Nvidia announced plans to build the world’s most powerful artificial intelligence supercomputer, given to predicting climate change. The system is called Earth-2 or E-2, which will create a digital twin of the earth in the Omniverse.
According to the latest report, this system will become Cambridge-1 focusing on the field of climate change research. Earlier this year, Nvidia launched the most powerful artificial intelligence supercomputer in the global medical research field in the UK. It has been adopted by many leading medical companies.
Stay tuned with us via Telegram
Huang Renxun said that in order to formulate the best strategies to mitigate and adapt to climate change, climate models that can predict climate changes in different regions of the world over decades are needed. Different from weather forecasts that mainly simulate atmospheric physics, climate models are based on physics, chemistry, and biology for decades of simulations for the atmosphere, water, ice, land, and human activities.
Currently, the resolution of climate simulation is 10 to 100 kilometers. But to simulate the entire global water cycle from oceans, sea ice, land surface and groundwater to the atmosphere and clouds, higher resolution is required. Simulating the clouds that reflect sunlight back into space requires meter-level resolution.
Moreover, that to achieve this resolution, it will require millions to billions of times more computing power than currently available. If calculated according to the conventional standard of 10 times increase in computing power every 5 years, it will take decades to achieve this goal.
Huang Renxun said that for the first time, NVIDIA has the technology to build an ultra-high-resolution climate model that can predict regional extreme weather changes in the next few decades at the speed of light. By combining GPU-accelerated computing, deep learning, and breakthroughs in neural networks with embedded physical information, as well as the three major technologies of artificial intelligence supercomputers, plus a large amount of observation and model data for learning, it can achieve a million-fold acceleration.
In an interview with the media, Huang Renxun said that Earth-2 was completely independently funded and built by NVIDIA. NVIDIA built it to enable software development in the modern artificial intelligence world. Earth-2 was created to pool the capabilities of NVIDIA researchers and contribute to helping combat climate change.
He revealed that NVIDIA will soon announce the architecture of Earth-2, and the new architecture will enable the company to build the most energy-efficient supercomputer ever. It is a supercomputer designed for Omniverse. If you imagine the earth as physical existence, Earth-2 is the engine of another world.
