/cdn.vox-cdn.com/uploads/chorus_image/image/72911954/49_Priya_Donti_Final_Dims.0.jpg)
When Priya Donti was asked to choose any superpower, she responded: “Maybe the ability to suck carbon dioxide out of the atmosphere and turn it into basalt.” Short of that, she’s working on turning a different kind of superhuman power into reality: using AI and machine learning to tackle climate change through her global nonprofit, Climate Change AI (CCAI).
The project, which she co-founded in 2019 and now runs, has already become a staple in the “AI for Good” example catalog, helping governments develop grant programs, companies hone in on opportunities, and individuals find areas to work or study in.
Donti decided to dedicate her life to climate change well before CCAI, though. In her first week of high school, her biology teacher led a curriculum on climate and sustainability, showing how the world’s most disadvantaged populations are also going to bear the worst of climate change. Donti, a second-generation Indian American, thought that about settled things. Her life’s work would focus on climate.
But then she went off to college and discovered computer science. “This was a bit of a conundrum for me,” she recalls, “because I didn’t really understand if and how computer science could play a role in addressing climate change.” A paper titled “Putting the ‘Smarts’ into the Smart Grid” showed Donti the way forward, however: AI and machine learning, the paper argued, could be critical tools in developing next-generation power grids for renewable energy.
That realization was the spark for her current work. In 2019, after a PhD on deep learning and electrical power systems, Donti was one of the lead authors on a paper that set the research agenda for how AI can help mitigate climate change. Broadly, she describes AI’s climate applications in five categories: wrangling huge datasets into actionable insights; optimizing complicated systems; speeding up innovation; running climate simulations much faster; and getting better at forecasting.
In practice, that means things like processing satellite imagery to spot methane leaks or monitor flooding, using neural networks to improve prediction accuracy for wildfire risks, and analyzing heaps of data for early detection of climate-induced locust outbreaks in Kenya. “We call on the machine learning community to join the global effort against climate change,” the paper’s abstract concludes.
Of course, just unleashing AI on the climate crisis won’t automatically do good. “AI is an accelerator of the systems in which it’s employed. If you employ it in an inequitable system or employ it to optimize the wrong thing, it’s just going to make those problems worse,” Donti told Popular Science. Not every problem merits an AI solution, especially given AI’s carbon footprint.
CCAI is developing AI to supplement, not replace, other approaches. Looking forward, the group is focusing on unwinding the bottlenecks outlined in Donti’s paper, from the gaps in data that machine learning can help fill to breaking disciplinary silos and convening a broad range of experts to work together.
This fall, Donti will also join MIT as an assistant professor of electrical engineering and computer science, teaching a new generation of students what she so serendipitously discovered: In the fight against climate change, there is a unique need to bring diverse interests and expertise together, and the resulting strength is a superpower all its own.
