Faced with these developments, we feel that the intersection of climate change and AI is a topic that will dominate headlines in the (near) future. Collaboration in these fields between scientists, governments, organizations and AI practitioners could have far reaching positive effects.

In this article, we will explore what AI can do for climate change, and what some of the pitfalls are that we have encountered. It will give you a top level overview as an introduction to this fascinating, and rapidly evolving field.

The time to act is now

As the saying goes, the best time to plant a tree was 20 years ago, the second best time is now. The 2020s will be a critical decade for climate change. Leading up to 2020 were the global climate protests, broadly motivated by young people concerned for their future. Now the world is facing an economic shock unlike any we’ve known — and calls to make the economic recovery a green recovery. If we can follow up on these intentions, then the willingness to act on climate change will lead to increasingly ambitious international commitments and goals set by national and local governments. These will have a cascading effect on organizations: as the deadlines for these commitments grow closer, the institutional and economic landscape will change. Countries are increasingly held to their commitments, such as the Netherlands. We are seeing that legislation, guidelines, and policies will mandate the need for having carbon neutrality at the core of future economies. These developments will have far-reaching impacts for governments, organizations, and citizens. It is in this context, that artificial intelligence can be a game changer.

Emerging initiatives on the intersection of climate change and AI

Over the past years, the possibilities of AI to positively impact climate change have been increasingly explored in the field. A number of initiatives have emerged which aim to utilize the advantages of AI with the goal of fighting climate change and attaining the Sustainable Development Goals (SDGs). Some prominent examples are initiatives such as Climate Change AI, the AI for Good Global Summit, and its application by NGOs such as the WWF. We are also seeing increased use case explorations such as Microsoft’s AI for Earth and municipal initiatives such as AI 4 Cities. These initiatives are exploring the realm of what’s possible for AI in the context of societal challenges.

This might make you wonder; how exactly can AI impact climate change? We explored this question broadly from two vantage points: climate mitigation, and adaptation.

AI for climate change mitigation

Climate change mitigation efforts are those that target reduction or prevention of the emission of greenhouse gases. This is usually realized through innovation, efficiency, and behavior change. New technologies and renewable energy play a major part in most scenarios, but as of yet, have failed to meet the requirements that are necessary to achieve the stated commitments. This is where AI could be a game changer. As the writers of “Tackling Climate Change with Machine Learning” present, there are myriad use cases for electricity systems, transportation, buildings, industry and land use. As of yet, there are still only few examples that have been productionalized, but the coming years will be critical for putting theory into practice.

AI for climate adaptation

Climate adaptation efforts are concerned with adapting our lifestyle and surroundings so that we are anticipating and hopefully negating the effects of climate change consequences. Examples are increased green areas in cities to counter the heat island effect, reopening waterways, or moving away from climate-vulnerable areas. There are a multitude of applications for AI to help decision makers and organizations: evaluating vulnerable areas based on sensor data and remote sensing, improved extreme weather and disaster modelling through upscaling, AI can aid in showing the effects of extreme weather, help analyze what has happened (and might happen) based on, for example, satellite images and photography, and help identify best practices in fields such as planning and adaptation for future occurrences.

Pitfalls and considerations

While it is tempting to only discuss the exciting possibilities, there are also several considerations to be made. Without listing them all, there are two that stand out. Firstly, a lot of the immediate solutions are based on efficiency metrics. A popular application area of artificial intelligence is focused on increasing the efficiency of processes. In practice however, oftentimes the gains made from efficiency actually lead to increased use, rather than a reduction in use (or increased sustainability). It is therefore important to keep in mind the potential cascading effects of efficiency measures, to understand whether they will indeed lead to a positive climate impact.

The second potential pitfall is that in some cases, the use of AI might actually cloud more innovative solutions. When entirely new practices and a radical shift in system design are needed, a model that is trained on existing practices might not be adequate. The focus on efficiency for instance, or the use of existing treasure-troves of data, could sometimes stand in the way of more innovative approaches to solving climate change issues. In short, the overall goal of sustainability needs to be first in mind, and the solutions need to be adapted to this purpose. This is where we believe that applying AI to climate change in a human(ity)-centered way, can make the difference.

Human(ity)-centered > technology-centered

Even though artificial intelligence can play a large role in contributing to the solutions for some of the climate challenges we’re facing, it’s important to keep in mind that climate change is not a technology problem. AI won’t “solve” climate change; however, it can be an important part of a larger solution. In our view, approaching the intersection of climate change and AI in a human(ity)-centered way, means a number of things:

• Deep interdisciplinary collaboration between government, academia, organisations; from climatologists, urban planners and policy makers to machine learning experts, designers and ethicists.
• Understanding the systems at play; enabling us to consider and mitigate potentially harmful cascading effects of introducing (technological) solutions.
• Considering the trade-off in terms of the carbon footprint of large AI models; by focusing on reducing & reusing AI models.

So back to the title: Can AI solve the climate crisis? The answer is no…. but it can certainly help. It is important to realize that AI is not a hammer for every nail. Approaching climate change from a human(ity)-centered AI lens, is to a large degree about understanding which problems can actually benefit from AI, and which ones… don’t.