2025: Surviving (and Thriving) in the Age of Intelligent Machines

The world asks if AI will replace all human work. This piece explores why labor might be safer than we think.

“The discovery and use of machinery may be injurious to the labouring class…”

— David Ricardo, Principles of Political Economy (1817)

I almost didn’t write this post. It seemed too deep, too heady. Too much in the weeds. But I decided to publish it anyway. The motivation for deciding to publish is personal: I have personally lived through the advent of roughly 1B new humans into the global labor market.

In the language we use for AI today, I lived through the entry in the labor market of 1B “generally intelligent agents.” We don’t usually name it like that (nor should we), but that’s what happened during the 1990’s in Southeast Asia as China scaled up and came online as a global economic power. One billion new workers entered the global economy all at once. And they were all directly competing with workers in Southeast Asia.

I visited China then. I lived in China’s shadow in Southeast Asia. I saw firsthand how that competitive pressure shifted labor markets in Indonesia and the Philippines. And it was not all bad. In fact, it was mostly good—more people in Southeast Asia over that decade ended up with better food, better transportation, TV’s, could put their kids farther through school—life got better on most of the measures of personal well-being that mattered to them.

And I keep thinking about that when I read AI doom and gloom articles. I think we are missing a wider perspective. And that’s why I published this piece. I won’t pretend it’s short, but I think this is perhaps the most important subject to write about these days, and I think it’s worth the effort to read.

---

For centuries, we’ve feared that machines would displace human workers for good. Each new wave of automation—from spinning jennies to steam engines, from assembly-line robots to neural networks—has triggered a fresh round of existential angst about the “end” of human labor.

Today, large-scale language models and increasingly capable AI agents have revived that anxiety like never before. In fact, the final trigger for this post (which I’ve been thinking about for months) came from Maxwell Tabarrok, who argued for comparative advantage as a key factor in AI automation discussions on January 9th.

But beneath the buzz about AI consciousness, alignment, and GPT superpowers lies a deeper economic question: Will AI be subject to the same forces—market, physical, thermodynamic—that have constrained every previous technology? Or can it break free, “escape velocity” style, through recursive self-improvement?

Some see near-infinite replicability leading to a world of superintelligent AI that drives humans to economic irrelevance. Yet, over 200 years of data suggest labor’s share of GDP has remained surprisingly constant even amidst sweeping automation revolutions. The key question we’re all facing in the jobs discussion is this: are we finally facing the exception to the rule, or will tomorrow’s AI just follow the path of past breakthroughs?

To boil it down even more simply in energy terms: does AI escape the laws of entropy eventually or not? If we think AI will not be bounded by physical limits, then it is pretty rational to suppose AI will figure out how to solve coordination problems and comparative advantage problems and out-compete the human race at every turn.

But if the fundamental laws of physics remain constants, then there will be limits on the energy efficiency of AI, and in that efficiency limit lies the opportunity for comparative advantage.

Part I: The Grand History of Automation—and a Persistent Labor Share

Kaldor’s Stylized Fact: Labor’s Share Around 50–60%

One of the most confounding observations in economic history is the near-constancy of labor’s share of GDP: around 50–60% for more than two centuries. Widespread mechanization, electrification, and digital computation have vastly raised worker productivity—and yet the fraction of income that goes to workers, rather than to capital and materials, barely budges over the long run.

Why? Economists point to four powerful countervailing forces that offset displacement by automation: