BY PIP BAITINGER
Pip Baitinger is a Master of International Relations student focusing on technology, transnational advocacy, and gender in IR. She is a US Air Force veteran where she worked in computer networking and cybersecurity.
Governments around the globe are in dire search for technological solutions to address impending dangers related to climate change. Many decision-makers and leaders see the potential of curbing climate catastrophes in emerging technologies such as Artificial Intelligence (AI). AI could have the potential to create carbon-neutral economies through its ability to analyze large and complex environmental data sets.[1] This ability to analyze large amounts of environmental data sets allows for more decision-making bandwidth to be focused on finding suitable and innovative solutions that humans might miss. The computing capacity of AI is also being applied toward protecting against biodiversity loss. Global Fishing Watch uses AI to visualize and relay patterns associated with overfishing. This AI system, partially created by Google, uses GPS and satellite data to create a map showcasing areas impacted by overfishing.[2] By visualizing the areas impacted by overfishing, AI allows governments to see where overfishing must be stopped to ensure biodiversity loss is at a minimum. Hence, optimists see AI as the driving technology in finding solutions to reduce global temperature rising and increase biodiversity.
However, this optimism should be met with skepticism. AI and the computer systems they run on contain just as many environmental costs as benefits that governments must confront before AI can be successfully utilized to benefit the environment and the people most at risk from the effects of ecological disasters. Governments must enact policies and guidelines on how AI should be developed in a world teetering on the brink of climate disaster. The European Union (EU) offers a practical, sustainable, and green AI governance framework. This framework is a valuable starting point for how global governments and international organizations should collaborate in creating norms and guardrails related to AI development and the environment.
How AI Contributes to Environmental Problems
1. Carbon Footprint and Water Consumption
In late 2022, OpenAI released the newest version of its generative text-based machine learning model to the public, ChatGPT or GPT-4. ChatGPT has thus far revolutionized the market of large language models and AI. This technology breakout has caused both wonder and concern over the future of such generative forms of artificial intelligence. On the upside, such a technology could revolutionize writing and creative spaces. However, this kind of technology and the popularity of such technology comes at a cost. Generative AI systems like ChatGPT must be trained to operate on vast data. To store such data, AI relies on the usage of data centers which have an extremely high carbon footprint. An estimated 500 metric tons of carbon dioxide have been produced through training ChatGPT’s older model, GPT-3.[3] Current estimates found that the energy used to train and run ChatGPT is equal to the carbon usage of 175,000 Danish citizens in one year.[4]
Generative AI and large language models also rely on high water usage levels for cooling the data storage hardware. In recent studies, GPT-3 required 750,000 liters of clean, fresh water for training. This is comparable to the amount of water needed to produce 320 Tesla vehicles. Additionally, the study compares the water usage of one conversation with ChatGPT, equating to the waste of a 500 mL bottle of water per conversation.[5] This water usage becomes increasingly problematic, given that more regions worldwide are experiencing climate-change-induced droughts.
2. Inequities of AI
Beyond its environmental footprint and energy consumption, AI’s most notable issue is its amplification of human biases and inequities. AI and machine learning algorithms are burdened by the biases of the individuals who program them. These biases can reinforce inequity and discrimination against marginalized groups. Despite the direct consequence of increased discrimination, AI’s biases could also increase income inequalities. Due to the high emissions and subsequent environmental damage associated with the richest percentiles of the world, increased income inequalities have environmental repercussions. AI has the potential to amplify these repercussions by allowing the world's wealthiest individuals the ability to pay high prices for the increased intelligence and productivity that AI systems provide. At the same time, AI threatens to automate repetitive and routine jobs. Jobs that are held by many middle- and low-income individuals.[6] These issues of AI contribute to furthering class inequalities that threaten environmental progress.
3. Migration and Bias
This bias and inequity are also problematic when used in the sphere of migration. With increasing extreme weather events and flooding in areas, especially in the global south, migration will become more prevalent.[7] Already, many nations have begun to deploy machine learning algorithms at borders to assist in applying their immigration policy. Countries like Canada and the United States utilize AI to parse asylum applications and monitor immigration.[8] Over time, it would make sense that more and more countries begin to utilize AI to assist in efforts to streamline immigration, especially given its expected increase due to climate change. This is where issues of programmed bias begin to take form. Since this bias is baked within the algorithm, there is a significant risk that this could induce greater bias and discrimination in countries’ immigration procedures, including discrimination based on race and gender.[9] Coupled with issues surrounding privacy protection for migrants,[10] AI’s deployment for immigration policy enforcement is problematic. If these biases are not dispelled in AI systems' programming and deployment processes, then it will perpetuate discriminatory policies in migration already accelerated by climate change.
The EU’s Commitment to Creating Green AI
Given the issues presented concerning AI’s threat of perpetuating and contributing to climate change, it is imperative that governments must take action to safeguard against these threats. It is also essential that innovative capabilities are not blocked by policy so that AI can lead to solutions in solving issues perpetuated by climate change. Many states are beginning to draft legislation and frameworks to protect against the issues that AI and machine learning present. However, many of these frameworks and policies do not focus on the intersections of climate change and AI. For example, the Biden White House presented its “AI Bill of Rights” in October 2022. This non-binding document outlines how AI should be regulated to safeguard against bias and inequity.[11] The EU has also introduced and is in the imminent stages of passing the AI Act governing against AI it deems ‘risky.’[12] The AI Act looks to be a trailblazing policy that will begin to create norms of governance surrounding AI. The legislation also represents the EU’s dedication to finding solutions that encourage innovations in AI while also considering the negative impacts it might have. This dedication is seen through earlier documents and policy recommendations surrounding sustainable and green AI.
In 2021, the EU published “The Role of Artificial Intelligence in the European Green New Deal.” This document outlines the risks and solutions concerning AI and its impact on the environment. Its recommendations for integrating AI into the EU's sustainability and environmental goals point to increasing resources to ensure that AI is developed sustainably within the EU by fostering pathways to encourage businesses and EU states to build sustainable practices in their AI development.[13] Thus far, this document represents one of the only instances of a government organization taking steps to combat AI's threats to the environment and worsening climate change. Along with this policy recommendation document, the EU has built-in ideas of intersecting AI development with sustainable practices. In what is sometimes referred to as a ‘twin transition,’ much of the EU’s policies in recent years have been centered on creating norms and frameworks on AI and climate change.[14] These frameworks create incentives for businesses to build sustainable AI and encourage innovation.
Conclusion
To conclude, the increasing development and deployment of AI worldwide present vital challenges in the fight against climate change. Along with AI’s massive carbon footprint and water usage, it also has issues of bias, inequity, and an ability to allow repressive states to repress climate activists more easily. However, AI still offers an opportunity to find innovative solutions to fixing problems caused by climate change. Governments have the opportunity to get ahead of businesses and enterprises in creating norms surrounding AI’s role in the environment. The EU is leading the way in creating initiatives and policies focusing on the intersections governing AI and climate change. Although the EU needs to further develop these policies and pass more legislation to establish green AI frameworks, governments and international organizations should consider the EU an example of safeguarding against AI’s environmental costs.
PHOTO CREDIT: Michael Cordedda licensed under CC BY 2.0.
References
[1] Megan Mastrola, "How AI Can Help Combat Climate Change," The Hub, March 7, 2023, https://hub.jhu.edu/2023/03/07/artificial-intelligence-combat-climate-change/#:~:text=Another%20application%20of%20AI%20to.
[2] Brian Sullivan, "Mapping Global Fishing Activity with Machine Learning," Google, September 15, 2016, https://blog.google/products/maps/mapping-global-fishing-activity-machine-learning/.
[3] Melissa Heikkilä, "We’re Getting a Better Idea of AI’s True Carbon Footprint," MIT Technology Review, November 14, 2022, https://www.technologyreview.com/2022/11/14/1063192/were-getting-a-better-idea-of-ais-true-carbon-footprint/.
[4] Kasper Groes Albin Ludvigsen, "ChatGPT’s Electricity Consumption," Medium, March 5, 2023, https://towardsdatascience.com/chatgpts-electricity-consumption-7873483feac4.
[5] Pengfei Li et al., "Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models," (Cornell University), April 2023, https://doi.org/10.48550/arxiv.2304.03271.
[6] Peter Dauvergne, AI in the Wild: Sustainability in the Age of Artificial Intelligence (Chicago: University of Chicago Press, 2022), 121-125.
[7] Mia Prange, "Climate Change Is Fueling Migration. Do Climate Migrants Have Legal Protections?" Council on Foreign Relations, December 19, 2022, https://www.cfr.org/in-brief/climate-change-fueling-migration-do-climate-migrants-have-legal-protections#:~:text=Why%20is%20climate%20migration%20on%20the%20rise%3F&text=Climate%20migration%20occurs%20when%20people.
[8] Roxana Akhmetova and Erin Harris, "Politics of Technology: The Use of Artificial Intelligence by US and Canadian Immigration Agencies and Their Impacts on Human Rights" (Edward Elgar Publishing, 2021), accessed April 22, 2021, https://china.elgaronline.com/display/edcoll/9781789909142/9781789909142.00008.xml.
[9] Peter Dauvergne, AI in the Wild: Sustainability in the Age of Artificial Intelligence, 120 - 124.
[10] Tuba Bircan and Emre Eren Korkmaz, "Big Data for Whose Sake? Governing Migration through Artificial Intelligence," Humanities and Social Sciences Communications 8, no. 1 (2021): 1-5, https://doi.org/10.1057/s41599-021-00910-x.
[11] The White House, "Blueprint for an AI Bill of Rights," The White House, 2022, https://www.whitehouse.gov/ostp/ai-bill-of-rights/.
[12] Foo Yun Chee, Martin Coulter, and Supantha Mukherjee, "EU Lawmakers’ Committees Agree Tougher Draft AI Rules," Reuters, May 11, 2023, sec. Technology, https://www.reuters.com/technology/eu-lawmakers-committees-agree-tougher-draft-ai-rules-2023-05-11/.
[13] Peter Gailhofer et al., "The Role of Artificial Intelligence in the European Green New Deal," European Parliament, IPOL_STU(2021)662906_EN.pdf.
[14] Josh Cowls et al., "The AI Gambit: Leveraging Artificial Intelligence to Combat Climate Change—Opportunities, Challenges, and Recommendations," AI & SOCIETY (October 2021), https://doi.org/10.1007/s00146-021-01294-x.
Works Cited
Akhmetova, Roxana, and Erin Harris. 2021. “Politics of Technology: The Use of Artificial Intelligence by US and Canadian Immigration Agencies and Their Impacts on Human Rights.” China.elgaronline.com. Edward Elgar Publishing. April 22, 2021. https://china.elgaronline.com/display/edcoll/9781789909142/9781789909142.00008.xml.
Bircan, Tuba, and Emre Eren Korkmaz. 2021. “Big Data for Whose Sake? Governing Migration through Artificial Intelligence.” Humanities and Social Sciences Communications 8 (1): 1–5. https://doi.org/10.1057/s41599-021-00910-x.
Chee, Foo Yun, Martin Coulter, and Supantha Mukherjee. 2023. “EU Lawmakers’ Committees Agree Tougher Draft AI Rules.” Reuters, May 11, 2023, sec. Technology. https://www.reuters.com/technology/eu-lawmakers-committees-agree-tougher-draft-ai-rules-2023-05-11/.
Cowls, Josh, Andreas Tsamados, Mariarosaria Taddeo, and Luciano Floridi. 2021. “The AI Gambit: Leveraging Artificial Intelligence to Combat Climate Change—Opportunities, Challenges, and Recommendations.” AI & SOCIETY, October. https://doi.org/10.1007/s00146-021-01294-x.
Dauvergne, Peter. 2020. AI in the Wild Sustainability in the Age of Artificial Intelligence. Cambridge, Massachusetts The Mit Press.
Gailhofer, Peter, Anke Herold, Jan Schemmel, Cara-Sophie Scherf, Cristina Urrutia, Andreas Köhler, and Sibylle Braungardt. 2021. “The Role of Artificial Intelligence in the European Green New Deal.” https://www.europarl.europa.eu/RegData/etudes/STUD/2021/662906/IPOL_STU(2021)662906_EN.pdf.
Heikkilä, Melissa. 2022. “We’re Getting a Better Idea of AI’s True Carbon Footprint.” MIT Technology Review. November 14, 2022. https://www.technologyreview.com/2022/11/14/1063192/were-getting-a-better-idea-of-ais-true-carbon-footprint/.
Li, Pengfei, Jianyi Yang, Mohammad A Islam, and Shaolei Ren. 2023. “Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models.” ArXiv (Cornell University), April. https://doi.org/10.48550/arxiv.2304.03271.
Ludvigsen, Kasper Groes Albin. 2023. “ChatGPT’s Electricity Consumption.” Medium. March 5, 2023. https://towardsdatascience.com/chatgpts-electricity-consumption-7873483feac4.
Mastrola, Megan , and 2023. 2023. “How AI Can Help Combat Climate Change.” The Hub. March 7, 2023. https://hub.jhu.edu/2023/03/07/artificial-intelligence-combat-climate-change/#:~:text=Another%20application%20of%20AI%20to.
Prange, Mia. 2022. “Climate Change Is Fueling Migration. Do Climate Migrants Have Legal Protections?” Council on Foreign Relations. December 19, 2022. https://www.cfr.org/in-brief/climate-change-fueling-migration-do-climate-migrants-have-legal-protections#:~:text=Why%20is%20climate%20migration%20on%20the%20rise%3F&text=Climate%20migration%20occurs%20when%20people.
Sullivan, Brian . 2016. “Mapping Global Fishing Activity with Machine Learning.” Google. September 15, 2016. https://blog.google/products/maps/mapping-global-fishing-activity-machine-learning/.
The White House. 2022. “Blueprint for an AI Bill of Rights.” The White House. 2022. https://www.whitehouse.gov/ostp/ai-bill-of-rights/.