Is AI a Climate Solution?

Aug 9

by Spencer Rubino-Finn

Amidst a record-breaking heatwave last month, Alison and I boarded an early morning Amtrak train due south. The destination? Washington, D.C. for the 2024 Climate Crossroads Summit at the National Academy of Sciences. This was the second annual summit held at the Academy, a two-day long event featuring a wide panel of experts, aimed at showcasing current solutions and innovations for combating climate change. The day’s agenda addressed topics ranging from transboundary water management to energy transition in the financial markets to planetary health. However, the topic that stuck with me the most was the first session of the day, “Artificial Intelligence and Climate Change.”

AI has become one of the biggest buzzwords in tech over the last 2 years, dominating news cycles and catapulting chipmakers, such as Nvidia, to become some of the most valuable companies in the world. As a result, I was curious to see where the dialogue on AI would wind up.

Technological innovation certainly should have a seat at the sustainability table, but this particular technology is not without tradeoffs. Emissions associated with cloud-based computing have continued to climb. Moderated by David Goldston, director of the MIT Washington Office, the discussion featured a brilliant panel of speakers including Priya Donti, Uyi Stewart, and Topaz Mukulu. The group touched on a wide range of subjects, including reasons to be hopeful about AI, the need to prioritize underrepresented communities, and the role governments and corporations can play in leveraging AI for climate change solutions. In discussing these measures, the panel provided a nuanced approach to viewing AI and climate interactions.

So, is AI a climate solution?

TLDR: AI is an accelerant, but whether it will accelerate “climate solutions” faster than it accelerates other climate negative tech, like oil and gas development, depends on how and where it’s deployed.

The Digital Divide hinders development of AI climate solutions where they’re needed most

When asked why AI might not bring meaningful solutions to climate problems, all three speakers gave similar answers: the technology leaves many communities, especially in the Global South behind. This “digital divide” is ever increasing as AI continues to expand. Donti stated that, “the community that is often bringing forth AI tools is often a very privileged [one], often centered in the Global North.” Donti went on to stress the importance of developing AI from the ground up, with underrepresented regions’ experiences and needs at the forefront. These areas of the world often suffer the most from the impacts of a changing climate, even while emitting the fewest greenhouse gasses (GHG).

The panel of speakers at the AI & Climate Change Session included David Goldston, Topaz Mukulu, Priya Donti, and Uyi Stewart (left to right).

AI is only as effective as the data it is being fed (the garbage in, garbage out methodology) and non-representative data should also be considered bad data, even if it was gathered accurately. For instance, the Global South is inheriting Integrated Assessment Models (complex models used to drive policy change) that were predominantly developed in Western nations with no input from other regions of the world. This can lead to a pattern where technology is constantly outpacing the feedback required to make it more equitable. Additionally, AI climate models and IPCC mitigation strategies often place an outsized burden on developing nations to curb fossil fuel usage while developed countries can continue emitting at grossly disproportionate levels. Inequality can show itself in other forms as well, as oftentimes this technology is simply not readily available in regions like sub-Saharan Africa.

Stewart gave an apt anecdote about a flood in Africa that arrived with no warning, due to a lack of predictive AI weather platforms in the region. This caused the displacement of over half a million people and forced many into camps. Once in displacement camps, children could not go to school and parents could not work, disrupting the entire society. These ripple effects of climate change are severe but could be partly alleviated with proper tools.

Many of us living in the United States receive mobile alerts about a torrential rainstorm or can easily access weather models that give us enough warning to take some kind of safety precaution. In these ways, AI can be a helpful tool in adapting to the effects of climate change but the data infrastructure in place needs to be representative. Mukulu remarked that only “5 percent of cities actually have the tools that they need to understand the data coming in on GHG inventories.”

Capacity building is one of the best ways to make sure that developing nations are equipped to use AI tools to solve their climate related problems. Elevating local decision makers to the forefront of technological expansion will be important to assure that AI exists as a tool and not a hindrance to these communities. Mukulu also mentioned that the areas covered comprehensively on maps are usually commercial hubs, with satellite imagery often not focusing on developing regions that may be more prone to crisis or climate vulnerability.

Other forms of inequity stemming from AI can arise as well. Most speech data and language models are created using English or Mandarin, leaving thousands of languages left out. Additionally, AI is not inherently climate positive. It is just an accelerant technology. AI is being used by oil and gas companies to accelerate fossil fuel resource identification, extraction, and marketing, ultimately further accelerating emissions and climate change.

While modern industrialization was a vehicle for some nations (particularly Western ones) to experience rapid growth, it left many regions behind and exploited countless others for gain. When it comes to AI, the potential is there for the same mistakes to be made, further making the case for implementing guardrails along the way to reduce exploitation and make access to the technology more equitable. Without boundaries in place, it will prove increasingly difficult to bridge the digital divide.

Whether AI is net positive or not for the planet depends on Governments and the Private Sector

AI can help fight climate change, but it will require guidance from governments and coordinated partnership efforts with the private sector. Governments (at the federal level in particular), will need to play a massive role in establishing the proper regulation necessary to push AI tools in the right direction. Stewart made it clear that governments can be doing more to highlight climate adaptation, transition, and resiliency plans. Climate change is already here, so while mitigation is still important, adaptation presents itself as a more fertile area for AI application.

Governments can spur innovation and create a more diversified ecosystem of AI providers. Donti stressed how divergent AI tools can be depending on their intended use. For instance, the models and data needed for an energy sector application differ greatly from the inputs required to create a tool that would aid disaster relief. These stark differences mean relying on a small group of AI firms to handle every possible application can hamstring sustainability efforts and prevent diversity of thought in the space. Particularly, as AI expands into more robust forms like smart cities, a digital ecosystem propelled by active government regulation will be important to prevent one or two companies from controlling the behavior of entire populations.

For the private sector, their role is “knowledge translation,” as Stewart put it, or professionals who can pass on information to government decision makers, consumers, or even the communities mentioned above.

AI literacy will be a sought after and often required skill in order to solve climate challenges using digital tools. However, these skills will need to be disseminated by the private sector to the greater public. As the drivers of AI advancement and owners of the big data required to build sufficient models, private sector firms have a responsibility to make AI learnings more accessible.

Open Data: A Critical Prerequisite

Open data is perhaps the most essential piece to this. Mukulu mentioned that due to underwriting activity, many financial firms have predictive modeling on areas that are marginalized or not reported on by Western media. If these corporations release the data (in anonymized or encrypted forms), governments will have more information to make effective funding and infrastructure decisions. On the flip side, there is also plenty of potential data on the ground level waiting to be collected. The Global South is sitting on a treasure trove of data that will make AI tools more effective and representative. However, if the private sector is going to collect data to add to their models, governments need to ensure that they are protecting the rights of the individuals whose data it is, and offering adequate compensation.

AI With a Purpose

AI can certainly help solve climate change related issues, but will it? AI is much more effective when it is used to solve specific needs. For example, IBM’s Deep Blue computing system has made major impacts in broad areas like financial modeling or pharmaceutical drug development, but it was originally conceived with a specific goal in mind – to beat the world’s best chess players. Most of the advancements in AI have been developed with these specific focuses and assumptions in mind, not the other way around.

Find the problem first and then use the AI solution second, instead of forcing existing AI solutions onto problems like climate change.

Being “issue-specific” is a great way to harness the power and utility of AI. Donti illustrated this concept using the company Climate Trace, a firm that leverages satellite imagery to track GHG emissions. For the company to be successful, they first needed industry-specific knowledge regarding heat outputs and emission types. Once these concepts were understood, it was much clearer where the analytical gaps were, and thus where AI could be inserted. All three panelists emphasized the need for both a comprehensive and purposeful approach to AI climate solutions.

When there is a lack of planning, we see the extreme negative effects of AI on the climate. AI-related emissions are rapidly increasing across the tech sector – OpenAI has reported a 30% CO2e increase since 2020, while Google’s emissions in 2023 were almost 50% higher than compared to four years prior. As a result of these emissions, more fossil fuels (coal in particular) are being burned to generate the electricity necessary to keep pace with the power demands of the AI expansion. Bloomberg reported this year that AI is using so much power that old coal plants are turning on and remaining on.

However, as Donti highlighted during the session, fossil fuel usage is not “inherent” to AI, power demand is, and emissions are the consequence of becoming overly reliant on non-renewable energy forms. AI use is not decreasing anytime soon, and its warming potential won't either as long as it is fueled by carbon-intensive energy sources, so, it is important to harness all the takeaways highlighted by the panel.

The bottom line: Effective AI and climate interaction should be representative of developing and underprivileged communities, supported by regulation, and strategic in scope. Coalescing around these principles will be an essential if AI is to drive sustainable change.

Companies to Check Out

AI’s contribution to the climate sector is much more than the generative chat bots that most of us associate with AI. These cool, innovative companies are leveraging AI for climate-positive applications!

Aionics is a battery and energy storage company “using next generation compute to design next generation materials.”
Climate Trace is “mobilizing the global tech community to track GHG emissions with unprecedented detail and speed.”
Open Street Map is charting “a map of the world, created by people like you and free to use under an open license.”
Station A “helps real estate owners and long-term tenants rapidly identify and install solar, battery, and EV charging projects.”

Alison Grantham