An artificial intelligence algorithm has predicted that crossing the critical threshold of global warming will occur sooner than previously thought. The finding comes from a study by Colorado State University and Stanford University, Gizmodo reports.
Researchers used a neural network trained on the outputs of climate-model data to forecast when the critical thresholds of global warming would be reached. The AI found that the Earth will be about 1.5°C warmer in roughly ten years. By the end of the century, the planet will warm by 2°C.
According to the Paris Agreement, global average temperatures should not rise by more than 1.5°C by 2050.
“We began with a particular interest in this threshold of global warming. Because of the Paris Agreement, there has been a lot of discussion and research about when we would reach 1.5°C”, the researchers said.
The team trained the neural network on the same climate-modeling dataset used by other organisations. The algorithm analysed historical temperature observations from around the world and considered low-, medium-, and high-emission scenarios.
The AI concluded that in any scenario, the planet would reach 1.5°C of warming between 2033 and 2035.
The algorithm also found that even under a low-emission scenario, the annual average temperature would rise by 2°C by 2054.
Researchers say the AI has revised the IPCC forecast. Previously the IPCC had projected a roughly 2°C increase in mean global temperatures under a zero-emission scenario for CO2.
“In the same scenario, the AI projects a higher likelihood of … . We are not asserting that the algorithm is right or that the IPCC is wrong, but this is the largest area of disagreement between the neural-network forecasts and the IPCC”, the researchers said.
Co-author Noah Diffenbaugh said that while the faster-warming projections are troubling, governments and agencies remain determined to avert a tipping point. He credited the Paris Agreement with pushing policymakers to accelerate emissions reductions.
“If you look at the IPCC’s 2014 report, the 21st century’s warming could have ended up at 4°C rather than 2°C had policymakers acted,” the scientist said.
Diffenbaugh added that the current trajectory is far less warming than it would have been before the Paris Agreement.
In October 2022, the AI helped hurricane victims Ian and Fiona secure monetary compensation.
In July, Ohio State University scientists trained the neural network to predict the extent of building damage after a storm.
Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the AI world!