Towards the sustainability of AI infrastructure

This research will explore how normative conceptions of sustainability within the AI industry shape corporate governance and policymaking on the environmental impacts of Artificial Intelligence (AI) in the US, EU and the UK. This is an important academic and policy question as AI is increasing deployed across social and political realms of society with little consideration of its environmental impact.

These include AI's large carbon impact through training and development, to the energy, water consumption and mineral extraction required to build and power its computational hardware and infrastructure. My work will intervene Science and Technology Studies (STS) and digital anthropological approaches to studying technology cultures, while linking theoretical insights to grounded discussions in AI ethics and policy.

I will extend existing work which has focused on carbon emissions rather than the broader environmental impacts of AI infrastructure, while bridging the gap between theoretical and empirical research, and the practice of AI. As nation-states pursue unprecedented investment into AI across the globe, fuelled by so-called "AI race" dynamics, my project addresses the timely and pertinent issue of AI infrastructure and climate change.

I pose three research questions: How do corporate and state actors in the AI industry make sense of the environmental impacts of AI? How does this shape the practices of AI governance? And what policy recommendations can be drawn from the local politics of AI in the broader governance of digital infrastructure in the age of AI?

To answer these questions I draw from digital ethnographic methods to explore normative conceptions of sustainability amongst three social groups in the Ai industry within the US, EU and the UK: 1) AI researchers, engineers and governance practitioners within industry, 2) AI policymakers within the government and civil society, and 3) local community groups who are impacted by the environmental impacts of AI infrastructure. I will use thematic analysis to analyse existing governance frameworks and policies that attempt to regulate AI, to map the experiences of these social groups to gaps in existing governance.

M project will provide novel theorisation of AI's environmental impacts visa-vis its infrastructure, questioning how engagement with local politics can inform policy approaches in pursuit of sustainable AI. The implications of this work extend beyond the realm of AI and its infrastructure to broader academic and policy questions about AI ethics and the governance of digital infrastructure during the current climate crisis.

I will conduct overseas fieldwork in the US and Europe (e.g.Spain and Germany). This will involve ethnographic engagement with AI researchers, policymakers and local communities impacted by AI infrastructure. This will involve interviews,

participant observation and involvement in the communities around AI companies, government organisations and research institutes, and NGOs working with local communities on datacentre resistance. I may also explore interviews and fieldwork in Southeast Asia where there is increasing community organisation around data centre investment by US AI companies, however this is pending on scope and feasibility of the research project.