Scientists from the University of California, Los Angeles, investigated how quickly AI makes weather forecasts and found that it surpassed meteorologists.

The GraphCast program from DeepMind researchers outperformed the traditional European weather forecasting system by 99%. This was investigated during 12,000 measurements.

The model makes accurate forecasts in a few minutes, while the most powerful traditional meteorological instruments work for hours, the Nature article writes.

In particular, the calculations show exactly how heat, air and water vapor move in the atmosphere. However, scientists warn that such AI-based tools are expensive and energy-intensive.

“Standard physical models are still needed to derive global weather estimates, which are initially used to train AI models,” the researchers say. “We anticipate that it will be another two to five years before humans can use machine learning-based predictions to make real-world decisions.”