This week, Nvidia announced plans to build the world’s most powerful artificial intelligence supercomputer for predicting climate change.
Now the head of the company, Jen-Hsun Huang, has spoken in more detail about the project on the Nvidia blog. It is noted that the Earth-2 system, or E-2, is designed to create a digital twin of the Earth on the Omniverse platform.
As Jen-Hsun Huang points out, climate simulations can now be tuned with a resolution of 10 to 100 kilometers, but more resolution is needed to model changes in the global water cycle – the movement of water in oceans, sea ice, land surface and groundwater through the atmosphere and clouds. Changes in this system are leading to increased storms and droughts.
Jen-Hsun Huang notes:
To simulate clouds reflecting sunlight back into space, you need meter resolution. Scientists estimate that these permissions will require millions and billions of times more computing power than what is currently available. It will take decades to achieve this with the usual dynamics of computational evolution (acceleration 10 times every five years). “
To achieve “millions of times faster speeds,” Nvidia plans to combine GPU-accelerated computing, deep learning and neural networks, and AI-powered supercomputers, along with massive amounts of observational and synthetic data for training.
The new system will be analogous to Cambridge-1, the world’s most powerful artificial intelligence supercomputer for medical research, which Nvidia unveiled in the UK this summer.
As a result, countries, cities and towns will be able to receive early warnings to adapt and improve infrastructure resilience.