""Philosophically, scientifically, it's a return to simplicity," Mike Pritchard, director of climate simulation at Nvidia, told reporters on a call before the meeting. "We're moving away from hand-tailored niche AI architectures and leaning into the future of simple, scalable, transformer architectures.""
""Because this model is trained directly on globally available geostationary satellite observations, rather than region-specific physics model outputs, Nowcasting's approach can be adapted anywhere on the planet"
"Nvidia claims that one model in particular, Earth-2 Medium Range, beats Google DeepMind's AI weather model, GenCast, on more than 70 variables."
Weather forecasts varied widely ahead of a major U.S. winter storm. Nvidia released Earth-2 AI forecasting models, including Earth-2 Medium Range, Nowcasting, and Global Data Assimilation. Nvidia claims Earth-2 Medium Range outperforms Google DeepMind's GenCast on more than 70 variables. GenCast was released in December 2024 and had been significantly more accurate than existing models capable of forecasts up to 15 days. Earth-2 is built on a new Atlas architecture and emphasizes simple, scalable transformer architectures instead of hand-tailored niche AI designs. Nowcasting uses globally available geostationary satellite observations to produce zero-to-six-hour forecasts adaptable anywhere on the planet. The suite aims to make forecasting faster and more accurate.
Read at TechCrunch
Unable to calculate read time
Collection
[
|
...
]