Karen Hao on AI's Imperial Agenda and the Nightmares of OpenAI

AI's Resemblance to Colonial Empires

• 🎯 AI companies are compared to historical colonial empires due to their consolidation of wealth and power by taking resources not their own, including intellectual property and private data without compensation.
• 💰 Empires exploit labor without sufficient pay to amass wealth, a parallel seen in the AI industry where data laborers in places like Kenya earn minimal wages for crucial annotation tasks.
• 🔒 A third parallel is the control of information flows, where AI research agendas are distorted by corporate interests, leading to censorship of inconvenient truths.
• 🚀 Finally, empires justify their existence through a narrative of moral or existential imperative, framing themselves as a "good empire" bringing progress, often by demonizing a competing "evil empire."

The Pace and Scale of AI Infrastructure

• ⚡ The speed of AI development, moving at the pace of "bits" rather than ships, is a key differentiator from classical empires.
• 🤯 Sam Altman's goal to build 250 gigawatts of data centers by 2033, estimated to cost $10 trillion, is highlighted as an almost incomprehensible scale, equivalent to nearly four dozen New York Cities of data centers.
• 🏢 Meta's plans for supercomputer facilities the size of Manhattan further illustrate the unprecedented infrastructure buildout controlled by a small group.

Environmental and Public Health Harms

• 🏭 Energy for AI infrastructure often comes from fossil fuels, including natural gas and coal, extending the life of polluting plants and providing a lifeline to the fossil fuel industry.
• 💧 Data centers require significant amounts of fresh water for cooling, often competing with communities in already water-scarce areas, leading to environmental and public health crises.
• ☣️ The expansion of AI infrastructure is accelerating existing crises, with potential for increased emissions, pollutants, and strain on vital resources.

The Myth of Superintelligence and AGI

• 🤔 The concept of superintelligence is critiqued as scientifically undefined and useful for executives to market their companies, as is Artificial General Intelligence (AGI).
• 🎭 Tech CEOs like Sam Altman fluidly redefine AGI based on their audience, using it to sell products, secure deals, or ward off regulation.
• 📉 Early explorations of redistributing AI windfall profits via a tax structure for universal basic income have been abandoned, replaced by a K-shaped economy where AI stocks soar while average Americans struggle.

Challenging the AI Empire

• 💡 The narrative that AI's trajectory is inevitable is challenged; the scale of current AI development is technically unnecessary.
• 🚲 A shift from building "rockets" (large-scale, exploitative AI) to "bicycles" (nuanced, beneficial AI) is proposed as a path to unlocking AI's benefits without its costs.
• ✊ Despite a lack of government regulation, collective action through protests, lawsuits, and community engagement is proving effective in blocking data center projects and demanding accountability from AI companies.
• 💬 Individuals have an active role in shaping AI's future by demanding AI systems that are not built on exploitation and extraction.