AI Forecasts Beat Experts

Microsoft has created an AI weather model, named Aurora, that surpasses traditional forecasting methods. The AI is capable of more accurately tracking air quality, weather patterns, and tropical storms. Aurora has outperformed seven major forecasting centres in predicting cyclone tracks for the 2022-2023 season. The model was trained on a vast amount of diverse data, leading to greater accuracy, especially in forecasting extreme weather events.

Aurora generates forecasts in seconds using high-bandwidth GPUs, which is approximately 5,000 times faster than traditional systems. While the initial training is costly, the operational expenses are significantly lower. The model accurately predicted all hurricanes in 2023, outperforming forecasting centres like the US National Hurricane Center. Aurora's capabilities extend beyond weather prediction, with potential applications in air quality and ocean dynamics.

Another machine learning system, Aardvark, is also emerging, demonstrating the potential to outperform traditional supercomputer models. These advancements signal a shift in Earth system predictions, offering more accurate forecasts at reduced costs.