This article examines the rise of the AI industry through the lens of journalist Karen Hao, author of Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI. Drawing on over 300 interviews, including 90 with current and former OpenAI employees, Hao argues that the current trajectory of artificial intelligence is being driven by an "imperial agenda" that prioritizes profit and consolidation over human flourishing.
The Myth of the "AI Empire"
Hao posits that major AI companies operate similarly to historical empires. They claim ownership over resources that do not belong to them—such as the intellectual property of artists and writers—and engage in massive "land grabs" to build the infrastructure required for their models.
Beyond resource acquisition, Hao identifies several core characteristics of these AI empires:
- Labor Exploitation: These companies rely on hundreds of thousands of workers worldwide to annotate data, often under grueling conditions. Furthermore, they design tools specifically to automate away human jobs, eroding labor rights.
- Monopolization of Knowledge: By bankrolling the majority of AI research, these corporations set the agenda for what is studied and often censor or silence researchers who produce findings inconvenient to their business models.
- Strategic Myth-Making: Executives use narratives of existential risk—often pitting themselves against an "evil" competitor like China or a rival corporation—to justify their need for unchecked power and resources.
The Power Struggle at OpenAI
Hao’s research highlights the internal volatility at OpenAI, particularly regarding the leadership of CEO Sam Altman. She describes a company culture characterized by intense polarization; employees and peers view Altman either as a transformative tech visionary or as a deeply manipulative figure.
The 2023 attempt to oust Altman was not, according to Hao, an arbitrary act. It stemmed from deep-seated concerns among board members and senior executives—including former chief scientist Ilya Sutskever—that Altman’s leadership was creating instability and that he was steering the company toward research outcomes that undermined its original mission of safety and public benefit.
The "Statistical Engine" Hypothesis
A central theme in Hao’s critique is the scientific assumption underpinning the current AI race. Many industry leaders, including those at OpenAI and DeepMind, subscribe to the hypothesis that the human brain is simply a "statistical engine."
Hao argues that this is a dangerous reduction. By treating human intelligence as something to be replicated by building larger and larger statistical models, these companies ignore the "jagged frontier" of AI—where models excel at specific tasks but fail to demonstrate true, general human intelligence. This pursuit of AGI (Artificial General Intelligence) is often defined inconsistently depending on whether the audience is Congress, consumers, or investors, allowing companies to pivot their goals whenever it serves their financial interests.
The Hidden Human Cost
The drive for AI dominance is currently exacting a significant toll on both the environment and vulnerable communities.
- Environmental Impact: Massive data centers required for training models consume gigawatts of power and vast amounts of fresh water, often in communities that are already facing resource scarcity or environmental justice issues.
- The Data Annotation Trap: As companies automate professional roles, many laid-off workers are finding themselves in the "data annotation" industry—a lower-wage sector where they are treated as extensions of the machines they are helping to build.
Hao notes that these workers often report feelings of dehumanization, anxiety, and a loss of agency, as their labor is harvested to accelerate the very automation that threatens their future employment.
Reimagining the Innovation Ecosystem
Hao rejects the idea that the only options are to "race at all costs" or to stop technological progress entirely. Instead, she advocates for a shift in how we build AI:
- "Bicycles" vs. "Rockets": Rather than focusing solely on massive, resource-heavy models (the "rockets" of AI), Hao calls for a focus on more efficient, targeted tools (the "bicycles"). She cites DeepMind’s AlphaFold—which revolutionized drug discovery using smaller, curated datasets—as a model for sustainable innovation.
- Democratic Accountability: Hao emphasizes that the public is not powerless. She points to the 80% of Americans who support AI regulation, as well as the grassroots protests against data centers and the legal challenges from artists and families, as proof that democratic contestation is working.
- Establishing Alternatives: The path forward involves breaking up the AI monopolies and creating an ecosystem where technology is designed to improve human flourishing rather than replace human dignity.
"The goal is not that we completely get rid of this technology," Hao concludes. "The goal is that these companies need to stop being empires."