Generating scenarios for extreme events, without extreme data

MIT engineers have created a machine-learning algorithm that generates plausible extreme weather events and worst-case scenarios without relying on historical extreme data. The tool predicts characteristics like duration, intensity, and impact area for rare events such as hurricanes or heat waves, aiding community preparedness.