Google has introduced an upgraded AI weather model, WeatherNext 2, set to enhance its popular products such as Search, Gemini, and Pixel phones. The new AI-enhanced forecasts have exhibited high accuracy levels compared to traditional physics-based models, enabling faster and more efficient predictions. Previously largely experimental, Google is now leveraging its AI weather predictions as a key feature across its product lineup.
The latest AI model, WeatherNext 2, can deliver forecasts eight times quicker than its predecessor while boasting a remarkable 99.9% accuracy rate in predicting variables like temperature and wind. Leveraging Google’s TPU chips, WeatherNext 2 can rapidly generate multiple forecast scenarios, a task that would typically require hours on a supercomputer using physics-based models.
Google’s approach involves utilizing a Functional Generative Network (FGN) in WeatherNext 2, streamlining the forecasting process. Unlike physics-based models that aim to replicate atmospheric physics, AI models like WeatherNext 2 analyze historical data patterns for predictive insights.
Source: The Verge