Google DeepMind and Google Research have just unveiled WeatherNext 2, an AI weather-forecasting system designed to deliver significantly faster and more precise global predictions.
According to Google, the new model produces forecasts up to eight times faster than before, with an hourly resolution that offers finer detail about changing conditions.
Weather impacts decision-making in many areas, from agriculture and energy management to travel planning and disaster response. To address this, WeatherNext 2 was developed using a new architecture that can generate hundreds of possible future weather scenarios from a single input. The model relies on independently trained neural networks and introduces carefully calibrated randomness.
Remarkably, each of these scenarios can be computed in under a minute on a single TPU, a pace that would take prestigious physics-based simulators hours to match. n terms of accuracy, WeatherNext 2 reportedly outperforms its predecessor, WeatherNext, across nearly all weather variables and forecast lead times, measuring improvements in temperature, wind, humidity and more, up to 15 days into the future.
The advance comes via a novel technique called a Functional Generative Network or FGN. Rather than simply estimating individual weather variables (known as “marginals”), the FGN structure allows the model to learn the interactions between them (“joints”) so it can forecast large-scale systems, such as heat domes, wind farms or sweeping storm regions.
Google is moving WeatherNext 2 out of the lab and into practical use. Forecast data will be made available through Earth Engine and BigQuery, and a limited early-access program is launching on Google Cloud’s Vertex AI platform for organizations seeking custom model inference.
Beyond consumer tools, WeatherNext 2 is expected to benefit researchers, governments, and businesses dealing with complex weather-sensitive challenges. Google says it is committed to making its advanced forecast tools available to a wider global audience while continuing research on even more powerful models.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
Good stuff! At least some genuine benefits from the AI, we are drowning in the AI slop, wasting energy and water resources in the process.