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

In the Prologue, Karen Hao uses a recent key anecdote in OpenAI’s history to introduce her central argument about AI development, which she will develop throughout Empire of AI. This structure is typical of reported news stories, which often open with an eye-catching anecdote exemplifying the overall topic of the article. Hao is an experienced reporter, and she uses existing reporting as well as her own to develop a moment-by-moment account of the events around the OpenAI board’s unsuccessful attempt to unseat Sam Altman as CEO of OpenAI. Her account of corporate intrigue uses a personal, intimate lens, including direct quotes and descriptions of peoples’ affects, as when she writes, “Sutskever looked solemn” (3) or when she quotes the chief operating officer (COO) in full, complete with interjections that connote uncertainty, “The tender—we’re, um, we’re going to see” (4). This level of detail is supported with extensive endnotes about existing reporting and a description of the sourcing of Hao’s own reporting, based on interviews with “eleven people who were present across each of the scenes recounted” (427). This anecdote illustrates the dominance of Sam Altman within the AI sector, the tensions between factions involved with AI development, and The Need for Accountability in Big Tech, as Altman’s status as an idol within the field of AI development makes it difficult for others to hold him accountable for ethical failures. 

The first half of the Prologue is written in a typical investigative reporting mode. Hao uses sourcing to describe the day-to-day events of an attempted corporate coup. This form of writing and reporting would not be out of place in a traditional business news source like The Wall Street Journal or corporate trade history. The second half of the Prologue takes a different tack and tone. Hao abandons the traditional “objective” reporting posture to take a more activist approach to the material. The shift to a more subjective approach is increasingly common in reporting, especially by a younger generation of writers. As Masha Gessen, an award-winning New Yorker writer and an expert on authoritarianism both in their native Russia and in the US, states, “moral clarity would be a much better guiding ideal for journalism than objectivity” (Hond, Paul. “Is Objectivity in Journalism Even Possible?” Columbia Magazine, 2023). Thus, Hao concludes her prologue with a moral argument about the neocolonial aspects of AI development in Silicon Valley and the need to recognize the labor and resource abuses inherent in a model created and perpetuated by wealthy white men who see themselves as the saviors of humanity. She also, controversially within the AI field, states that “AGI is largely rhetorical—a fantastical, all-purpose excuse for OpenAI to continue pushing for ever more wealth and power” (19). This claim underpins her argument for Redefining AI Safety Around Present-Day Harms. The vocal handwringing of industry leaders about the apocalyptic dangers of AGI are largely a distraction from more pressing and real concerns, including the huge environmental costs of AI data centers and the exploitation of workers.

Hao uses her unique professional background to craft her argument. Hao is a proponent of AI technology and has a history of working in the tech field herself. Because she has written favorably about AI in the past, early on in her reporting of AI she had the ability to conduct what is known as “access journalism,” meaning she was given access to leaders in the field like Sam Altman because they believed she would write favorably about them. However, she also pursued research outside of the Silicon Valley bubble to directly report about the environmental and labor issues with AI in the Global South. This provides her with both an insider and outsider view of AI systems, which is reflected in the bifurcated structure of the Prologue and is a tendency throughout Empire of AI.