Google’s new tool, Groundsource, aims to enhance flash flood prediction using its Gemini language model to analyze over 5 million global news articles for flood reports. By converting this data into a geo-tagged timeline, it integrates current weather forecasts to assess flash flood risks in specific areas. Although the model’s accuracy remains unverified, early users report it aiding rapid responses to localized weather events. Groundsource currently highlights urban flood risks across 150 countries via Google’s Flood Hub and shares data with emergency response agencies. However, it has limitations, such as only covering a 20-square-kilometer area and lacking integration with local radar data, which can track precipitation in real time. Juliet Rothenberg from Google’s Resilience team expresses aspirations for this technology to predict other weather-related phenomena, demonstrating Google’s innovative approach to leveraging AI for disaster preparedness and resilience.
Source link
