Google built a flash-flood prediction tool using Gemini and old news reports
Google has developed Groundsource, a new AI-powered methodology utilizing its Gemini language model to predict flash floods. The system processes millions of global news reports to create a dataset of flood events, which is then used to train a model that forecasts flood likelihood based on current weather data. While it has limitations in precision and area coverage compared to some existing systems, it aims to assist areas lacking advanced weather-sensing infrastructure and has potential applications for predicting other natural phenomena.
Engadget
- Entities
- Flash Flood Prediction
- Artificial Intelligence
- Google Research
- Gemini
- DeepMind
- US National Weather Service
Original source
This article was reported and published by Engadget. feed.gg links to it as part of a story cluster — full text, images and rights remain with the publisher.