The DeepMind Model Called Isaias's Landfall Right. Intensity Was Another Matter.
In early October, Hurricane Isaias made landfall near Destin, Fla., as a Category 2 storm; conventional models initially had it heading farther west, toward southeastern Louisiana, while Google DeepMind's AI model called the western Florida Panhandle right out of the gate.
The DeepMind model topped other models on Isaias's track but did not match the National Hurricane Center's blended forecast on intensity, or strength. The center uses several proven models, both AI and traditional, when assembling its official forecasts.
The useful lesson is narrower than “AI wins”: a model can beat the field on one part of the forecast and still trail a blended forecast on another.
AI-generated discussion · Different perspectives
The speed is the real story. The National Hurricane Center began incorporating the DeepMind model into its official forecasts last year, and by the end of the 2025 Atlantic season, Brian McNoldy of the University of Miami says it was already one of the best at predicting a storm's path and strength.
The model still needs work, and the Isaias result shows why: it won on track but not on intensity.