It’s Déjà Vu All Over Again: The Parallels of AI and Globalization

The slow-burning crisis over the risks of artificial intelligence (AI) reached a fever pitch this past week after Anthropic CEO Dario Amodei published an essay calling for the leading labs to pace the frontier of AI development. Not since OpenAI’s rollout of ChatGPT in late 2022 has a single piece of AI news so dominated the headlines and escaped out of the Bay Area technological bubble. Everyone from Senator Bernie Sanders and President Donald Trump to the Chinese Communist Party and the most important CEOs around the world felt compelled to weigh in on the implications of Amodei’s missive. 

This national conversation around the downside risks of AI has taken off like wildfire. From the potential effect of data centers on water and energy supplies, to the potential impact of AI deployment on jobs, to fundamental questions about what it means to have machines that seemingly can think for themselves—and better than humans—it seems as if the whole country is seized with these questions. 

The debate is not new. But even as some of the concerns about labor force displacement have yet to materialize, recent statements by researchers involved in AI’s development about the 10 percent likelihood of existential risk to humanity—along with recent examples of rogue AI agents hacking the internet—have taken the level of anxiety to a new level. Almost absent from the debate has been a discussion of the other side of the ledger: the benefits of AI development, including for medical research and personalized learning. 

At one level, this is all profoundly unprecedented: this technology is truly revolutionary. At the same time, though, this entire experience feels eerily familiar to me. In a number of important respects, the risks and fears around AI, as well as the public pushback against it, echo the debate over globalization the past few decades. 

First and foremost, both globalization and AI bring with them fears over major job loss and displacement. No matter how much the promoters of free trade promised that globalization was, in the aggregate, good for both the global economy and the American economy, those who saw local jobs move abroad remained unconvinced. 

The same is true for AI. But rather than replacing one job with another, some AI boosters promise an abundant future in which the productivity gains of the technology make work more efficient, if not superfluous, for a prosperous economy. This vision of utopia offered by AI optimists is unconvincing, however. People derive significance, meaning, and purpose from work, and the notion of a life playing video games and cashing universal basic income checks simply does not resonate with most people. Instead, it fuels anxiety and insecurity. 

Second, globalization and AI both raise important distribution issues. Globalization, for all the criticism it aroused, did lower the prices of goods and services for all Americans. Everything from food to phones benefited from a more tightly integrated global economy in which chips from Taiwan and bananas from Guatemala transited the globe, eventually ending up in your backpack or pocket for pennies on the dollar.  

And as it turns out, affordability is in fact something Americans care about. One of the challenges with globalization, though, is that while the benefits were broadly shared, they were not equally shared and were largely invisible. No one walked out of Walmart exclaiming, “Thank goodness for globalization!” Rather, globalization was seen as benefiting large firms, investors, and a narrow slice of individuals much more substantially than the broader populace. 

We are already seeing the same dynamic play out with regard to AI. But in the case of AI, trillions of dollars are being poured into the industry, enriching a handful of individuals and companies long before the country writ large receives the material benefits. The productivity boom that AI proponents promise will benefit every American will be years, if not decades, behind the minting of more billionaires in San Francisco. It is possible that the rapid development of AI will make the nation as a whole more prosperous, just as globalization did. But so long as the benefits are perceived to be going to a small number of individuals and companies, there will be a challenge building the necessary popular buy-in that would make those general benefits politically sustainable. 

Beyond pure economics, both globalization and AI have the effect of reducing individuals’ sense of agency. In the case of globalization, local economies and domestic suppliers—where the sources of goods and materials were broadly known and familiar—gave way to global, diffuse supply chains, the decisions over which were made in faraway places by unknown individuals. That proved unsettling. Part of the opposition to globalization reflects a desire to take back control. 

With AI, this is perhaps even more true. The accuracy of a large language model is based on the weights of that model, which are the billions or even trillions of internal parameters that get adjusted through a process known as training, so that the model produces the right output for a given input. That already sounds complex enough. But layer on top of that the fact that the system that crunches those data points and spits out a correct answer is basically a black box to us all. We don’t understand how and why a neural network produces the answer it does. In this instance, we are talking about handing over control of huge swaths of our lives and our economy to a system that is fundamentally impenetrable and unintelligible. 

Finally, there is something similar in the sense of inevitability around both globalization and AI. The emergence of globalization had a lot more to do with the invention of the shipping container (which made it possible to disperse supply chains all over the world) and the spread of broadband (which similarly opened the door to providing services from anywhere the internet lived) than it did with any trade policy. And the economic incentives to focus on efficiency are only now being balanced with the importance of national security, resilience, and the diversification of supply chains. 

We are hearing similar echoes with AI. The economic drive for productivity, along with the pace of improvement and potential for AI to advance itself without a human involved—a process known as recursive self-improvement—all add to this sense of inevitability. And then there is the reality that the United States is not alone. All of this is taking place in the context of a multidimensional competition with China. 

If the fears are so similar, we shouldn’t be surprised that the pushback is too. 

The most obvious similarity between the anger toward AI and globalization is its populism. In both cases, the opposition and proponents do not break down along conventional partisan lines. Just last week, Senator Sanders and former Trump White House strategist Steve Bannon—two men who happen to broadly align with one another on questions of globalization and just about nothing else—met to discuss slowing down the rate of AI progress. At the same time, elites in both parties are confident that if they could only properly explain to people the benefits of this Great New Thing, there would be no backlash at all. 

So, can AI avoid some of the pitfalls that have plagued globalization? 

First, if AI is going to avoid some of the same backlash, there should be much more serious attention paid to what needs to be done—by the government, private sector, educational institutions, and others—to help workers survive and thrive in a rapidly changing economy. The response to globalization was inadequate. Today, there is too much admiring of the problem and hand-wringing about AI, but far too little actual work being done to find and support solutions. There is one notable exception: RAISE US, a new bipartisan nonprofit focused on helping Americans transition to the AI economy, led by former Commerce Secretary and CFR Distinguished Fellow Gina Raimondo and former Indiana Governor Eric Holcomb. 

Second, there is the role of the federal government. When it comes to globalization, President Trump has moved aggressively to reshape U.S. trade policy. With the broad application of tariffs, we are engaged in a grand experiment as to whether it leads to a broader reindustrialization of the U.S. economy. When it comes to AI, the government has been schizophrenic about whether to let technology march ahead unimpeded or insert itself in the review and approval of models, to be laissez-faire or to take equity stakes and revenue shares in particular companies. 

Finally, as with global trade, AI would benefit from a set of international standards, ideally one that the United States plays a leadership role in designing. Were the United States not engaged in a competition with China—if not to achieve superintelligence first, then to control the diffusion of tech stacks around the world—we might all be happy to slow down until we have solved the safety and alignment questions. But we do not have that luxury. 

After a lot of effort, the Biden administration successfully reached agreement with China on the threshold issue of maintaining humans in the loop for nuclear weapons deployment. That might seem like a no-brainer, but it was difficult to achieve. U.S. Secretary of the Treasury Scott Bessent will be meeting with his Chinese counterpart in New York City on Sunday to discuss, among other things, cooperation on AI. This is an important step, and one can only hope that it is the beginning of broader, fruitful conversations. 

Let me know what you think about the backlash to AI and what this column should cover next by replying to [email protected].