The growth of artificial intelligence (AI), particularly ChatGPT, has led many people to wonder why AI models are now capable of composing poetry and producing artwork, but we humans are still stuck with boring tasks like cleaning floors and washing dishes. The displeasure expressed by users is due to the way AI technology has been developed and commercialized rather than any deliberate attempt by academics.
However, it is important to recognize that many AI developers are working hard to make this technology more practical and helpful. The Huawei Cloud team recently made a breakthrough by publishing an AI model in the journal Nature that focuses on a very practical problem: weather forecasting.
One of the most challenging issues in weather forecasting is the unpredictability of atmospheric systems. While current numerical weather prediction (NWP) approaches have made significant progress in short-term forecasting, medium- and long-term forecasts remain limited. NWP models require large computational resources and specific initial conditions, which can lead to errors.
The Huawei Cloud team solved these problems with their Pangu weather model, published in Nature on 5 July 2023. Their AI-based medium-term weather forecast achieved “second-level” global weather forecasting accuracy, which is 10,000 times faster than traditional numerical forecasting.
The HUAWEI CLOUD Pangu large-scale meteorological model is built around a 3D deep neural network (3D Earth-Specific Transformer) trained with 40 years of global weather data. The model’s innovative technique combines height information into a new dimension, allowing for a better understanding of three-dimensional weather patterns and more accurate forecasts. A hierarchical temporal aggregation technique also minimizes errors in medium-term forecasts.
Experts have hailed Huawei’s study as a significant achievement in AI weather forecasting. The Pangu model’s predictions have outperformed actual observations in a variety of meteorological research settings.
While the Pangu model is an important step forward, it still relies on NWP for training and hasn’t completely overtaken it. Experts recommend further study and improvements to the huge Pangu meteorological model.
Accurate weather forecasts are critical for agriculture, aviation, energy and disaster preparedness. The success of AI-driven weather forecasting, such as the Pangu model, can lead to better agricultural planning, more effective air traffic routing, and lower airline operating costs.
Finally, AI technology, as demonstrated by Huawei Cloud’s Pangu weather model, has the potential to improve weather forecasting while addressing the problems faced by traditional techniques. Continued development and research in this area could lead to increasingly accurate and lucrative weather forecasting with far-reaching social and economic implications.
