Paid subscribers can access the podcast version:
How the American Empire Plans to Win the AI Race (And What Hurdle They Must Overcome)
Thanks for checking out this episode, and I’d value your feedback.
Two things may be true at once: There is legitimacy to mounting domestic backlash against AI, and the AI economy need not, and must not, burst in America. Opinions aside, we’ll grant this conclusion to better understand the strategy behind it.
The AI economy requires deeper AI integration, and integration cannot occur without addressing the issue of worker displacement upfront.
Just as Anthropic is modeling deep AI integration to nudge non-frontier companies to follow suit, America must do the same to model before potential global consumers—both the national and global levels point to solving worker displacement.
The Legitimate Concerns
In my writings, I’ve gestured at several of the legitimate concerns:
While Anthropic has shown that deeper AI integration yields productivity and new opportunities, they have also indicated that financial incentives and job security are needed to spark the will to integrate; yet, many non-frontier companies are not positioned to provide these incentives.
Leaders of frontier AI too often signal the coming disruption of AI, yet they treat disruption as if it is a positive, metamorphic force. The conditions of Silicon Valley—high concentration of technological opportunities, startups, and investors—buffers the shock of the fast iteration of ventures. Most Americans don’t live with these conditions, and their aversion to this kind of risk is natural (and not a defect). The “spirit of disruption” framing is incredibly ineffective for consolidating support for the AI economy.
Similarly, the gradual release of frontier AI models through a chat interface has allowed the public to incrementally adopt and develop positive associations with AI. Google reported in late July that 90% of the workforce has integrated AI in some way. Clearly, value was perceived in these tools. However, it also gave people the false perception that AI would augment their professional development without side-effects. At some point, workers realized that AI would, at best, relocate their work, and, perhaps more likely, displace them—either directly as their work is automated, or indirectly as a new AI organization absorbs the integration, displacing those in pre-AI segments of the business.
A massive gap has emerged between what Anthropic’s workers (and similar) perceive as the potential for AI to replace skilled work and what non-frontier workers perceive. Even business leaders in corporate America are skeptical of AI’s potential impact, citing that knowledge workers possess irreplaceable business intelligence. If they are right, then there is no coming payoff in the AI economy, for we would have reached that by now. If they are wrong, and I believe they are, then corporate America will feel on the precipice of displacement. In either event, it adds to the domestic backlash.
Outside of my writings, the following has also entered the public imagination:
Besides concerns about the workforce, there is the general bucket of AI-induced harm that has led to legal action against frontier AI. For instance, Texas Attorney General Ken Paxton recently sued OpenAI.
Data centers have become unpopular, as Aaron Renn has documented so well.
The Geopolitical Concern
All of these concerns are no small thing, and the vague notions of universal basic income and other promises emanating from frontier leaders dilutes into a kind of “hype slop.” Clearly, a better vision could be positively presented to stir within American workers a concrete commitment to support the AI economy. Such a positive vision, however, requires moving from private interests to the common good.
The AI economy is not merely about hopes of increased productivity, greater material prosperity, and humane breakthroughs; it is at the center of an intense geopolitical arms race—a race to become a dominant empire of a new kind. In a fantastic paper, Fergusson, Klaassen, and Schmidt write, “By controlling the large language models, chips, physical infrastructure, and capital that provide the foundations for advanced machine cognition, Beijing and Washington are not only attaining new heights of economic performance and power projection; they are also setting the terms of all future scientific and intellectual progress. The term “AI superpowers” fails to capture the influence and reach that these rivals now wield. Even without active alliance building or coercion, they can make all other countries that seek access to advanced AI beholden to them.” There is more at stake than improved worker experience and economic payoff; both the Biden and Trump administrations have pursued technical supremacy in an interest to defend the free world from totalitarian regimes, namely China. This race, and its framing, is analogous to that of the Cold War. The meta payoff is economic domination that counters trends of rising debt and economic flattening.
As Fergusson, Klaassen, and Schmidt note, America has the upperhand in AI software whereas China has the edge in robotics and other embodied AI technologies; yet, the two empires are close in capabilities. Moreover, America excels in its ability to amass private funding and funnel premier talent. Because of this, the United States can fund the AI build-out without touching its own savings. China, by contrast, marshalls its AI innovation directly via the state, streamlining uniformity while demanding state-directed financing. The calls to pace the American AI frontier—and all the resulting proposal for regulation mechanisms—must be understood in this context. The Trump administration’s approach of resisting state regulation while streamlining the development of data centers calculates that innovation and funding from the private sector will allow the United States to gain a technological edge over China without the bottleneck of government funding and quality control. The administration does maintain strict oversight, however, when it comes to open access to AI for national defense and export controls.
The race involves more than having an edge in capability; the meta payoff concerns adoption of the developed technology. Here, more trade-offs exist than merely capabilities. Fergusson, Klaassen, and Schmidt highlight that China incentives adoption through cheap and mostly open technologies whereas American technology is expensive and mostly closed. This makes sense given what has been said above. The American approach incentivizes innovation through private markets where financial reward is sought after—and rewards cannot happen if the technology is mostly open. While the higher costs of American technology raise the bar for adoption, the edge can be maintained so long as the technology is superior and worth the premium. By contrast, China risks overcapacity and the evaporation of any margin given that competitors can leverage their open technology, which has historical precedent.
The Internal Tension
America’s risk brings us back to our original set of concerns. While America’s expensive-and-closed offering can be fruitful at a global scale, the technology being offered is largely concentrated in the frontier AI companies. Once computers could be mass-produced swiftly and cheaply, the market for software products grew and invited many players. Companies that could marshall the development and sales of quality software to solve domain-specific needs could reap rewards, similar to non-technical goods and services. Hence, even the digital revolution, momentous as it was, did not drastically change the formula of amassing workers skilled in business intelligence to design quality products and accelerate their development and sales. This also meant that companies retained a recognizable organizational shape—and all the functions and methodologies that come with that shape—regardless if software was their primary product. Moreover, software became an important tool to make the functions within that organizational shape more efficient. Opportunities abounded.
The dawn of the AI economy does not present as smooth a tradeoff for corporate America as the transition from manufacturing to software:
The underlying models cannot be mass-produced easily like computers.
The models are capable of absorbing the execution of business intelligence that informed the shape of organizations.
The value of models is in their sheer power which amassed by the concentration of resources, requiring the harnessing of energy, critical minerals, advanced chips, and capital on a grand scale. The market is inherently narrow.
While many assumed that AI models would expose a mere tool that could accelerate the software market that had been well-established, the tool makes software very easy to produce, exposing to the software marker to the same margin-deterioration risk that China’s cheap-and-open approach faces.
As a result of this, the American advantage in the digital revolution to marshall corporate organizations with quality talent to develop and sell software is threatened. Manufacturing, model development, and now the harnessing of resources enabling AI models are the dominant markets that remain; yet, the business intelligence traditionally needed to marshall these markets can now be absorbed by AI.
The tension, therefore, is that the federal administration is focused on the meta competition with China, resulting in limited government regulation and swift development of AI models in the private sector. Combined with strict export controls, this gives America the attractive expensive, closed package—so long as the technology is worth the premium. However, the creates an internal tension: enabling frontier AI threatens the software market and any slice of the corporate shape that has been driving by human business intelligence. Whether implicitly or explicitly, corporate America sees the writing on the wall and is resisting AI integration.
Undoubtedly, other nations will face this same tension. If AI absorbs domestic workers rather than offering a tool to augment them, then consumption collapses alongside those workers. In effect, this raises the question whether AI is a race for a global market or purely an arms race. The American-China race assumes that it is the former, as their distinctive policies reflect.
The Path Forward
Putting it all together, providing a solution to the issue of worker displacement would give America an edge in the quest for global consumption of AI technology since that issue is the bottleneck. It simultaneously would resolve the brewing backlash internally. Merely appealing to the China-America race will elicit counterproductive skepticism; merely appealing to seemingly utopian promises of universal income and the like will do the same. Americans need a tangible path forward to sustain their livelihoods, even if it requires a period of transition.
Here are some practical recommendations to that end, leaving deliberations around federal-private trade-offs aside:
There must be some mechanism to protect American workers from being absorbed by AI. China has issued that its technology companies resist laying off workers. Either a similar policy needs to be enforced, or a mechanism to swiftly transition workers must be marshalled.
Since the AI economy necessarily concentrates wealth, the redistribution of wealth from frontier companies into other segments of the economy could provide additional financial security for workers and encourage deeper AI integration. While this has been seen as leftist in a digital economy, the essential differences in an AI economy make this a sensible rather than anarchic policy.
While counterintuitive, Americans would do well to have less exposure to AI. This allows labor to concentrate into applications of AI that are undoubtedly for the common good, creating an edge through depth over breadth. Moreover, it ensures that the virtues necessary to travel this path don’t atrophy.
Both the global and national interplay, and its importance, needs to embedded in an accessible, positive vision.
Similarly, legitimate criticisms around AI should be channeled into this productive vision rather than devolving into anarchic protests to halt AI altogether.
My hope is that this vindicates the concerns I have traced while tying them together into something that serves the common good. This is a difficult challenge, and the leadership in frontier AI deserves criticism; yet, we must be careful, again, to not let the anarchic reign. AI integration, the preservation of human workers and human intelligence, and the eradication AI gluttony can all complement each other. It is an experiment, no doubt, but what has made America great is her leading by example in global experiments. May we see this happen once again.
Next compare the American plan with that of the UK:
Role Reversal: The UK Adopts Early American Protectionism in Its Ambitious Plan for AI
Comparing the AI Plans of America and the UK