Microsoft has revealed new details about its new weather AI model, Aurora, which can accurately predict air quality, hurricanes, typhoons and other atmospheric phenomena. The company says the system outperforms traditional weather methods in both speed and accuracy, citing research in the journal Nature.
According to Microsoft, Aurora has been trained on more than a million hours of data from satellites, radar, weather stations, and computer simulations, allowing it to adapt to new conditions and predict specific weather events. For example, in experiments, the model predicted Typhoon Doksuri four days before it actually hit the Philippines, beating experts' predictions.
Aurora also successfully predicted a sandstorm in Iraq in 2022 and outperformed the US National Hurricane Center in predicting the tracks of tropical cyclones. Importantly, this model works much faster than traditional systems, generating forecasts in seconds, while conventional methods require hours of computation on supercomputers.
Microsoft has already integrated Aurora into its MSN Weather app through a custom model. It can generate hourly forecasts and add additional weather parameters, including precipitation and cloud movement.