Link: Nvidia announces its Earth-2 Medium Range weather model, built on its Atlas architecture, claiming it outperforms Google DeepMind's GenCast in 70+ variables (Tim De Chant/TechCrunch)

Nvidia's new Earth-2 weather forecasting models claim to enhance speed and accuracy in weather predictions, ideally timed with a major U.S. winter storm. These AI-driven models, announced at a recent meeting in Houston, allegedly outperform Google DeepMind's GenCast in forecasting over 70 variables.

The Earth-2 Medium Range model utilizes Nvidia's innovative Atlas architecture, a simplified, scalable approach to weather forecasting, moving away from complex AI systems. This shift was highlighted by Mike Pritchard, Nvidia's director of climate simulation.

Alongside the Medium Range model, Nvidia introduced a Nowcasting model for real-time weather predictions and a Global Data Assimilation model that utilizes data from global weather sources. The latter model processes data quickly on GPUs, vastly reducing the reliance on supercomputers.

The Nowcasting model's training on globally available satellite observations allows it to adapt to various geographic regions, enhancing forecasts worldwide. This could significantly aid regions prone to severe weather in making timely decisions.

These new models complement Nvidia's existing forecasting tools, aimed at democratizing access to sophisticated weather predictions beyond affluent nations and large corporations. According to Pritchard, the models are already being utilized by meteorologists and companies to refine local forecasts.

Nvidia's suite of models represents a critical shift towards scalable weather technologies, offering flexibility for users ranging from governmental bodies to private companies in managing weather-related challenges effectively. #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.