The algorithms behind generative AI tools like DallE, when combined with physics-based data, might be used to develop higher ways to model the Earth’s climate. Computer scientists in Seattle and San Diego have now used this mix to create a model that’s able to predicting climate patterns over 100 years 25 times faster than the state-of-the-art.
Specifically, the model, called Spherical DYffusion, can project 100 years of climate patterns in 25 hours-a simulation that may take weeks for other models. As well as, existing state-of-the-art models have to run on supercomputers. This model can run on GPU clusters in a research lab.
“Data-driven deep learning models are on the verge of remodeling global weather and climate modeling,” the researchers from the University of California San Diego and the Allen Institute for AI, write.
The research team is presenting their work on the NeurIPS conference 2024, Dec. 9 to fifteen in Vancouver, Canada.
Climate simulations are currently very expensive to generate due to their complexity. Because of this, scientists and policymakers can only run simulations for a limited period of time and consider only limited scenarios.
Certainly one of the researchers’ key insights was that generative AI models, resembling diffusion models, could possibly be used for ensemble climate projections. They combined this with a Spherical Neural Operator, a neural network model designed to work with data on a sphere.
The resulting model starts off with knowledge of climate patterns after which applies a series of transformations based on learned data to predict future patterns.
“Certainly one of the most important benefits over a standard diffusion model (DM) is that our model is way more efficient. It could be possible to generate just as realistic and accurate predictions with conventional DMs but not with such speed,” the researchers write.
Along with running much faster than state-of-the-art, the model can be nearly as accurate without being anywhere near as computationally expensive.
There are some limitations to the model that researchers aim to beat in its next iterations, resembling including more elements of their simulations. Next steps include simulating how the atmosphere responds to CO2.
“We emulated the atmosphere, which is one of the vital vital elements in a climate model,” said Rose Yu, a school member within the UC San Diego Department of Computer Science and Engineering and certainly one of the paper’s senior authors.
The work stems from an internship that certainly one of Yu’s Ph.D. students, Salva Ruhling Cachay, did on the Allen Institute for AI (Ai2).
