Aurora, the foundational AI model developed by Microsoft researchers as a basis for weather forecasting, now predicts air quality at rapid speed and precision. The firm has provided the information in a new blog post.
Aurora has previously been utilized to forecast weather phenomena, including hurricanes, cyclones, but now it has been tuned to make forecasts beyond the bounds of conventional weather forecasting, including air pollution. Microsoft explained this model initially acquires basic knowledge from large and diversified sets of information and subsequently gets refined using small and particular information.
Aurora's source code and model weights are now publicly available. A special version that provides hourly forecasts (such as cloud cover) is now integrated into the MSN Weather app.
Aurora is trained on more than 1 million hours of satellite, radar and weather station data, as well as historical weather simulations and forecasts. Its encoder architecture transforms huge amounts of data from different sources into a standard format, allowing the AI to make accurate predictions.