Investigative journalist Karen Hao reveals the unsettling parallels between modern AI giants and the exploitative empires of history. From the boardroom drama at OpenAI to the grueling reality of data annotation, this article exposes how the “AI Revolution” is being built on the backs of a new digital underclass and at the expense of our planet.

Core Question: Are we building a utopian future for humanity, or are we simply enabling a new form of digital imperialism that centralizes power in the hands of a few tech elites?

Highlights

• The strategic “myth-making” behind AGI and how definitions are shifted to manipulate regulators and investors.
• The “Digital Underclass”: Why highly skilled professionals are being replaced by high-pressure data annotation roles.
• Environmental Racism: The massive energy and water demands of AI “super-facilities” located in vulnerable communities.
• “Bicycles vs. Rockets”: The argument for specialized, efficient AI tools over resource-heavy, generalized “everything machines.”

⏱️ Reading time: approx. 12 minutes · Saves you about 100 minutes vs. watching.

Section 1: The Architect of the Empire

Sam Altman and the Power Struggles of OpenAI

The history of OpenAI is less a story of pure scientific pursuit and more a series of calculated political maneuvers. Karen Hao, through over 300 interviews, describes a “cult of personality” surrounding Sam Altman. She notes that those within the industry are polarized: some see him as the Steve Jobs of his generation, while others view him as a master manipulator who leverages “existential risk” narratives to consolidate control. The drama of Altman’s brief firing by the OpenAI board wasn’t about a single event, but a culmination of concerns regarding his transparency and a perceived pattern of pitting teams against each other.

A significant part of the empire’s power lies in the ambiguity of “Artificial General Intelligence” (AGI). Since the field’s inception in 1956, there has been no scientific consensus on what “intelligence” actually is. This allows companies like OpenAI to redefine AGI whenever convenient. When speaking to Congress, it is a tool to cure cancer; when speaking to Microsoft, it is a revenue engine; and when speaking to consumers, it is a “dazzling assistant.” By keeping the goalposts moving, these empires avoid accountability while maintaining a sense of divine mission.

The relationship between Altman and early co-founders like Elon Musk highlights the “imperial” tactic of using shared fears to mobilize resources. Altman reportedly mirrored Musk’s alarmist language about AI “summoning the demon” to secure initial funding and partnership, only to later muscle him out when the transition to a for-profit entity began. This pattern of shifting allegiances and “dangling carrots” of access is how the empire maintains its narrative and quashes dissent from journalists and researchers alike.

💡 Digging Deeper

Q: Why did the board actually fire Sam Altman?
A: According to Hao’s sources, it was due to “instability” and a lack of transparency regarding Altman’s side projects, like the OpenAI startup fund which was actually under his personal name, and a culture of internal competition that hindered safety.

Q: What is the “myth of the demon”?
A: It is the narrative that AI could destroy the world, used by tech leaders to argue that only they should be trusted to build it safely, effectively barring others from participating in the development process.

Q: How does OpenAI control the media?
A: They use “access journalism” as a carrot, rewarding favorable coverage with interviews and withholding it from critics, which Hao experienced firsthand after her critical reporting in 2020.

Section 2: The New Imperialism

Land Grabs, Labor Exploitation, and Environmental Costs

The “Empire of AI” is not just a metaphor; it describes the physical and social footprint of companies like Google, Meta, and OpenAI. These entities engage in “land grabs” to build massive supercomputer facilities, often in rural or marginalized communities. A prime example is the “Stargate” initiative—a $500 billion plan for AI infrastructure. These facilities, like the one in Abilene, Texas, or Musk’s “Colossus” in Memphis, consume more power and water than mid-sized cities, often decreasing grid reliability for local residents and polluting the air with methane turbines.

The labor required to build these systems is equally exploitative. Behind the “magic” of ChatGPT are hundreds of thousands of data annotators. This is the “hidden” labor force that manually labels images and corrects text so the models can “learn.” These workers are often pitted against each other in a gig-economy race to the bottom, losing their dignity as they are treated more like biological components of a machine than human beings with expertise.

Furthermore, these empires monopolize knowledge production. By bankrolling the majority of AI researchers, the industry ensures that the scientific agenda favors their profitable “brute force” scaling methods. Researchers who find “inconvenient” truths—such as the environmental costs or the inherent biases in Large Language Models (LLMs)—are often censored or fired, as was the case with Dr. Timnit Gebru at Google.

💡 Digging Deeper

Q: What is “data annotation”?
A: It is the process where humans manually label data (e.g., “this is a pedestrian,” “this is an angry tone”) so AI models can recognize patterns. Without this human labor, modern AI wouldn’t exist.

Q: How is AI affecting local environments?
A: Data centers require millions of gallons of water for cooling and massive amounts of electricity, which can lead to higher utility bills and toxins in the air for nearby communities.

Q: Why does Hao call them “empires”?
A: Because they claim resources (data/IP) that aren’t theirs, exploit labor without fair exchange, and project their worldview onto billions of people without democratic consent.

Section 3: The Hollowing of the Middle Class

The Broken Career Ladder and the “Bicycle” Alternative

The impact of AI on employment is often framed as a binary: either it takes all jobs or creates new ones. The reality is more nuanced and concerning. AI is “gouging out” the middle of the career ladder. Entry-level and mid-tier roles in finance, law, and media are being automated, while the new jobs being created are often at the extremes: either very high-level “orchestrators” or very low-level, high-stress data annotators. This “broken ladder” makes it nearly impossible for new graduates to gain the experience needed for senior roles.

However, the “brute force” scaling approach favored by the empires is not the only way. Hao advocates for the “Bicycles of AI.” While current LLMs are like “rockets”—expensive, resource-heavy, and prone to “hallucinations”—specialized tools like DeepMind’s AlphaFold are “bicycles.” These tools use curated, smaller datasets to solve specific problems, like protein folding, with far less environmental impact and far higher societal utility.

The future of work may rely on our ability to prioritize “irreplaceably human” skills. While AI can handle spreadsheets and code, it cannot replicate real-life connection, empathy, and community building. We are seeing a “post-peak social media” trend where younger generations are moving toward “dark social” and IRL (In Real Life) experiences. The ultimate goal should be to break up the AI empires so that technology serves to enhance these human connections rather than replace them.

💡 Digging Deeper

Q: What did the CEO of Klarna say about AI and jobs?
A: Sebastian Siemiatkowski noted that his company shrunk from 7,000 to 3,000 employees while revenue doubled, proving that AI can handle 70% of customer service, though he remains “optimistic” about the long-term.

Q: Why are highly educated people doing data annotation?
A: As AI displaces white-collar jobs, many PhDs and doctors find themselves with no other options, working for third-party firms to train the very models that replaced them.

Q: What is a “Bicycle of AI”?
A: A specialized AI tool built on a small, curated dataset for a specific task (like medical diagnosis) rather than a general-purpose “everything machine” that requires massive resources.

Key Takeaways

The AI industry, led by figures like Sam Altman, is currently operating under an “imperial” framework. This involves extracting value from artists (IP), laborers (data annotation), and the environment (water and power) without a fair exchange. The narrative of “existential risk” is often used as a tool for regulatory capture, ensuring that only a few giants have the right to build these transformative technologies.

True progress will require breaking up these digital empires and shifting our focus toward specialized, efficient AI tools. We must advocate for democratic participation in AI governance and protect the “human rungs” of the career ladder. The goal of technology should be human flourishing and connection—not the mechanization of our lives for the profit of a few.

Ultimately, the power to change the trajectory of AI lies in our collective refusal to let these companies operate flawlessly without scrutiny. Whether through withholding data, protesting local data centers, or demanding regulation, the public must reassert its agency to ensure that the “Nightmares” of AI don’t overshadow its “Dreams.”

Q&A

Q1: Is AGI actually near?
A: According to Hao, the definition is so vague that companies can claim they’ve reached it whenever they want. Scientifically, we still don’t even have a consensus definition of human intelligence.

Q2: How much power do these AI data centers use?
A: Large-scale projects like Meta’s in Louisiana or OpenAI’s in Texas can consume as much power as 20% to 50% of the average demand of New York City.

Q3: What happened to the researchers who criticized Big AI?
A: Many, like Timnit Gebru and Margaret Mitchell, were fired or forced out of companies like Google after publishing research that highlighted the risks and costs of large language models.

Q4: Will AI replace surgeons and lawyers?
A: While tech leaders predict total replacement, Hao argues that the best outcomes involve AI as a “bicycle” tool in the hands of human experts, rather than a full replacement.

Q5: What is “Stargate”?
A: A massive initiative involving $500 billion in investment for AI computing infrastructure, aimed at ensuring the US maintains dominance over rival nations like China.

Q6: Why is data annotation described as “inhumane”?
A: Workers are often monitored via Slack, paid low wages, and forced to work at extreme speeds with no breaks, causing high anxiety and a loss of personal agency.

Q7: Can we stop the “AI Empire”?
A: Hao suggests it’s not about stopping the technology, but about regulating the “imperial” business practices—breaking up monopolies, protecting IP, and forcing a fair exchange of value.