MIT's Priya Donti on How AI can help Boost Renewable Energy, Monitor Environment
Priya Donti, assistant professor, MIT, and co-founder and chair of the non-profit Climate Change AI, talks to Nabodita Ganguly about how AI can help fight climate change, collect data for climate-tech start-ups and more. Edited excerpts
Climate artificial intelligence (AI) sounds a bit counterintuitive, especially because AI is known to consume a lot of energy.
When I think about the relationship between AI and climate change, there are a few key aspects. One is that AI can be used to support various strategies for climate action. This includes helping us improve the production of solar and wind power, as well as making power grids more efficient so we can better utilise renewable energy.
It also includes better monitoring of events such as deforestation and floods using satellite imagery. Then there are applications focused on optimising heating and cooling systems in buildings. There have been several demonstration projects in India and elsewhere showing that adding intelligence to these systems can reduce energy use by more than 20%.
Where does India currently stand on using AI to manage and optimise power grids?
I don’t have precise numbers or specific examples for the Indian context, but at a high level. I think that’s because India is experiencing such strong growth, along with the need to provide electricity access to a large and growing population, there has been a major focus on adding more power to the grid from any available source.
One positive development is that over the past 5–10 years, the cost of solar power has fallen significantly. In many cases, it is now cheaper than natural gas and much cheaper than coal.
As a result, using solar power along with battery storage is often the most economically efficient option for adding new capacity.
In the US, much of the climate work today aligns both with environmental goals and economic incentives
Many AI start-ups struggle with data in niche sectors like climate. How do you tackle data challenges while ensuring reliability in critical systems?
While there is often a lot of data available, it may not always be clean or usable. The issue is not just about data availability, but also about accessibility. In many cases, data is collected but not made easily available to others. For example, in the biodiversity space, bioacoustic data, recordings of environmental sounds, is collected to help identify species and monitor ecosystems.
This involves large volumes of continuous audio data, often gathered by NGOs. However, these organisations may not have the storage capacity or infrastructure to make this data widely accessible or to handle requests from others who want to analyse it.
There are also challenges around data licensing. Even when data exists, it is not always clear who can use it, whether it can be used for commercial purposes or if it is restricted to academic use.
So overall, data remains a critical issue. It requires targeted efforts in better data collection and cleaning, as well as stronger infrastructure for storing, sharing and accessing data. It also requires more clarity around licensing frameworks.
Another important aspect is data sharing and reuse. In some cases, even if data is limited in one location, a similar region may have more available data. For instance, if you are forecasting electricity demand in one part of Haryana and have limited data, another nearby region may have richer datasets with similar patterns. In such cases, sharing data, augmenting datasets, or even sharing machine-learning models can help bridge the gap.
In my own work, particularly in the context of US and European power grids, data is relatively abundant because these systems collect a large amount of operational data. However, access to this data is often restricted, especially for those outside grid operators or utilities. This is due to privacy and security concerns.
Given the current political climate in the US, many expected funding for climate start-ups to slow, but that hasn’t fully happened.
I would say that in the US, much of the climate work today aligns both with environmental goals and economic incentives. In many cases, it is the right thing to do from a climate perspective as well as from a cost perspective.
For example, deploying solar power is both cleaner and far cheaper than fossil fuels. In such areas, whether you are building a company or doing research, you can make a strong case that this work improves the affordability and reliability of power. So, in these cases, broader societal goals are aligned with climate goals, and there is clear synergy.
However, challenges arise in areas where something is beneficial for the climate but not economically attractive. Biodiversity is a good example. In such cases, it becomes harder to sustain work through market-driven funding alone. This is where philanthropy, corporate sustainability initiatives or CSR [corporate social responsibility] funding need to play a role in bridging the gap.
Is there some sort of fear right now working on something like climate change in the US?
It’s not necessarily a fear, but rather an unfortunate reality. Work that is purely climate-focused, especially when not aligned with other prior-itised economic or policy goals, tends to be harder to sustain. The funding ecosystem is challenging, and as a result, some companies are shutting down, making the situation difficult.
At the same time, the US is only one part of the global landscape. It is encouraging to see other governments continuing to prioritise climate sustainability.