“Empire of AI”: Karen Hao on How AI Is Threatening Democracy & Creating a New Colonial World

The AI industry, particularly companies like OpenAI, operates like a colonial power by seizing and extracting vast resources. This includes the intellectual property of artists and writers, the data of countless individuals, and critical physical resources like land, energy, and freshwater. This extraction fuels their 'scale at all costs' vision of AI development.

Hao quotes, 'The empires of AI are not engaged in the same overt violence and brutality that marked this history, but they too seize and extract precious resources to feed their vision of artificial intelligence: the work of artists and writers, the data of countless individuals... the land, energy, and water required to house and run massive data centers and supercomputers.'

The rapid expansion of AI computational infrastructure places immense strain on global energy and water resources. McKinsey reports that within five years, AI will require two to six times the annual energy consumed by California, largely serviced by fossil fuels, leading to extended lifespans for coal plants and unlicensed methane gas turbines. Furthermore, two-thirds of new data centers are being built in water-scarce regions, tapping into public drinking water supplies.

McKenzie's report on energy demand (03:56) and Bloomberg's analysis on data center placement in water-scarce areas (05:05) are cited. Example: Elon Musk's Colossus supercomputer in Memphis uses 35 unlicensed methane gas turbines, polluting the community's air.

AI companies rely on data annotation firms in the Global South to clean and moderate vast amounts of internet-scraped data. Workers in places like Kenya are paid minimal wages to read and categorize extremely graphic and harmful content (racist, hateful, violent, sexual), leading to severe psychological trauma, mirroring the issues faced by social media content moderators.

OpenAI contracted middleman firms in Kenya where workers read 'reams of the worst text on the internet' and AI-generated harmful content to train content filters, paid 'a few bucks an hour, if at all,' resulting in deep psychological trauma.

Sam Altman, OpenAI's CEO, is characterized as a highly strategic figure who built OpenAI by leveraging Silicon Valley's 'scale at all costs' mentality. He transformed the company from a nonprofit to a for-profit entity to secure massive capital, positioning himself at the center of AI trends. He also strategically engages with governments, as seen in his participation in the Trump administration's 'Stargate project' and trips to the Middle East to secure land and energy for OpenAI's aggressive computational needs.

Altman's career trajectory from Y Combinator to OpenAI CEO, his creation of a for-profit arm within the nonprofit (14:19), his ability to raise 'hundreds of billions and even trillions' (14:41), and his politicking with the Trump administration and Gulf trip (27:15, 28:22) are cited.

AI technology, despite its actual capabilities, is perceived by executives as a tool to replace jobs, leading to widespread layoffs. Companies like OpenAI explicitly aim to develop 'highly autonomous systems that outperform humans in most economically valuable work.' This trend is breaking the career ladder, making it difficult for new college graduates to find entry-level positions, as industries opt for AI models over human workers.

OpenAI's definition of AGI (16:11) and Sam Altman's statements about AI models becoming 'senior colleagues at a law firm' (18:52) are presented as evidence of the intent to automate jobs.

US export controls aimed at hindering China's AI development have inadvertently spurred Chinese innovation. Companies like Highfly, with its DeepSeek model, have demonstrated the ability to achieve comparable AI model capabilities with two orders of magnitude less computational resources, energy, and data than their American counterparts. This disproves Silicon Valley's assertion that 'scale at all costs' is the only path to advanced AI.

Highfly's DeepSeek model cost around $6 million to train, compared to hundreds of millions or tens of billions for OpenAI models, using existing engineering sophistication rather than fundamentally new techniques.

Despite feeling a 'loss of agency,' communities are actively resisting the AI industry's extractive practices. Examples include Chilean water activists successfully blocking Google's data center for four to five years, artists suing companies for intellectual property theft, and Kenyan workers unionizing against exploitation. These actions demonstrate that communities can reclaim agency and force AI companies to adjust their approaches.

The Chilean community's activism against Google's data center (08:18), artists suing for IP (54:10), and Kenyan workers unionizing (54:20) are cited as examples of effective resistance.