Recently, Google's DeepMind research team introduced a new AI model called GenCast, designed to improve weather forecasting.
According to the paper published in the journal Nature, the researchers said the model demonstrated superior performance compared to the world's leading weather forecasting system - the ENS (Ensemble Prediction System) of the European Centre for Medium-Range Weather Forecasts (ECMWF).
GenCast is built with advanced forecasting capabilities that leverage deep learning and complex data processing. Unlike traditional forecasting models that are “deterministic and provide a single best estimate of future weather,” GenCast creates “an ensemble of 50 or more predictions,” each representing a possible weather pattern.
This allows the model to generate “multiple future weather scenarios,” improving the accuracy of its forecasts.
The DeepMind team said they trained GenCast using weather data collected up to 2018. When comparing GenCast's forecasts with real-world data from 2019, the results showed that this AI model was more accurate than ENS in 97.2% of the tested cases.
Google also revealed that GenCast is now part of a suite of AI-based savings tools the company is integrating into products like Google Search and Google Maps.
Google plans to release real-time reports and historical data from GenCast, allowing people to use them in research and build their own models.
The deployment of GenCast marks a major step forward for Google’s AI-powered, detailed real-time forecasting technology. With its high accuracy and ability to generate complex predictions across a wide range of conditions, the model has the potential to be a game-changer in global weather forecasting.
