ECNETNews reports that a groundbreaking Google AI tool, NeuralCGM, is set to revolutionize climate prediction. Developed by Google Research, this innovative tool employs physics-based modeling combined with artificial intelligence to deliver rapid and precise simulations of the Earth’s atmosphere.
Unlike traditional climate models that divide the planet into large, pixelated sections, making it challenging to accurately forecast localized phenomena such as clouds and turbulence, NeuralCGM merges advanced physics with AI to tackle these challenges effectively. In a 2020 evaluation, NeuralCGM outperformed its non-AI counterpart, X-SHiELD, predicting annual temperature and humidity levels with 15-50% greater accuracy and significantly faster—executing predictions in just 8 minutes compared to 20 days for X-SHiELD.
NeuralCGM is designed to function on a single AI chip, known as a TPU (Tensor Processing Unit), which contrasts sharply with many high-resolution atmospheric models that require access to costly supercomputers and numerous CPUs (Central Processing Units). This portability means that researchers around the globe will have easier access to cutting-edge climate modeling, while the improved efficiency may also lead to lower energy consumption.
This development follows Google’s recent environmental report, which revealed that the company’s reliance on energy-intensive AI technologies over the last five years has resulted in nearly a 50% increase in emissions.
Topics
Climate Technology
AI Innovations
