for Hurricane 'Irene' and Typhoon 'Kamagi'. James Franklin, a retired senior forecaster at the NHC, intends to collaborate with Google to enhance the model’s usability for forecasters. He notes that many view the model’s outputs as a 'black box', lacking transparency in its decision-making process. Meanwhile, Google maintains confidentiality around its core technology, while the U.S. and European governments are also actively developing their own AI-driven meteorological models.