DeepMind, Google's AI research team, has introduced a cutting-edge weather prediction model called GenCast, which it claims surpasses the accuracy of the world’s leading forecasting system.
According to a study published in Nature, GenCast outperforms the European Centre for Medium-Range Weather Forecasts’ ENS, regarded as the top operational weather prediction system globally. The researchers determined GenCast’s superiority by training it on pre-2018 weather data and testing it on 2019 forecasts, where it proved more accurate 97.2 percent of the time.
In a blog post, DeepMind explained how GenCast advances weather forecasting. Unlike its earlier deterministic model, which provided a single best estimate, GenCast generates “an ensemble of 50 or more predictions, each representing a possible weather trajectory.” This method creates a "complex probability distribution of future weather scenarios," enhancing the scope and precision of forecasts.
Google is integrating GenCast into its suite of AI-based weather tools, with plans to incorporate its capabilities into Google Search and Maps. Additionally, the company intends to make GenCast’s real-time and historical forecasts publicly available, enabling researchers and developers to leverage the technology for their own models and analyses.
(Source: TechCrunch)
BD-Pratidin English/Mazdud