Waterloo Region Record

Team using AI to predict tipping points

Climate change, cancer and world economies among scenarios tech can help as early warning system, researchers say

ROBERT WILLIAMS Robert Williams is a Waterloo Region-based reporter for The Record. Reach him via email: robertwilliams@torstar.ca

WATERLOO — Imagine knowing when something terrible was going to happen before it happened.

A way of understanding the “tipping points” that can lead to action to divert problems, or in cases when it’s too late to reverse, allow the people most impacted the opportunity to prepare.

Now picture this being applied to some of the world’s biggest systems — climate change, domestic and global economies or disease states in the human body.

That’s the aim of a new multidisciplinary study published in the Proceedings of the National Academy of Sciences of the United States of America, where a team or researchers from the University of Waterloo and beyond have developed a deep learning system — or artificial intelligence — to provide early warning signs of tipping points in real-world systems.

“Our basic insight was to realize that because all kinds of complex systems behave in similar ways close to tipping points no matter what type of system it is, we can use that insight from mathematic theory to improve the ability of the AI algorithms to detect tipping points,” said Chris Bauch, a professor of applied mathematics at the University of Waterloo and co-author of the study.

In the realm of climate change action, the discovery of these tipping points could help governments and other international bodies make better decisions on crisis-mitigation before they happen.

To find these warning signs for climate change, the team can use specific data sets — the more data they can run through the system, the more accurate the result — on topics like melting Artic permafrost, oceanic current systems or global temperature rises.

The approach, which is a hybrid model, which includes both AI and established mathematical theories on tipping points, goes beyond what either of the two disciplines could accomplish on their own, said Bauch.

The idea came from Bauch and his wife, Madhur Anand, an ecologist and the director of the Guelph Institute for Environmental Research. The two made waves in the scientific community last year when they used game theory to model different ways of prioritizing vaccine distribution to determine what would save the most lives.

“It definitely gives us a leg up,” Anand said of tipping points. “But of course, it’s up to humanity in terms of what we do with this knowledge. I just hope that these new findings will lead to equitable, positive change.”

Thomas Bury, formerly a student at Waterloo now with McGill University, was brought on to the project to utilize his expertise in both mathematical theories on tipping points and computer science.

The team then reached out to collaborators with environmental system-specific knowledge from the United Kingdom, Netherlands and India to help explain and contextualize the data sets the algorithm would be working with.

In these early stages, Bauch explained, it’s important to study situations that have already undergone tipping points so they can test the predictions against the data to see how accurate it is.

“Ultimately, the goal is that if we can detect a tipping point approaching, then hopefully we can avoid it,” he said. “And even if we can’t avoid it, then we can plan for it so that the worst impacts are mitigated.”

This goes beyond climate change action, explained Bury, and includes things like algal blooms, the collapse of fisheries and epileptic seizures.

Its use in the medical realm specifically holds promise.

“The transition from health to disease can arise through a tipping point, suggesting our approach could have therapeutic application,” he said.

This has real-life application for medical subjects like depression, cancer and cardiovascular events.

“The improvement of wearable devices in collecting massive amounts of physiological data makes this an appealing area for testing our algorithm,” he said.

LOCAL

en-ca

2021-09-25T07:00:00.0000000Z

2021-09-25T07:00:00.0000000Z

https://waterloorecord.pressreader.com/article/281552294007973

Toronto Star Newspapers Limited