NVIDIA, the Santa Clara giant specializing in the production of graphics processors, necessarily follows the movement of AI. Its CEO, Jensen Huang, recently declared that he wanted to put this technology to work in the fight against global warming.
This happened during the Berlin summit devoted to the Earth Visualization Engines Initiatives (EVE), a set of technologies allowing the recovery of central data relating to the environment. This with the aim of proposing solutions adapted to the resolution of the problems brought to light on occasion. Jensen Huang underlined the major role that AI and accelerated computing must play in the field of climate research.
The challenges of the climate situation
NVIDIA intends to rely on interactive visualizations, colossal databases and rapid climate simulations to effectively equip decision-makers to face the challenge of climate change. The idea is to use these technologies so that researchers can offer solutions to political actors.
NVIDIA places particular emphasis on accelerating climate simulations. For this, the researchers need precise analysis areas (a few square kilometres) and that this geographical data is coupled with a rapid climate simulation offering a high level of detail at the local scale.
To achieve this acceleration objective, NVIDIA relies on in-house technology, the Grace Hopper Superchip GH200 GPU (not Supercheap, above all, spare us the pun!). This processor was designed specifically for AI to work on large scale on high performance computing. In the long term, NVIDIA’s goal is to connect several of these chips to provide entire networks that act as ecosystems. This novel design will accelerate the work of climatology researchers.
Understanding the climate through the power of AI
Jensen Huang also unveiled an Open Source framework, the NVIDIA Modulus. This little gem of technology can build machine learning models based on the laws of physics. It will be one of the elements of this new research assistance project. Another such tool is FourCastNet, a more classic weather forecasting model that relies on real-time databases.
By combining traditional simulations with these new solutions, it will be possible to establish more detailed long-term forecasts. FourCastNet has already succeeded in accurately predicting the trajectory of Hurricane Harvey in 2017. It therefore theoretically has little to prove in the future jobs entrusted to it.
NVIDIA isn’t just there to flatter the retinas and egos of gamers with 4.5 kilo GPUs. The proof is with this new collaboration that the American company will develop with professionals in the field of climatology. Real humanist and ecological momentum or yet another attempt at greenwashing to make people forget the cryptocurrency farming episode of 2021?
Sources: TechPowerUp, Eve4climate
6