Browse
Log in / Register
Share
Date: March 18, 2025
From the University of Sydney’s “Thinking Outside the Box” series, Professor Matthew Beck challenges the rosy narratives surrounding generative AI. While big tech touts efficiency and productivity gains, Beck argues that the hidden costs—biased training data, skyrocketing resource consumption, job losses, and diminished accountability—demand far more scrutiny. This piece, adapted for a U.S. audience, weaves in recent American developments to examine whether we are building a future for machines at the expense of people.
The promise of generative AI often begins with a clean slate: a digital brain free of human error. Yet the reality is messier. Much of the data used to train large language models is historical, meaning it carries the same race, gender, and age prejudices that have long plagued society. High-profile incidents—such as a leading AI chatbot quickly adopting racist and conspiratorial language—have shown how easily these systems absorb bad habits from flawed data. In the United States, an automated benefits system in several states has incorrectly flagged thousands of applicants for fraud, echoing the same “algorithmic injustice” that plagued Australia’s Robodebt scandal. When a machine is treated as an objective authority, the burden of proof shifts to the human—often with devastating consequences.
Behind every generative AI tool lies a data center, and those centers are insatiable. In Northern Virginia, the world’s largest data center hub, competition for land has pushed up prices and strained local water supplies. A single large AI training run can consume as much electricity as hundreds of homes use in a year, and the cooling needs of these facilities drain millions of gallons of water daily. In drought-prone regions like the American Southwest, the question is no longer hypothetical: should limited water and energy go to human communities or to machine expansion? Recent studies show that data centers in areas like Loudoun County are already contributing to higher utility rates and environmental degradation, yet there are no federal mandates requiring them to report water consumption.
OpenAI CEO Sam Altman has proposed that artificial intelligence may one day be treated like electricity or water—metered, paid for, and provided as a monopoly utility. It is a revealing idea. Public utilities exist because governments recognize certain services as essential and too important to leave entirely to private profit. Yet Altman’s vision imagines the reverse: privately owned AI systems becoming indispensable while the public funds the infrastructure—subsidized land, tax breaks, energy grants—and corporations keep the value. In the U.S., this dynamic is already visible: Amazon, Google, and Microsoft secure massive tax incentives for data centers while local communities see few direct benefits. The god complex is real, and it is being subsidized with taxpayer dollars.
Generative AI is reshaping the labor force faster than any policy can respond. In 2024 alone, U.S. tech companies cut over 250,000 jobs while simultaneously investing billions in automation. AT&T, Google, and Microsoft have all announced layoffs tied to AI efficiency gains. Meanwhile, the retail and financial sectors are testing chatbots and robotic process automation to replace customer service roles. The result is not merely job displacement—it is a structural shift. As displaced workers retrain and flood into healthcare, skilled trades, and personal services, those sectors face downward pressure on wages and rising underemployment. A future without work may sound idyllic, but research shows that work provides routine, social connection, and a sense of purpose that are central to well-being.
Who pays for the massive retraining effort needed? In the U.S., community colleges and public universities are already stretched thin, yet they are expected to absorb millions of displaced workers. At the same time, federal funding for higher education has not kept pace with inflation. Private companies, which reap the rewards of automation, often avoid the cost of retraining. That bill falls on taxpayers and already strained institutions. It is a classic case of privatizing gains and socializing losses—a pattern that has played out from the decline of manufacturing to the rise of the gig economy.
Every few months, another major organization confirms it cannot protect the data it already holds. In 2024, a breach at a leading health insurer exposed records of over 50 million Americans. Now generative AI is adding a new layer of risk: AI-powered scams can clone voices, write convincing phishing emails, and automate cyberattacks at scale. The same technology that fuels productivity also arms malicious actors. Anthropic has reportedly developed AI so capable at cyber operations that it was deemed too dangerous to release. As our digital footprints grow—passwords, biometrics, health records, financial histories, voice prints—so does the surface area for exploitation. The question looms: will we one day retreat from digital convenience because the risk outweighs the reward?
None of this means AI should be rejected outright. Generative AI has legitimate benefits, from accelerating drug discovery to automating rote tasks. But the current trajectory—fueled by corporate profits and a narrative of inevitable progress—deserves active, informed skepticism. The decisions made today—about regulation, education, taxation, and resource allocation—will shape whether AI serves human flourishing or simply deepens inequality. We are not passive observers. We can decide what kind of future we accept, and what kind we refuse.
The hardest question remains: if machines take over most productive work, and the benefits are hoarded by a minority of owners, who will buy the goods and services? Can you barter with a machine? And if so, what would a machine want?
