Google DeepMind AI Model Beats Top Global Weather Forecasting System
Google's DeepMind introduces GenCast, an AI weather prediction model that outperforms the leading global forecasting system 97.2% of the time. The tech giant plans to integrate the model into Search and Maps while releasing its data for public research.
Google’s DeepMind team unveils GenCast, a new artificial intelligence model designed for weather prediction. Unlike previous deterministic models that provide a single best estimate, GenCast generates an ensemble of 50 or more predictions to create a complex probability distribution of future weather scenarios.
According to a paper published in Nature, GenCast outperforms the European Centre for Medium-Range Weather Forecasts’ ENS system, which holds the title of the world’s top operational forecasting system. After training on historical data up to 2018, the AI model proves more accurate than ENS for 2019 forecasts 97.2 percent of the time.
Google integrates GenCast into its existing suite of AI-based weather models and rolls the technology out across Google Search and Maps. The company also plans to release both real-time and historical forecasts from GenCast so that researchers and developers can utilize the data in their own projects.