The predictions made by Aurora are of similar quality to those made by conventional models, which are used by policymakers to track air pollution and protect against related health risks. The model was trained on over a million hours of data from six weather and climate models and can generate a ten-day global weather forecast alongside the air-pollution prediction. Further research will determine whether AI models like Aurora, trained on diverse data sets, perform better than those trained on a single data set.
Key takeaways:
- Microsoft has developed an AI model named Aurora that can accurately predict global weather and air pollution in less than a minute.
- Aurora is the first entirely AI model to generate a global pollution forecast, a task more complex than weather forecasting.
- The AI model requires significantly less computational power than conventional models and can predict levels of six major air pollutants worldwide, spanning five days.
- Aurora's predictions are of similar quality to conventional models and are used by policymakers to track air pollution and protect against related health harms.