MIT AI forecasts extreme weather without historical data

Engineers at MIT have developed an AI tool that forecasts extreme weather without relying on historical data. Graduate student Kai Chang and Professor Themis Sapsis created this tool, which produces maps of events not recorded in a region's historical data. The tool, named η-learning, uses an algorithm that combines point statistics with spatial maps to estimate the intensity and duration of potential events. Details of this method were published on August 20 in Nature Communications, highlighting the limitations of existing models that rely on past data.