Nvidia A new one has been announced Generated AI The model simulates the Earth’s climate at a new level of detail, a major milestone in the use of AI for climate science.
Called cbottle (climate in bottles), the model is the first similar model to simulate global atmospheric conditions at a resolution of kilometers in a resolution of kilometers. Digital Twins For climate scientists.
Cbottle is part of NVIDIA’s Earth-2 platform, providing scientists with an interactive tool to better predict and mitigate the impact of climate change.
Over the past 50 years, the model has been trained in atmospheric states and can generate climate models in response to specific inputs such as time of day, time of year, and sea temperature.
While typical climate models require high levels of computational energy and time, Cbottle uses generated AI and GPUs to produce high-resolution climate predictions in very few times and has higher energy efficiency.
“The model can be filled in missing or damaged climate data, correct biased climate models, super-resolution low-resolution climate data, and comprehensive information based on patterns and previous observations,” Nvidia said in a statement at the press conference. “Cbottle’s extreme data efficiency can be trained in just four weeks of climate simulation.”
According to NVIDIA, a single weather sample can reduce the PB of simulated data by 3,000 times, allowing large amounts of data compression without compromising accuracy.
“In the face of a rapidly changing climate, the latest advances in Earth-2 represent a shift in our ability to understand, predict and adapt to the world around us,” said Bjorn Stevens, director of the Max Planck Institute for Meteorology.
“By leveraging NVIDIA’s advanced AI and accelerated computing, we are building a digital twin of the planet – marking a new era when climate science becomes accessible and actionable, enabling informed decisions that preserve our collective future.”
The model has been field tested during the Global KM-scale Hackathon organized by the Global Climate Research Program, which covers 10 research centers in eight countries.
Currently, scientific institutions including the Max-Planck-Institute and the Allen AI Institute (AI2) are currently working with Cbottle to transform Earth observation and climate data into “queried and interactive” AI systems.
“Cbottle is an elegant use for generating AI and an exciting new resource for effectively simulating local extreme weather such as flooding or hot winds spreading wildfires,” said Christopher Bretherton, senior director of climate modeling at AI2.
The Cbottle Foundation model is now available for early entry into climate AI researchers and can be accessed via preprints on GitHub and Arxiv.
This article first appeared in a sister publication of AI Business Today’s Internet of Things world.
