Artificial intelligence may reveal the critical point of climate change.

Chris Bauch, a professor of applied mathematics at the University of Waterloo, is the co-author of a recent research paper, which reports the results of a new deep learning algorithm. Bauch said that the focus of this study is the threshold of rapid or irreversible changes in the system.

"We found that the new algorithm can not only predict the critical point more accurately than the existing methods, but also provide the information of state types beyond the critical point," Bauch said. "Many of these tipping points are unwelcome, and we hope to stop them if possible."

Some critical points usually associated with uncontrolled climate change include the melting of permafrost in the Arctic, which may release a large amount of methane and stimulate further rapid warming; The ocean current system collapses, which may cause the weather pattern to change almost immediately; Or the disintegration of the ice sheet may lead to rapid changes in sea level.

According to the researchers, the innovative method of this artificial intelligence is that it is programmed to learn not only one type of critical point, but also the characteristics of general critical points.

This method gains strength from the mixture of artificial intelligence and critical point mathematical theory, which is more than any other single method. After training AI in what they described as a "possible critical point universe" (including about 500,000 models), researchers tested it on specific real-world critical points (including historical climate core samples) in various systems.

"When we approach the critical point of danger, our improved method may trigger a red flag," said Timothy Lenton, director of the Institute of Global Systems at the University of Exeter and one of the co-authors of the study. "Providing better warning of the critical point of climate can help society adapt and reduce their vulnerability to the coming events, even if they cannot avoid it."

Deep learning has made great progress in pattern recognition and classification, and researchers have transformed critical point detection into pattern recognition for the first time. This is to try to detect the patterns that appear before the critical point, so that the machine learning algorithm can judge whether the critical point is coming.

"People are familiar with the critical point of the climate system, but there are critical points in ecology, epidemiology and even the stock market," said Thomas Bury, a postdoctoral researcher at McGill University and another co-author of the paper. "We have learned that artificial intelligence is very good at detecting the critical point characteristics of various complex systems."

Madhur Anand, another researcher of the project and director of Guelph Environmental Research Institute, said that the new deep learning algorithm is "a changer in the ability to predict major changes, including those related to climate change".

Now that their AI has understood how the critical point works, the team is working on the next stage, that is, providing them with data on contemporary climate change trends. But Anand warned what might happen with this knowledge.

"This definitely gives us an advantage," she said. "But of course, in terms of how we use this knowledge, it depends on human beings. I only hope that these new discoveries will bring about fair and positive changes. "