To what extent can artificial intelligence help tackle climate change today?
Now Reading
To what extent can artificial intelligence help tackle climate change today?

To what extent can artificial intelligence help tackle climate change today?

While artificial intelligence (AI) is often associated with the spawning of robots that will take our jobs, Terminator’s Skynet, or the unblinking red eyes of Hal 9000 in 2001: A Space Odyssey, its true and immediate effects are best seen by simply observing the innovations — ones that prove that software can do a variety of tasks better than humans can.

If one thing is clear, it’s that artificial intelligence has the potential to disrupt every industry, which leads to a big question that should matter to all of us:

To what extent can a powerful technology like artificial intelligence be used to help us tackle climate change?

To learn more about how we can leverage artificial intelligence to tackle climate change, I had to chat with Priya Donti, who’s completing a Ph.D. in Computer Science and Public Policy at Carnegie Mellon University, focused on the role machine learning can play in climate change mitigation solutions. Donti is also a co-chair of Climate Change AI, an organization that unites “volunteers from academia and industry who believe in using machine learning, where it is relevant, to help tackle the climate crisis.”

Priya Donti, Carnegie Mellon University Computer Science and Public Policy PhD Candidate and co-chair of Climate Change AI
Priya Donti, Carnegie Mellon University Computer Science and Public Policy PhD Candidate and co-chair of Climate Change AI | Photo Courtesy of Priya Donti

Our conversation, which has been edited for length and clarity, discusses the risks, the rewards, and the limitations of using artificial intelligence to combat climate change.


CHRIS CHEN: As it relates to tackling climate change, what is artificial intelligence capable of doing so far?

PRIYA DONTI: There are many cases in which machine learning technologies can be and are being applied today. For example, electricity system operators are already using machine learning-based forecasts of renewable energy and electricity demand to manage power grids with large amounts of renewables.

Similarly, many remote sensing techniques can be applied out of the box to gather climate-relevant data and insights from satellite imagery. In order to have real-world impact, it is important that projects be developed in conjunction with relevant stakeholders, such as power system operators and policymakers, who can apply them with high impact.

It’s interesting you mention policy, which is typically thought of in a people-driven context. How do you see AI changing that landscape?

AI can help provide useful insights and analyses to decision-makers as they formulate climate policies. For instance, AI can help fill information gaps in cases where on-the-ground information is not available, by analyzing satellite imagery to pinpoint sources of carbon dioxide emissions, map the locations of rooftop solar panels, or infer the energy efficiency characteristics of buildings.

These analyses can be conducted inexpensively and at scale; as a result, while the resulting estimates should not be viewed as replacements for targeted data-collection efforts, they can provide much-needed insights in the meantime.

In addition, AI can help down-scale climate models to provide local governments insights on how climate change will affect their jurisdictions; analyze legislation and political texts to discern the effects of previous climate policies; and speed up simulation and optimization models that are often used to trade-off between different policy options.

Aside from policy, can AI help us reduce carbon emissions or achieve some kind of direct climate-positive effect?

Electricity grids are a delicate balancing act, where the amount of electricity fed into the grid must equal the amount of electricity consumed from the grid at any given moment. This balance can be difficult for power system operators to achieve, especially as we depend more on solar, wind, and other electricity sources whose power output varies from moment to moment based on the weather.

AI can help power system operators better balance power grids, and therefore handle larger amounts of time-varying renewable electricity sources. For instance, AI can combine information such as historical weather, satellite imagery of clouds, and the outputs of physical weather models to forecast the amount of solar and wind power that will be produced in the near future. This application, called nowcasting, can give power system operators much-needed lead time to make decisions about how to best balance the grid.

AI can also enable strategies such as demand response, where electricity consumption can be intelligently reduced or shifted in time depending on the cost or carbon intensity of the electricity that is actually available. Specifically, AI can help optimize equipment such as industrial chillers or smart thermostats, so they prioritize using electricity at beneficial times. 

AI will require large amounts of personal data to accurately forecast climate trends, right? What kind of security concerns can arise as a result?

In many cases, the data used in climate change applications, such as satellite imagery or historical weather data, is non-personal in nature. That said, in contexts where personal data is gathered — such as in smart cities — it is extremely important to put policies in place that govern what data can be gathered in the first place, who owns this data, and how it is protected. 

Another important consideration is equity. AI tools are only as good as the quality of the data they use and the sociopolitical systems within which they’re implemented. For instance, if a machine learning algorithm is given racially biased data, or data that contains information about some geographical locations but not others, the algorithm will likely pick up on these biases, potentially replicating and even amplifying them.

Separately, AI technologies can often be leveraged more effectively by those with money or power than those without, and this difference in access can exacerbate existing inequities. These considerations around data bias and the democratization of AI tools are extremely important to consider from an ethical and policy perspective.

Aside from the ethics, what other hurdles are ahead?

There are a number of hurdles related to meaningfully deploying AI “in the real world” as it relates to climate change. Real-world data can be messy, distributed, disorganized, and often private, and therefore not in a form that is friendly for AI algorithms to use.

Like all societally-oriented work, impactful projects in this space must also be thoughtfully co-developed with relevant stakeholders and in multi-disciplinary teams, and building these connections can be challenging.

Finally, there are many applications at the intersection of climate change and AI that are potentially impactful, but are long-term or lack clear monetization strategies, and are therefore difficult to scale under current funding and deployment models. Climate Change AI was created with the goal of addressing some of these hurdles by sharing resources and facilitating cross-disciplinary teamwork. However, much remains to be done.

As funding for AI is increasing rapidly, is this trend reflected in your work too? Are there a lot of resources available to researchers in your field?

While AI research funding as a whole is definitely increasing, the array of stakeholders working in the AI and climate change space is wide and diverse, and funding is not always available uniformly to all these stakeholders.

For instance, we’ve heard from discussions with city governments that they do not always have the monetary resources to build in-house AI capacity. In the startup space, there is a large gap in funding for longer-term or higher-risk technologies, or those that are in an early stage where the technology is well-established, but there isn’t yet a clear monetization model.

There is a huge opportunity here to provide deployment-focused funding that ensures innovations make it out of research labs and are applied meaningfully on the ground.

At the end of the day, do you view AI as a silver bullet to tackling climate change? 

No — in the end, AI is simply a tool that, like many other tools, can help inform or accelerate existing climate change strategies within policy or engineering. Climate change is a multi-faceted problem with multi-faceted solutions, and addressing it will require rapid action from all corners of society. I would encourage everyone to leverage their unique skills to help tackle climate change.


Did you enjoy this interview? Then please consider subscribing so we can keep funding our independent journalism. As a subscriber, you’ll get our most exclusive members-only stories and commentary from industry experts.

© 2020 Medius Ventures LLC. All Rights Reserved.
Scroll To Top
Copy link