Tackling climate change with AI – Interview with Priya Donti | Agoria

Tackling climate change with AI – Interview with Priya Donti

Published on 31/08/21 by Pierre Martens
The IPCC 6th assessment report (AR) leaves little to the imagination. The earth is warming faster than previously thought and the window to avoid catastrophic outcomes is closing.
Some key highlights:
- In 2010-2019, the global-mean surface air temperature (GSAT) is already 0.9°C–1.2°C higher than in 1850–1900.
- Climate change is affecting every inhabited region across the globe, but the effects are not the same in all regions.
- The world will pass 1.5°C somewhere between 2030 and 2035, depending on the future emissions scenario.
Global mean sea level (GMSL) is rising, and the rate is faster than over any preceding century in at least the last three millennia.- Even if enough carbon were removed from the atmosphere that global emissions become net negative, some climate change impacts, such as sea level rise, will be not reversed for at least several centuries.
Read the 6th IPCC report

It is clear that we will need both mitigation strategies, to reduce CO2-emissions as quickly as possible, as well as adaptation strategies, to cope with climate change impacts like the sea level rise, hot extremes, heavy precipitation, agricultural drought, etc. 

We sat down with Priya Donti, co-founder and chair of climate change AI, to understand which role AI can play in tackling climate change. 

Who is Priya Donti?
Priya Donti is a Ph. D. student in Computer Science and Public Policy at Carnegie Mellon University, and a U.S. Department of Energy Computational Science Graduate Fellow. She is also a co-founder and chair of Climate Change AI, an initiative to catalyse impactful work in climate change and machine learning.

" My Ph.D. work focuses on machine learning for forecasting, optimization, and control in high-renewables power grids. Specifically, my research explores methods to incorporate the physics and hard constraints associated with electric power systems into deep learning models.

On 1 October, Priya Donti will give the keynote presentation during Agoria’s “AI in Energy” launch event

When did you start with the Climate Change AI initiative and what was your motivation? 
“It all began 2 years ago when we launched a paper, in which we outlined the various ways in which machine learning can help to mitigate - and adapt to - climate change, together with around 20 researchers. 
Many people in the AI-community are concerned about climate change, so were looking to figure out how they could use their expertise to make a contribution. On the climate change side on the other hand, various streams of data are emerging, and energy and climate experts around the world are seeing the possibilities offered by machine learning. With the Climate Change AI initiative, we wanted to bring these 2 ecosystems together.”
Today, the new IPCC report appeared, demonstrating that global warming is happening at a faster rate than earlier anticipated. Do you believe we will be able to limit global warming to, let's say, 1,5°C?
« I continue to be optimistic that we can do something about it. Climate change doesn't have an on/off switch. We can chose how much warming we will accept as a society. Obviously we need to accelerate efforts to reduce emissions, and in addition, we also need to think of adaptation measures. »
Can you give some examples of how AI can tackle climate change? 
« On the mitigation side, AI can e.g. help to automate heating & cooling systems in buildings, based on the combination of sensor data, weather forecasts, occupancy patterns, etc. There are also use cases that can improve the balancing of the electricity grid, e.g. by better forecasting renewable energy production and energy demand. However, AI is not a silver bullet. Where it is applicable it is typically only a part of the solution. Implementing smart controls in a building without insulation would not be very useful. 
On the adaptation side, AI can help to reduce the computation time associated with large climate models, e.g. by approximating parts of these models and determining better estimates for certain parameters of these models. This can help to reduce the complexity of climate change models and help to downscale them. This is important because local communities and countries need to understand what the localized effects of climate change are going to be in their area. Climate change is already affecting every inhabited region across the globe, but the effects are not equal in every region. 
Another emerging area is the use of machine learning for insurance by e.g. determining the risk of flooding and forest fires in certain areas. 
Finally on the policy side, AI can help to analyse policy documents of individual governments and companies, to analyse patterns, help determine which policies worked or not, and provide info to citizens on how they can best advocate for certain policies to their representatives or companies. » 

"AI is not a silver bullet. Where it is applicable it is typically only a part of the solution.
Implementing smart controls in a building without insulation would not be very useful."

Priya Donti, co-founder and chair of climate change AI

Are you not concerned with the energy consumption associated with digital technologies like AI and blockchain? Will this not impact the positive contributions they can make? 
“It strongly depends on the use case. We discussed some machine learning cases that can make a positive contribution to climate action, but ML that is used to accelerate the extraction and use of oil & gas, or to advertise certain consumption products, can obviously have a large negative impact on the climate.  
Regarding the carbon footprint of machine learning, you need to consider not only the datacenters and their energy consumption, but also the hardware that is deployed for running these ML applications in the field and their energy (and resource) requirements. While some of the largest ML models are quite energy intensive, most the ML algorithms that we discussed in our paper are actually rather small. To consider the impact of machine learning, you therefore need to look at it holistically and weigh the positive vs. the negative impacts. 
Another important element that is not often discussed, is the equity of access to machine learning models, so people in all regions can contribute to their development. Large machine learning models are often costly to run, which means many institutions have difficulty contributing to their development. This is also linked to climate justice. Underprivileged populations often face the worst impacts of climate change, but these same populations can often be locked out of technological development processes aimed to address climate change - unless care is taken to explicitly counteract that.” 
You have a dataset wish list on your website. Is access to data difficult?  
“After writing the Climate Change AI paper, we composed a dataset wish list to understand what the state of the data is like for the applications in the paper. There are lots of bottlenecks when it comes to data. It's often scattered, privately owned, or not properly licensed. Maybe there are safety or security concerns. Often companies and organizations are sitting on a lot of data that they think might be valuable someday, even though they're not using it yet. Companies, utilities and end-consumers concerned about climate change should reflect on how they can make their data publicly available for the benefit of the greater community.
In a lot of physical settings, it's not only about data, but also about having access to realistic simulation environments. Especially in the electric power sector, there are not so many options to see if your solution works in practice.”
Can (Belgian) people in get involved? 
“Yes, of course. We run a series of workshops at the major machine learning conferences, with typically 1000 to 2000 attendees from all around the world. It's a global movement of people who want to use their ML skills, or people from the climate side. Every other week we also organize a virtual happy hour as well as a monthly webinar series, and recently we launched a community platform, to facilitate the matching of skills within the community. You can also sign up for our newsletter.
Finally, you don't need to be an AI or climate expert to make a contribution. Every citizen can think how they can support in additional ways, like encouraging your company to take action, or speaking to your representatives.” 

Launch event AI in Energy
If you want to learn more about the different ways in which AI can help to tackle climate change, tune in on 1 October for the launch event of ‘AI in Energy’, a free online course on the transformative impact of AI in the energy sector. 

Federal Minister of Energy Tinne Van der Straeten will launch the course during a virtual event on October 1st (15h00 - 16h30).

After the opening speech by Tinne Van der Straeten, Priya Donti, the co-founder and chair of Climate Change AI, will talk about the different ways in which AI can help to tackle climate change.

Interested? Register now here 
Was this article useful?