Climate lawsuits & AI's environmental impact // The Week in Sustainability #35

Supreme Court and climate lawsuits This week, the U.S. Supreme Court declined appeals from Exxon Mobil, Suncor Energy, and Chevron to move lawsuits against them to federal courts. The companies are facing five separate cases from the state of Rhode Island and municipalities in Maryland, Colorado, California, and Hawaii. These lawsuits claim that the oil companies concealed and misrepresented the dangers of burning fossil fuels and violated consumer protection laws. Many of the municipalities seek to recover money for climate-related disasters. The Supreme Court’s denial deals a significant blow to the oil companies’ efforts—a win for the plaintiffs, who now expect the companies to defend themselves in front of local juries. Although this ruling appears to be in stark contrast to the Supreme Court’s decision last year to limit the EPA’s control over power plant emissions, it’s important to note that this is more of a procedural ruling and not indicative of the court’s position on the lawsuits at hand. Keeping these cases at the state level is likely in the best interests of all parties involved.   AI’s environmental impact The outsized impact that large AI models like ChatGPT have on the environment has gained recent attention. With their meteoric rise in tech and beyond, generative pre-trained transformers (GPT)—a subset of large language models—have come under scrutiny due to their substantial carbon emissions. MIT Technology Review found that a single large model can emit carbon equivalent to what several hundred acres of trees would sequester in a year.  The emissions from AI across sectors are now comparable to those of the aviation industry—it’s mainly due to the amount of data and computation required to train these models and the energy-intensive data centers in which they operate. Some solutions to reduce AI’s environmental impact include moving training to cleaner data centers, like those powered by hydroelectric energy, and sharing models in a more open-source manner to reduce the need for additional training. However, the proprietary nature of these models makes this less likely. There are calls for a federal body to oversee the development and use of AI, which could also help develop rules to reduce the emissions impact of AI.  While AI has the potential to compensate for its environmental impact through emissions reductions in other sectors, it’s essential to consider other environmental factors, such as water use, lithium, and environmental justice, as their proliferation continues to progress.