Rethinking the Inevitability of AI: Historicizing Climate Impact and Discussing Best Practices for Energy and Water Consumption

For all that artificial intelligence (AI) does, and claims to do, it is still currently a technology whose potential value is dwarfed by its costs—both economic, and more importantly, environmental. The discussions surrounding AI on Grounds and in popular discourse have understandably focused on potential social and labor effects. They nod to the massive environmental costs and harms of these systems, but rarely meaningfully integrate these with questions surrounding AI’s larger utility. As a result, these costs are often not considered, or even measured, when establishing best practices regarding how, whether, when, and how much to adopt and use general AI.

This project will consider what could and potentially should be the state of AI in U.S. society and globally in five years, 10 years, or 20 years, instead of taking it as a given that AI will be left to proliferate along the lines of earlier technologies, driven primarily by market forces and military spending. This project is interdisciplinary in nature, using a historical and STS approach to research and frame the comparisons and costs of AI, while remaining mindful of the legal, economic, and social frameworks for this technology. The goal of this project is to make more transparent and legible the unspoken ramifications of a relatively new and incredibly resource-hungry technology whose widespread use is being presented as a given for the future of academia, industry, government, and even the arts. This project questions the ahistorical and presentist framing of AI’s potential benefits, focusing on comparing and framing the environmental costs of this technology with what has come before this latest burst in development and usage.

We will write a history of AI’s environmental impacts that focuses on energy and water use, as well as resource allocation and extraction required for the manufacture, sale, and shipping of underlying components. Our goal in this research is to contextualize the seemingly dramatic and unprecedented shift in computing resources and natural resources towards a new digital product that is being marketed as a panacea for the concept of human work itself, even as it remains largely untested. We will then host a public workshop and conference in Spring/Summer of 2024 to bring together an interdisciplinary group of researchers, who are doing leading-edge work on this topic and associated topics, both on Grounds and at other institutions. In keeping with environmentally conscious ethos of project this workshop will not require attendees to travel to UVA to participate and will be convened as a hybrid-online conference.