Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI

Hao’s core argument in Empire of AI is that OpenAI and other big tech companies pursue AI technology as a form of neocolonialism. OpenAI established the Western model of AI development, in which companies race to gobble up as much data as possible in pursuit of the possibly chimerical goal of AGI. Drawing on the work of Stanford AI researcher Ria Kalluri, Hao argues that to achieve the “scale” required to stay ahead of its competitors, OpenAI has controlled “knowledge, resources, and influence” (418) in a way similar to historical colonial models of exploitation.

Hao illustrates how OpenAI has shaped AI research through its intense focus on scale and AGI, leading to the neglect of other subject areas like efficacy and bias across the field. The company’s allocation of computing power is a clear example of how it uses its resources to push its interests and agenda. Hao notes that OpenAI had a series of projects dedicated to improving the efficiency of its models. However, “the project ate up significant computational resources” so it was “scrapped” in favor of the continuing development of commercial products based on scale (269). She also argues that OpenAI’s concentrated resources shape the research field, as universities are unable to compete with the scale of information OpenAI controls, leading academic researchers to leave academia for to the private sector and “atrophying independent academic research” (133).