伝統的な予測は依然として極端な天候のためのAIを打ち負かす

The study pitted some of the leading AI weather models against a database of recent extreme events and found that they often fail to predict record-breaking heat, wind, or cold. This is because they're trained using decades of past data, which doesn't account for extreme events that haven't been observed before. Traditional physics-based forecasting, on the other hand, uses complex mathematical models to represent the physical world and can more readily adapt to new conditions. While AI models can outperform traditional models for typical weather forecasting or extreme weather that isn't wildly outside the range of past events, they still struggle to predict the most extreme events. Researchers are exploring ways to improve the accuracy of forecasting the most extreme of extreme weather, including adding data to training sets that shows what record-breaking events could look like. As new models roll out, it's important to undergo independent evaluation and benchmarking to ensure they're accurate and reliable.

Source: Traditional forecasting still beats AI for the most extreme weather