Google Develops AI System Using News Data to Forecast Urban Flash Floods

Google has developed Groundsource, an AI-driven flood forecasting system that uses Gemini, a large language model, to analyze over 5 million news articles and identify 2.6 million historical flood events across 150 countries. This structured dataset addresses the challenge of predicting rapid, localized urban flash floods, which traditional monitoring struggles to track. The data supports a neural network model integrated into Google's Flood Hub, aiding early warnings and emergency responses, though some prediction limitations remain.