DeepMind’s AI model is surprising weather scientists by picking up signals in lower-resolution data to predict storm intensity. While researchers admit the model acts as a "black box," its ability to produce a vast range of potential scenarios helps capture the "butterfly effect" where small deviations lead to major changes.

Last year, the model created 50 scenarios per storm; now, it generates 1,000 scenarios. This level of detail is something current numerical models cannot achieve with existing computing power. While the model is a powerful new tool, experts emphasize it is part of a larger toolbox and that the human element remains critical for translating forecasts into impact assessments.

To foster scientific discovery, Google DeepMind is open-sourcing the Weather Next models. This allows the research community to improve the technology and uncover fresh insights into how cyclones work, potentially using AI to poke into the laws of the universe.