No Permanent Underclass in the AI Era

Lately, a popular online quip has circulated, suggesting that individuals who fail to grasp artificial intelligence concepts will end up forming a so-called "permanent underclass." This term describes the emerging group of society's have-nots—those destined to be sidelined amid the impending surge o

Lately, a popular online quip has circulated, suggesting that individuals who fail to grasp artificial intelligence concepts will end up forming a so-called "permanent underclass." This term describes the emerging group of society's have-nots—those destined to be sidelined amid the impending surge of AI technologies. In this scenario, a select portion of the population would enjoy the benefits of large language model-driven agents handling their workloads entirely, while the rest languish in unemployment and unending financial hardship.

Though this idea sounds plausible on the surface, it lacks any solid foundation in historical patterns. Consider, for instance, the dramatic transformation in the United States agricultural sector: the proportion of the workforce engaged in farming plummeted from about 90 percent in 1790 to under 2 percent in the present day. Imagine approaching someone in 1790 and informing them that nearly 98 percent of all farming positions would vanish over time; they would struggle immensely to foresee the diverse occupations that former farmers and their descendants now hold.

No one back then could have anticipated professions such as social media coordinators, property brokers, data analysts, or the myriad other specialized jobs that simply did not exist in that era. This historical blind spot underscores why the current wave of alarmism surrounding AI is fundamentally flawed. The ongoing technological evolution promises to generate a multitude of novel job categories and amplify requirements in established fields. This dynamic accounts for the recent uptick of more than 10 percent in software development job advertisements over the past year, even as AI tools become more integral to coding processes. As Kenton Varda, a key technical figure at Cloudflare, insightfully observed:

Worries that software developer jobs are going away are backwards. There is SO MUCH software to build right now, that previously wasn’t possible (uses AI directly) or wasn’t cost-effective (too niche). We’re going to have more developers, and orders of magnitude more software.

This pattern aligns with what economists call Jevons’ paradox, a principle stating that enhancements in the efficiency of a resource—such as the process of developing software—often result in greater, rather than lesser, overall usage of that resource. We can expect this effect to ripple through numerous professions and sectors thanks to AI advancements. Ultimately, this will elevate human prosperity in the years ahead.

Critics might counter that "this time it's different" since AI targets knowledge-based tasks rather than mere physical labor. If that's the case, they ask, what meaningful work will remain for humans?

This perspective carries weight, yet analogous concerns echoed during the mechanization of manual jobs. In early 19th-century England, a movement known as the Luddites emerged, with workers smashing weaving machinery out of fear for their livelihoods in the textile industry.

Just as the Luddites could not envision the opportunities that would arise to supplant their trades, we face a similar limitation today. Undoubtedly, new vocations will surface, demanding unique human capabilities that lie beyond our current imagination. These roles will not rival large language models head-on but will instead complement and amplify their strengths.

Moreover, although AI's transformative pace may outstrip previous innovations, the adaptation period will accelerate accordingly. With information disseminating at unprecedented speeds in our modern world—far surpassing anything achievable in bygone eras—workers can pivot, upskill, and realign their careers more swiftly than ever before.

David Oks penned an excellent analysis exploring why AI's influence on employment markets will prove less devastating than many anticipate:

…the relevant question for labor impacts is not whether AI can do the tasks that humans can do, but rather whether the aggregate output of humans working with AI is inferior to what AI can produce alone.

Thankfully, we remain in an era where the synergy of AI and human effort surpasses AI operating in isolation.

If doubts persist, reflect on how past technological leaps have reshaped societies for the better. Throughout documented history, no instance exists where a widely embraced innovation coincided with a rising tide of poverty rates over the long term.

Quite the contrary: in the past 50 years, the global count of individuals enduring extreme poverty has dropped by approximately 66 percent since the 1970s, even as revolutionary technologies proliferated across the globe.

Admittedly, certain technologies have triggered brief, localized disruptions or downturns. The nascent stages of the Industrial Revolution, for example, brought harsher conditions for many laborers—exemplified by the Luddite uprisings. Nevertheless, these challenges proved transient and never crystallized into a lasting underclass.

Some readers might dismiss this evidence, contending that extreme poverty fails as a benchmark. Instead, they emphasize relative wealth disparities over absolute gains. Even if absolute destitution vanishes entirely, could not a comparative underclass still persist?

There is merit to this view to an extent. Such divides already characterize our world, with a tiny fraction enjoying private jets, luxury yachts, and freedom from traditional employment. Fortunately, these elevated statuses rarely endure indefinitely.

The Temporary Elite

A primary reason for my optimism regarding any potential AI-driven societal stratification—even if it materializes—lies in its probable brevity. Examining history reveals that family fortunes invariably wax and wane across generations. Wealth and influence amassed in one era often dissipate in the subsequent ones. Those assuming that affluent dynasties maintain their status eternally should note this striking anecdote from Fortune’s Children:

When 120 of Cornelius Vanderbilt’s descendants gathered at Vanderbilt University in 1973 for the first family reunion, there was not a millionaire among them.

Cornelius Vanderbilt rose from humble origins to claim the title of the world's wealthiest individual. Yet, his colossal empire, rooted in the railroads of his day, faded within a handful of generations as newer industries eclipsed it.

This vignette highlights how even today's prospective AI beneficiaries—those poised at the pinnacle—will likely forfeit their advantages eventually. It evokes Warren Buffett's wry observation:

I try to invest in businesses that are so wonderful that an idiot can run them. Because sooner or later, one will.

Indeed, regardless of your current prosperity or achievements, some future heir will embody that "idiot," squandering the legacy through mismanagement or misfortune. Traits like exceptional ability, sharp intellect, and balanced disposition tend toward regression to the mean across lineages, rendering sustained dynastic wealth exceptionally rare.

This historical truth fuels my doubt about any enduring AI-spawned class chasm. Should one arise, its lifespan would be fleeting.

There will be no permanent underclass because there is no permanent elite.

All pinnacles of privilege and authority are inherently ephemeral. Time and again, history affirms this reality.

Therefore, cease fretting over being overtaken by progress and concentrate instead on amplifying your value and utility in an evolving landscape.