Microsoft’s Aurora AI predicts weather in seconds using 1M hours of data

Microsoft has launched Aurora, an AI weather forecasting model that the company claims delivers more accurate predictions at significantly lower computational costs than traditional forecasting methods. According to Microsoft’s research, Aurora can predict air quality, ocean waves, tropical cyclone tracks, and high-resolution weather patterns using orders of magnitude less processing power than conventional approaches.

What you should know: Aurora represents a foundational shift in weather prediction technology, trained on over one million hours of diverse meteorological data.

• The model processes information from satellites, radar, weather stations, simulations, and forecasts to generate predictions in seconds rather than hours.
• Microsoft has already integrated Aurora into its MSN Weather service and made the model’s source code and weights publicly available.
• One startup has reportedly used the model to map renewable energy markets, demonstrating its practical applications beyond basic weather forecasting.

How it works: Aurora functions as a foundation model that can be fine-tuned for various downstream weather-related tasks.

• The AI analyzes millions of hours of historical data and makes multiple intertwined calculations simultaneously, rather than processing information separately and combining results.
• This approach enables the model to deliver smaller-resolution and more precise predictions compared to traditional “coarse resolution” forecasts.
• The model can map out air quality, wave height, typhoon patterns, and other weather phenomena with greater geographical precision.

Why this matters: Traditional weather forecasting requires enormous computational resources and time, making it expensive and limiting precision for small-scale geographical areas.

• Current smartphone and TV forecasts are fairly zoomed out, making it difficult to predict how weather events might impact specific locations.
• AI weather models could revolutionize the industry by dramatically reducing costs while improving accuracy, but only if they prove reliable enough to protect communities from weather-related harm.

The big picture: Microsoft claims Aurora “outperforms operational forecasts” in research testing, though the meteorological community remains cautiously optimistic about AI weather models.

• Both traditional and AI-powered forecasts still require review and delivery by human meteorologists, meaning the overall prediction process would likely remain unchanged.
• AI weather models were considered unreliable and impractical just a few years ago, leaving some experts unconvinced despite recent advances.

What experts are saying: The meteorological community acknowledges AI weather forecasting’s potential while maintaining healthy skepticism about current capabilities.

• “We’re all in the hype right now,” Amy McGovern, a computer scientist and meteorologist not involved in Aurora’s development, told The New York Times.
• “AI weather is amazing. But I think there’s still a long way to go,” McGovern added, reflecting broader industry caution about rushing to adopt AI weather models.