There’s no denying that artificial intelligence (AI) is reshaping how we do things, and while there’s an opportunity to embrace AI as tools with potential to enhance green technology and sustainability efforts, there are inherent risks in simply using it.

As the development and deployment of AI accelerate rapidly, along with the data centers needed to power its notoriously inefficient processing, people are becoming increasingly aware of its environmental impacts and realizing that finding ways to balance the benefits with costs requires significant thought and strategic planning, vs. the “full steam ahead” approach currently taken by tech giants. 

The role of AI in environmental progress and impact

How is AI beneficial? It has been around since the 1950s, but its development and availability have accelerated rapidly in recent decades, notably in the last few years, making it readily available to all. 

Scientists, including environmental scientists, have, for decades, used massive datasets to enable modeling, classify, identify patterns, accelerate conservation efforts and much more. For example, researchers at the University of Michigan School for Environment and Sustainability (SEAS) are using AI to transform the once laborious process of identifying and measuring bird skeletal specimens into a fully automated one. This helps researchers understand the evolution of birds, thus contributing to better planning of responses to global change. 

In another example, SEAS researchers used AI to lay the groundwork to accelerate wildlife conservation efforts through social media. 

SEAS Professor Bill Currie, who is co-PI of the Schmidt AI in Science Postdoctoral Fellowship, which provides early-career researchers with intensive training and experience in AI applications and methodologies for scientific research, says that funding this type of research that fosters breakthroughs can make lasting scientific and societal impacts.  

“AI tools, including various forms of machine learning and neural networks, are enabling faculty to address new types of research questions in fields as diverse as astronomy, materials science, epidemiology and conservation science,” noted Currie, who added, “the Schmidt Fellowship program housed in the Michigan Institute for Data and AI in Society is also developing training in responsible AI.”   

Balancing AI, justice and sustainability

Striking a balance between benefits and harms is key to the sustainability of AI use. A growing awareness of not just the demands on the electrical grid and water consumption, but also the fact that the energy being used is heavily reliant on fossil fuels, has more and more people speaking up about a desire to take a more thoughtful approach. Additionally, the rapid expansion of data centers creates a “speed to power” challenge, the bottleneck in the U.S. electric grid where capacity can’t keep up with energy demand, which also slows down the deployment of decarbonized options and progress toward meeting clean energy targets. 

There’s also an environmental justice component to the concerns. Historically, marginalized, urban, low-income and people of color have been disproportionately affected by fossil-fuel-powered facilities, and the expansion of data centers is following a similar trend. Michelle Martinez (MS ’08), lecturer and director of the Tishman Center for Social Justice and the Environment at SEAS, says AI exacerbates these realities.

“Environmental justice can be unrepresented, misrepresented or completely invisibilized by artificial intelligence systems. From oral history to unreported environmental harms, there are various information systems or worldviews not represented in the digital world,” says Martinez. “The mass invisibilization that can recapitulate outdated ideas about who is impacted, how or why, and reframe the ways communities respond in democratic and creative ways, is dangerous. AI used for data surveillance is now being catalysed to criminalize communities, gathering facial recognition and using data without consent.”

The question of consent, who is impacted and how, and so much more, has contributed to a growing consensus of people who generally feel that they are being left on the sidelines when it comes to the AI boom. That rapid development and deployment are happening whether they want it or not, and whether important questions are being addressed or not. Where will it all lead? Who will be in control? Will it be safe? Are those in charge making sure we’re protecting the environment? Whose in charge anyway? 

With careful consideration and environmental justice concerns in mind, some experts believe that AI’s full lifecycle can likely be designed to limit its environmental impact. Some ways to do this include creating energy-efficient algorithms, managing data responsibly, shifting electrical loads to off-peak time to minimize grid impact and prioritizing green energy sources to help mitigate AI’s carbon footprint. However, this would require a willingness to slow down and to provide space for engagement and collaboration with communities to build best practices for a more strategic approach, enabling growth while also allowing for the measurement of risks, an understanding of trade-offs, and planning for a sustainable and just future.