The Rise of the AI Empire
So much of what is happening today in the AI industry is extremely inhumane. While some might play devil's advocate and argue that the civilization that accelerates its research with artificial intelligence will logically become the superior civilization, that is a flawed premise. This is a specific prediction being made by figures like Sam Altman and Mark Zuckerberg, and they all share a common feature: they profit enormously from this myth. Internal documents show that these companies are purposely trying to create a specific feeling within the public so they can extract and exploit resources. To address this, we need to break up the empires of AI.
The tech industry has been under scrutiny for over eight years, involving interviews with over 250 people, including former and current OpenAI employees and executives. There are many parallels between the modern empires of AI and the empires of old. These companies lay claim to the intellectual property of artists, writers, and creators in the pursuit of training their models. Furthermore, they exploit an extraordinary amount of labor, which effectively breaks the career ladder. Someone might get laid off and then find themselves working to train a model on the very job they were just replaced in, which only perpetuates more layoffs.
While tech leaders promise that new, unimaginable jobs will be created, many of the jobs currently being generated are far worse than the ones they replaced. Beyond labor, there is an environmental and public health crisis created by these companies. They spend hundreds of millions of dollars to kill any legislation that stands in their way and frequently censor researchers whose findings are inconvenient to the corporate agenda. This is not to say that these technologies don't have utility, but the current production methods are exacting a great deal of harm on people. Research shows that the same capabilities could be developed in a way that avoids these unintended consequences.
The Ambiguity of Artificial General Intelligence
To understand where we are, we have to look back at when AI started as a field in 1956. A group of scientists gathered at Dartmouth University to chase a new scientific ambition, and an assistant professor named John McCarthy decided to name the discipline "artificial intelligence." He had previously tried to call it "Automata Studies," but his colleagues were concerned that the new name pegged the discipline to recreating human intelligence. The problem then, as it is now, is that there is no scientific consensus on what human intelligence actually is. There is no singular definition from psychology, biology, or neurology.
Historically, every attempt to quantify and rank human intelligence has been driven by a desire to prove that certain groups of people are inferior to others. Because there are no goalposts for the field, companies can use the term Artificial General Intelligence (AGI) however they want. They define and redefine the term based on what is convenient for their current audience. For OpenAI, AGI is a system that cures cancer or solves poverty when talking to Congress. When talking to consumers, it is the world's best digital assistant. When dealing with Microsoft, it was defined as a system that would generate $100 billion in revenue.
OpenAI’s own website defines AGI as highly autonomous systems that outperform humans at most economically valuable work. This is not a coherent vision of a single technology; these are different definitions spoken to mobilize audiences, ward off regulation, or secure capital. Back in 2015, Sam Altman explicitly outlined the existential risk of AI, calling superhuman machine intelligence the greatest threat to human existence. However, he was often mirroring the language of Elon Musk at the time to convince him to co-found OpenAI. Altman pivoted his stated fears to match Musk’s central concerns to build trust.
Power Struggles and the Firing of Sam Altman
Elon Musk ultimately co-founded OpenAI with Altman, but he now feels manipulated, believing Altman engineered his language to gain Musk's trust. During the subsequent years, documents from the legal battle between Musk and Altman revealed that Musk was essentially muscled out of the company. Originally, Ilya Sutskever and Greg Brockman had to decide who should lead the new for-profit entity they were creating. They initially chose Musk to be the CEO, but Altman appealed personally to Brockman, a long-time friend.
Altman convinced Brockman that having someone as erratic and unpredictable as Musk in charge of a powerful future technology would be dangerous. Brockman then convinced Sutskever, and the two switched their allegiances to Altman. This led to Musk's departure because he refused to stay if he wasn't the CEO. This highlights the polarization surrounding Sam Altman; people either view him as the Steve Jobs of this generation or as a manipulative liar. It usually depends on whether that person's vision of the future aligns with Altman's goals.
The internal tension eventually led to the board firing Sam Altman. Ilya Sutskever, the chief scientist, had become deeply concerned about Altman's behavior and how it led to poor research outcomes and instability. He eventually approached an independent board member, Helen Toner, to discuss these concerns. Sutskever felt that Altman was creating a chaotic environment by pitting teams against each other and being dishonest. While some might view this as just a fast-paced startup culture, the board felt that for a company building potentially world-breaking technology, this level of instability was unacceptable.
The Myth of the Statistical Brain
A major point of contention in the AI world is the hypothesis regarding how the human brain works. Ilya Sutskever and his mentor, Geoffrey Hinton, believe that the human brain is essentially a giant statistical model. This is a hypothesis, not a proven scientific fact, yet it drives the entire industry's strategy. They believe that if they just build bigger and bigger statistical engines, they will eventually reach and exceed human intelligence. Sutskever has even used charts showing a linear relationship between brain size and species intelligence to justify this "brute force" approach.
Critics argue that this is a very reductive way to view human consciousness. However, because these companies believe the hypothesis is true, they are hoovering up massive amounts of data and building gargantuan data centers. We have to ask why we are trying to build systems that are merely duplicative of humans. The purpose of technology throughout history has been to improve human flourishing, not to replace people entirely. Instead of building "everything machines," we could be focusing on specific AI systems that accelerate drug discovery or improve healthcare without the unintended consequences of trying to mimic a digital brain.
The "empire" metaphor is the only one that truly encapsulates the scale of these companies. They lay claim to resources that aren't theirs, such as personal data and the creative works of artists. They also exploit global labor, contracting hundreds of thousands of workers to label data in grueling conditions. Furthermore, they monopolize knowledge by bankrolling the majority of the world's AI researchers. If most climate scientists were funded by fossil fuel companies, we wouldn't trust their findings; similarly, the AI industry often censors researchers, like Dr. Timnit Gebru, whose work is inconvenient to their agenda.
Labor Exploitation and Data Annotation
The reality of AI development is far less "magic" and far more labor-intensive than the public realizes. Much of the progress depends on data annotation, where human workers teach models how to respond. For example, before ChatGPT could function as a chatbot, thousands of people had to manually type responses to show the model how to hold a dialogue. This work is often outsourced to third-party firms that pit workers against each other to get the tasks done as cheaply as possible.
Many highly educated professionals, including lawyers, doctors, and award-winning directors, are now doing this work secretly to put food on the table after being laid off. This creates a "monster" of an industry where workers feel mechanized and atomized. One worker recounted how the pressure to complete tasks was so high that she found herself screaming at her child for a distraction, feeling that her humanity was being squeezed out by the machine she was helping to build. This breaks the career ladder because entry-level and mid-tier jobs are being gouged out, leaving only high-order orchestrators and low-order manual laborers.
When companies like Klarna announce that they have reduced their workforce from 7,000 to 3,000 through the use of AI, they often claim "natural attrition." However, the CEO of Klarna, Sebastian Siemiatkowski, noted that while AI handles 70% of customer service, they still see a future where human interaction is regarded as a premium VIP service. While he remains optimistic about a richer society, the short-term reality is that many people are losing their livelihoods and sense of dignity as their expertise is harvested to train the models that replace them.
Environmental Costs and the Path Forward
The physical infrastructure of AI is creating a public health crisis. Companies are building colossal supercomputer facilities, like the Stargate initiative, which require massive amounts of power and water. A single facility in Abilene, Texas, could consume more than 20% of the power required by a city the size of New York. In Memphis, Elon Musk built the Colossus supercomputer using 35 methane gas turbines, which pumped toxins into a working-class community that wasn't even notified of the project. These communities now face increased utility bills, decreased grid reliability, and worsening respiratory illnesses.
We should think of AI like we think of transportation. Currently, we are building "rockets"—resource-heavy models that exact an extraordinary cost. Instead, we should be building "bicycles of AI." An example of this is AlphaFold, which uses small, curated datasets to predict protein folding. It provides enormous scientific benefits without the massive environmental footprint or the need for exploitative labor. We don't need a "rocket" to solve every problem; we need efficient, specialized tools that serve the public good.
The goal is not to get rid of AI entirely, but to stop companies from operating as empires. An empire is predicated on the idea of extraction without fair exchange. We see grassroots movements already pushing back, with people protesting data centers and artists suing for their intellectual property. We need to break up these monopolies and forge new paths for AI development that are broadly beneficial. While AI can be an incredible tool for entrepreneurs, we must be intentional about its social and environmental impact to ensure it actually improves the human condition rather than diminishing it.