Filters
Filtered by tag: neural networks × Clear
MIT AI forecasts extreme weather without historical data
Matched: AI
MIT engineers have developed an AI tool called η-learning (Extreme Event Aware) that forecasts extreme weather without requiring historical disaster data. Created by graduate student Kai Chang and Professor Themis Sapsis, it generates statistically plausible maps of unprecedented events, including estimated duration, intensity, and affected area. Tested on 25 years of US precipitation data, it can model storms exceeding any recorded levels, helping planners stress-test infrastructure against worst-case scenarios.