Introduction: Beneath the gleaming facade of Artificial General Intelligence, a new empire is rising, powered by unprecedented capital and an almost religious fervor. But as billions are poured into a future many experts doubt will ever arrive, we must ask: at what cost are these digital cathedrals being built, and who truly benefits?

Key Points

• The “benefit all humanity” promise of AGI functions primarily as an imperial ideology, justifying the consolidation of immense corporate power and resource extraction rather than serving as a verifiable scientific goal.
• OpenAI’s “winner-takes-all” mentality has fundamentally distorted AI research, prioritizing brute-force scaling via vast data and compute over fundamental algorithmic innovation, efficiency, or safety.
• Despite astronomical investments and the capture of top research talent by corporations, the tangible benefits of AGI remain largely unfulfilled, while real-world harms—from exploited labor to environmental strain and societal disruption—are increasingly evident.

In-Depth Analysis

AGI, as envisioned by its most zealous proponents, isn’t merely a technological frontier; it’s an economic and political strategy masquerading as a benevolent mission. Karen Hao’s “empire” metaphor is less hyperbole and more a critical lens through which to understand the “why” of today’s AI industry: the relentless pursuit of power, control, and resource extraction, all cloaked in the noble rhetoric of elevating humanity. This isn’t groundbreaking in the history of grand technological ambitions, but the scale, speed, and almost messianic fervor surrounding AGI make it uniquely potent.

OpenAI’s founding mission, ostensibly for “beneficial AGI,” has, by many accounts, morphed into a quest for market dominance where “speed over anything else” is the governing principle. This isn’t innovation; it’s a brute-force siege on the computational limits of the planet. The “intellectually cheap” path of throwing ever more data and compute at existing techniques isn’t a testament to ingenuity, but to the availability of an almost limitless supply of capital. This approach, while effective in generating impressive demos, bypasses the arduous work of fundamental algorithmic breakthroughs and sidesteps critical questions of efficiency, ethics, and long-term sustainability. It’s the technological equivalent of building a skyscraper by just stacking more bricks faster, rather than redesigning the foundation for stability and resilience.

The consequences of this corporate-driven, speed-obsessed trajectory are stark. Academia, once the wellspring of diverse and independent AI research, has been hollowed out, its brightest minds absorbed into corporate labs. The entire discipline’s agenda now reflects corporate priorities, stifling truly scientific exploration in favor of product roadmaps and market share. The irony is profound: while tens of billions are poured into this AGI quest, tangible, verifiable benefits to “all humanity” remain largely theoretical. Instead, we see the very real collateral damage: low-wage workers traumatized by horrific content, energy grids strained to their limits, and societal fabrics frayed by job displacement and the proliferation of convincing, yet delusory, AI outputs. Hao’s pointed example of Google DeepMind’s AlphaFold, a genuinely transformative AI developed with substantially less infrastructure and without the associated ethical quagmire, stands as a stark indictment of the industry’s preferred path. The “empire” isn’t just expanding; it’s externalizing its costs onto the global commons and vulnerable populations, all while preaching a gospel of future abundance that feels increasingly out of touch with present realities.

Contrasting Viewpoint

While critics like Hao paint a stark picture of unchecked ambition and mounting harms, proponents of the current AI trajectory would argue that such analyses overlook the undeniable, immediate benefits already realized. Large Language Models (LLMs) like ChatGPT, despite their imperfections, have demonstrably boosted productivity across various sectors, automating mundane tasks and accelerating research. The sheer accessibility of these tools has democratized capabilities previously reserved for specialists, empowering millions. They would contend that the pursuit of AGI, while ambitious, is a necessary moonshot – a grand challenge that drives innovation and pushes the boundaries of human potential, with the promise of unlocking solutions to humanity’s most complex problems, from climate change to disease. The substantial investments, they’d argue, are merely the cost of pioneering a transformative technology, and the associated harms are either temporary growing pains or solvable challenges that will be addressed as the technology matures and regulatory frameworks catch up. Furthermore, many still firmly believe in the geopolitical necessity of leading the AI race, framing it as a critical component of national security and economic competitiveness in a rapidly evolving global landscape.

Future Outlook

Looking ahead 1-2 years, the current trajectory suggests an unwavering commitment to scaling, with companies continuing to pour billions into larger models and more infrastructure. We can expect to see further incremental improvements in LLM capabilities – more nuanced outputs, better multimodal understanding – but genuine AGI remains a distant, perhaps mythical, horizon. The biggest hurdles won’t be purely technical, but rather economic and ethical. The astronomical compute costs will continue to escalate, limiting true innovation to a handful of well-funded behemoths, while the mounting societal and environmental harms will intensify calls for stricter regulation and a fundamental re-evaluation of the industry’s priorities. The conflict between profit motives and public good within entities like OpenAI will become more acute, potentially leading to further internal dissent or external challenges. While the “empire” will undoubtedly continue to expand its influence, the cracks in its ideological foundation—the growing chasm between its benevolent rhetoric and its tangible impact—will become increasingly difficult to ignore. The crucial question will be whether alternative, more ethical and efficient AI development paths, akin to AlphaFold, can gain sufficient traction to divert the industry from its current, resource-intensive, and morally precarious course.

For more context, see our deep dive on [[The Environmental Footprint of Generative AI]].

Further Reading

---

Original Source: Karen Hao on the Empire of AI, AGI evangelists, and the cost of belief (TechCrunch AI)

阅读中文版 (Read Chinese Version)