The founder and CEO of Nvidia, Jensen Huang, said these things at a conference of climate scientists in Berlin
Artificial intelligence and accelerated computing will allow climate researchers to perform "miracles" that are needed to achieve breakthroughs in the field - so said the founder and CEO of Nvidia, Jensen Huang, at a conference of climate scientists in Berlin.
Huang spoke of the urgent need for three "miracles", as he defined it. The first is the ability to simulate the climate at a fast enough pace, and with high resolution - for example on a scale of 2 square kilometers (compared to dozens of kilometers today). The second - the ability to imitate the laws of physics of climate systems with the help of artificial intelligence, and to predict all possible scenarios at record speed and with high accuracy. The third - the ability to visualize (visualize) all the interactive data on platforms, such as Nvidia Omniverse, and make it accessible to decision makers, policy makers, businesses and researchers.
At the conference, Huang spoke about the Nvidia GH200 Grace Hopper Superchip - a processor for accelerated computing that was developed specifically for very large artificial intelligence tasks and accelerated computing applications. By connecting a large number of such processors together, Nvidia can offer climate researchers very powerful computing systems that enable their complex models to run. "For the software - all these processors connected together look like one huge, exceptionally powerful processor," Huang said.
He also spoke about Nvidia Modulus, an open source platform that allows researchers to train and refine models based on the laws of physics for climate research, about FourCastNet - a unique climate prediction model, and how models based on artificial intelligence can learn the laws of physics from real data collected in the field. The model was able, for example, to accurately predict the path of a hurricane by weighting raw data, and predict an unprecedented heat wave in North Africa three weeks before it occurred.
According to Huang, such technologies from Nvidia help to make this unique knowledge more accessible to researchers, and accelerate the use of digital twins - very complex models that simulate systems in the real world, and not only of climate but also, for example, warehouses of retail chains such as Amazon or predictions of absorption levels of G5 networks in dense cities.
Nvidia is today building more powerful systems for training artificial intelligence models, simulating physical problems and interactive visualization to address these grand challenges, Huang said. "These new supercomputers are gradually connecting to the network. This is the most innovative computing technology imaginable," he added.
More of the topic in Hayadan: