Priya Donti is building the infrastructure of the future.
As an academic researcher, Donti builds algorithms for grids powered by renewable energy, combining her long-held aspiration to work on climate solutions with her background in computer science. As a nonprofit founder, she builds community, forging opportunities for others to apply their skills in climate as well.
“AI and computation are rarely an end-to-end solution,” says Donti. “People trained in computing should be well-versed in more than just literal computing—they should know what they’re developing and using it for.”
Priya Donti is a 2023 AI2050 Early Career Fellow and Assistant Professor at MIT Electrical Engineering and Computer Science and Laboratory for Information & Decision Systems. Her scholarship was recognized with the Association for Computing Machinery (ACM) SIGEnergy Doctoral Dissertation Award, the Siebel Scholarship, and the U.S. Department of Energy Computational Science Graduate Fellowship. Her papers have also won awards at the International Conference for Machine Learning (ICML) and the NeurIPS workshop on AI for Social Good. In 2021, she was selected for MIT Technology Review’s “35 Innovators Under 35”.
Her research develops machine learning methods for high-renewables power grids. She is also the co-founder of the nonprofit Climate Change AI, which facilitates collaborations at the intersection of machine learning and climate change. Donti’s AI2050 project develops novel machine learning techniques and simulations for power systems to enable safe deployment in the real world. This project addresses Hard Problem #4, which concerns using AI to solve humanity’s most pressing challenges.
The AI2050 initiative gratefully acknowledges Fayth Tan for assistance in producing this community perspective.
What are the challenges in decarbonizing our electrical infrastructure?
In a power grid, the amount of power injected into the power grid has to exactly equal the amount of power that’s being consumed—at every moment. This is a hard balance to maintain. There are variations in how people use energy based on the weather, social norms and behaviors, and industrial processes. In addition, we are often seeking to decarbonize energy supply with sources of power like solar and wind, which have variable outputs, which means that they’re not perfectly controllable or predictable.
However, the balance of power supply and consumption must still be maintained. Algorithms mitigate this by pulling different levers you have on the power grid, such as by figuring out when to charge or discharge batteries, modulating how much electricity supply enters the grid, adjusting power grid topologies, or shifting when heating and cooling occurs in buildings.
What do you need to take into account when designing these algorithms?
Algorithms need to be scalable to deal with the many dynamic variables in the system, and must accommodate the physics of the power grid. Besides balancing supply and demand, other constraints need to be satisfied, such as not overloading any line on the grid beyond its capacity. Failing to satisfy these constraints might result in part of the power grid blacking out, leading to the loss of lives and billions of dollars.
In addition to your research, you co-founded a nonprofit, Climate Change AI, and co-wrote a paper, titled “Tackling Climate Change in Machine Learning”. How did that paper come about?
While I was presenting at the NeurIPS Machine Learning Conference on AI in 2018, I got an email from [AI2050 Early Career Fellow] David Rolnick, who eventually became one of the co-founders of Climate Change AI. He organized a small lunch for people who hadn’t worked on climate change, but were interested in applying their skills to it. David wanted to facilitate that exploration and extend that to the rest of the community.
We followed that lunch up with a workshop at another major machine learning conference in mid-2019, the International Conference for Machine Learning (ICML). The people we looped in eventually formed the founding team of Climate Change AI. At NeurIPS, we had 40 people at lunch, but at ICML, we had hundreds of people lining up for the workshop. There was huge interest in working on climate, but also many questions about how to gain expertise, or find collaborators and data.
After the workshop, my co-authors and I decided to provide resources, entry points, and opportunities for people to get into this space, and it snowballed from there.
What impact have you seen from the work of Climate Change AI?
I think it gave people the tools, or even the courage and community, to start working on climate.
At the policy level, our paper has informed initiatives and programs incorporating AI into climate funding, including efforts such as ARPA-E’s DIFFERENTIATE and Vinnova’s “AI in the Service of Climate. The UK National AI Strategy cited our paper as the rationale for including climate into their AI strategy. It’s been amazing to see our work shape national and international policy, and see the area grow in the last few years. Even within Climate Change AI, 10,000 people registered for our 2023 summer school, where we organized both virtual programming and an in-person interdisciplinary collaboration on climate change-related projects.
Climate change often makes people feel overwhelmed or anxious. Do you share that experience, and if so, how do you manage those feelings?
Climate anxiety is real and overwhelming. It can feel disempowering. But taking action on climate and finding a community of people who are doing it [too] can make problems feel more tractable and give you a sense that real change is occurring.
You’re also working on goals with tight timelines. Not achieving net zero by 2030 or 2050 doesn’t mean that we’ve lost, but it does create a sense of urgency. It feels tempting to dive in head first all the time, but it’s also important to incorporate care so you can work sustainably instead of burning out.
If you encountered a student who wanted to do something about climate change, what advice would you give them?
People paralyze themselves by saying, “I want to do something about climate change, but I don’t know where to start.” They think they need to discover the optimal way to work on climate and build their whole life around it.
The reality is that climate change is a huge problem. Whether you’re a scientist, engineer, policymaker or artist—your skills are needed. There are undoubtedly people who are in your specific field who are also working on climate. Check out what they’re doing.
Also, let what you enjoy on a day-to-day basis anchor you. Something that theoretically sounds like the right thing to do isn’t much use if you aren’t personally motivated by it. If there’s something that motivates you to get up every day, that’s a better thing to do for you.