What AI Can Do for Climate Change, and What Climate Change Can Do for AI
The April 4, 2022 report from the U.N. Intergovernmental Panel on Climate Change makes it clear that it is “now or never” for the planet. We are “firmly on track toward an unlivable world,” U.N. Secretary-General Antonio Guterres said in releasing the report. There’s every chance that global temperatures will soar by 3 degrees Celsius, twice as much as the agreed-upon 1.5 C limit. Unless we take drastic steps and cut down emissions by 43 percent within this decade, the full force of this existential threat will be upon us.
In this context, it is interesting that some researchers have taken artificial intelligence—a technology often considered an existential threat in its own right– and tried to turn it into a vehicle for climate action. Since I’m writing a book on nonhuman actors in the 21st century, I have a more-than-passing interest in how these experiments turn out and what they say about our collective future. Could it be that climate change is the catalyst that transforms AI, challenging it to be more crisis-responsive, more focused on innovations addressing large-scale hazards? Such technology could be just what we need at this juncture. It could generate emergency action very different from the profit-driven, bias-amplifying and misinformation-spreading algorithmic regime we are familiar with. Giving much larger play to input from the field, from networks of engaged participants, this “climate AI” could be a game changer in the tech ecosystem, as in the physical ecosystems now facing their worst risks.
This new ethos was reflected in a report written by the Global Partnership on Artificial Intelligence and presented at the COP26 climate summit last November. The 15 co-authors, researchers and activists from 12 countries, argue that while we need to be vigilant about racial and gender biases and the tendency of big data to perpetuate inequities, AI can nonetheless play a key role in prediction, mitigation, and adaptation, in ways we can’t afford to ignore.
Take the visualization platform This Climate Does Not Exist, timed to coincide with the COP26 report. Built by motivated young researchers, this project demonstrates that a harmful technology can be repurposed to make climate change personal, visceral and unforgettable for the general public. Using the same machine learning algorithm that swaps visual and audio data to produce fabricated, hyperrealistic videos called deepfakes, it generates similarly real-looking views of floods or wildfires for any street address.
We have become “a bit blasé” about climate disasters when they happen to strangers, says lead researcher Sasha Luccioni, a postdoc at the Quebec Artificial Intelligence Institute. Seeing our own houses standing in several feet of water makes us take notice.
While the full impact of This Climate Does Not Exist remains to be seen, what seems clear is that AI is phenomenally adaptive, able to pivot to entirely different demands under crisis conditions. It’s most innovative coming from practitioners galvanized by fast-approaching disasters.
