Google WeatherNext Extends Cyclone Forecast Accuracy by 1 Day as AI Gains Ground in Weather
Updated
Updated · The Guardian · Sep 10
Google WeatherNext Extends Cyclone Forecast Accuracy by 1 Day as AI Gains Ground in Weather
3 articles · Updated · The Guardian · Sep 10
Summary
Nature-published results show Google’s WeatherNext can deliver a three-day hurricane or typhoon forecast with the accuracy older systems achieved only at two days.
That extra day comes from AI learning patterns from historical data rather than running the full physics calculations used in numerical weather models.
The approach also cuts computing demands, offering forecasts faster and at lower cost even as conventional models still rely on heavy supercomputer processing.
Weather forecasting has become one of the clearer early wins for AI, with newer systems increasingly matching traditional models—especially on longer-range predictions—and facing relatively little resistance.
Google's AI predicts cyclones faster than supercomputers, but can it accurately foresee unprecedented climate anomalies completely missing from its historical training data?
Would officials risk billion-dollar evacuation orders based entirely on a black-box AI that predicts a devastating hurricane earlier than traditional models?