Empire of AI: OpenAI, Colonialism Lens, and the Environmental Costs of Big Data Centers

In this Science Quickly episode, Kendra Pierre-Louis interviews Karen Howe about Empire of AI, a frame that views OpenAI and similar firms as modern empires driven by capital, ideology, and global data extraction. The conversation covers the scale and risks of AI development, especially environmental impacts from data centers and the exploitation of workers in the AI supply chain in places like Kenya and Bangladesh. They discuss governance gaps, the potential and limits of regulation, and a hopeful alternative built on specialized AI tools that support human decision-making rather than pursuing a speculative AGI. The episode blends analysis with international perspectives and a critical lens on power in tech.

Overview and the Empire Lens

The interview centers on Empire of AI, a framework that treats major AI firms as empires with vast economic and political reach. Karen Howe argues that these empires accumulate power through dispossession and extract resources, data, and labor from around the world. The conversation situates OpenAI within a broader critique of capitalism, ideology, and the pursuit of artificial general intelligence, challenging the idea that AGI is a neutral, inevitable solution to human problems. A key point is that OpenAI and peers are not merely building tools; they are shaping social and political structures through resource extraction and control over data, labor, and narrative.

"An enormous amount of economic and political power and that's specifically that they amass that power through the dispossession of the majority of the rest of the world." - Karen Howe

Environmental and Labor Realities

The discussion then pivots to environmental costs and the human labor behind AI. Howe highlights the energy demands of data centers and the use of fossil fuels if growth continues unchecked, with environmental justice concerns for communities near facilities. She cites rising water use and the location of new centers in water-scarce regions, along with air and water pollution from power sources such as methane-fired turbines. These realities are not speculative but evident in current projects, including what workers in places like Kenya and Bangladesh experience in content moderation and reinforcement learning from human feedback processes. The labor conditions described are exploitative, with workers enduring psychological strain and precarious incomes while generating vast value for tech firms.

"This is a heavy Industry, like this is, it is extremely toxic to the environment and to public health around the world." - Karen Howe

Regulation, Accountability, and a Path Forward

Despite regulatory dissatisfaction, Howe argues for bottoms-up governance and public accountability as crucial responses in the absence of decisive policy action. The conversation includes international perspectives and historical parallels, such as labor standards movements that used consumer pressure to improve conditions. Howe emphasizes that even if AGI remains unlikely or distant, the present harms require immediate action. She outlines a hopeful alternative in specialized AI designed to address climate, energy, and other societal challenges, rather than chasing a universal, power-consolidating general intelligence. The final sections explore how governance can evolve through public pressure, corporate responsibility, and the deployment of climate-focused AI tools that respect local contexts and culture.

"the possibility of an artificial general intelligence emerging and solving everything is astronomically small" - Karen Howe