Google has introduced a new flash flood forecasting model that leverages its Gemini large language model to analyze 5 million global news articles and has created a geographically tagged dataset called "Groundsource." The dataset trained an LSTM neural network to predict flood probability, which is currently being deployed across 150 countries/regions through Google's Flood Hub. Although the model has lower resolution than systems like the National Weather Service and does not use local radar, it was specifically designed for regions lacking expensive infrastructure. Officials who tested the system reported shortened response times, and Google hopes this approach can also be applied to other natural phenomena, such as heat waves.