An artificial-intelligence model from Google DeepMind correctly forecast Hurricane Isaias’ path toward Florida’s western Panhandle, beating traditional models that initially projected the storm would move toward southeastern Louisiana. Isaias made landfall near Destin, Florida, in early October as a Category 2 hurricane. The AI model was less accurate in predicting the storm’s intensity than the National Hurricane Center’s blended forecast, underscoring why forecasters are not handing decisions over to software. The National Hurricane Center began incorporating the Google DeepMind system into its official forecasts last year. Its performance has quickly made it one of the leading tools in the field, according to University of Miami hurricane researcher Brian McNoldy and meteorologist Michael Lowry, who has advised the Federal Emergency Management Agency and the hurricane center. Conventional hurricane models use current atmospheric readings—including temperature, humidity, pressure and wind shear—and apply physics-based equations to estimate a storm’s path and strength. AI systems instead draw heavily on patterns identified in thousands of past storms and atmospheric records. That approach can make AI predictions difficult to explain. Lowry has described the models as a kind of “black box”: forecasters may see a correct answer without knowing why the system reached it, or why it might fail next time. He also cautioned that many AI weather models remain unproven; Google DeepMind’s system and one developed by the European Centre for Medium-Range Weather Forecasts are among the strongest performers so far. The hurricane center uses a mix of established AI and physics-based models when building official forecasts. Director Michael Brennan said storms present different challenges, making a consensus of the best available tools more useful than relying on any single model. Human forecasters issue updates every six hours during active hurricanes and interpret shifts between model runs into practical warnings. Brennan said that continuity matters because residents need clear guidance about hazards, not just a changing set of computer projections. The AI model’s progress is drawing attention partly because it can be updated far more quickly than traditional forecasting systems, which are generally revised no more than annually. Lowry said advances that once might have taken a decade are now arriving within a year. He sees the technology as a boost to meteorology, not a threat to forecasters’ jobs. For coastal communities, the stakes are concrete: better forecasts can help residents and emergency officials prepare, but no single model gets every part of a storm right. Isaias showed both the promise of AI in predicting a hurricane’s route and the continuing need for human expertise in assessing its strength and translating forecasts into action.