Throughout the past year, I’ve encountered several unnerving truths — that American voters are quicker to excuse rebellions than rising prices, that turning 37 entails having one problematic knee (and another not particularly robust one), and that the New York Mets struck a dark bargain in the autumn of 1986, for which their devotees will perpetually suffer.
However, my most disturbing awareness might be this: Machines are now capable of performing large portions of my job more effectively than I am.
For a decade, I’ve earned a portion of my livelihood by converting news reports and scholarly papers into digestible explanations, which office employees could browse while feigning productivity. The talents that this occupation demands — the ability to swiftly process information, produce concise content, and project an air of expertise on diverse subjects — are abundantly present in ChatGPT.
Even worse, it can leverage these capabilities with far greater speed than I can. ChatGPT needs just five seconds and a bit of energy to detail the impact of President Donald Trump’s tariffs on the dollar. To accomplish the same, I require approximately 48 hours, six meals, 37 social media breaks, and at least three doses of medication. And unlike myself, the chatbot can promptly rephrase its breakdown in the manner of a pirate.
But journalists are hardly the only ones dreading our impending professional redundancy (which, admittedly, has long been a favored topic in the profession). Intellectual workers of every kind — from software developers to finance experts to filmmakers — are growing concerned about competition from robots. And with valid cause.
During investor briefings, executives are frequently highlighting strategies to meet their staffing demands by leveraging artificial intelligence at unprecedented levels. Goldman Sachs is curtailing hiring and expediting AI implementation. The financial tech company Klarna has employed AI to reduce its workforce by 40 percent. Salesforce reports that machines are now able to handle 50 percent of the company’s workload.
Widespread joblessness caused by AI remains a science fiction trope. Nevertheless, the technology seems to be decelerating recruitment in greatly affected industries, especially for entry-level positions. And employment statistics suggest something peculiar is transpiring: For the past 35 years, recent university graduates have almost invariably experienced lower unemployment rates than the general American workforce. Now, a recent college graduate is more likely to be involuntarily jobless than the average US worker.
Simultaneously, the prominent AI research labs are promising the emergence of an artificial general intelligence (AGI) — that is, a machine that can outperform humans in all mental tasks — in the foreseeable future. Anthropic CEO Dario Amodei suggests AI will probably eliminate half of all beginner white-collar roles by 2030. Investors appear to be placing substantial bets on these predictions. This year alone, companies are poised to invest $375 billion in AI infrastructure.
Among apprehensive knowledge workers, these developments have stirred up semi-ironic discussions about the advent of a “permanent underclass”: Once AI renders nearly all human labor commercially worthless, the majority will be doomed to unending servitude and instability. No employer will compensate you for tasks that a machine can accomplish more effectively. And no market system will permit you to ascend the economic ladder if your work possesses no value. In Silicon Valley, this logic has given rise to a uniquely bleak version of hustle culture: Amass your wealth within the next five years, and you’ll secure a position in the eternal aristocracy of AI proprietors — fail, and you’ll forever be at their mercy.
These assertions are largely speculative. It is not certain that modern AI labs have viable economic models, let alone the capability to develop fully competent machines. Yet, of all the doomsday visions offered by pessimistic futurists, AGI-induced neofeudalism appears to me as one of the more realistic possibilities.
AGI may or may not choose to eliminate the human species. But it will diminish the worth of human labor, almost inherently. We do not know how ordinary individuals will fare in a world where the affluent can manage without their talents and endeavors. But it is sensible to worry that the response is “not especially well.” After all, there are already societies in which employees have limited economic power over elites. And they are generally unpleasant environments for the typical person.
It’s therefore worthwhile to delve into exactly how AI might create an unchangeable oligarchy — and what actions can be taken to stop it from occurring.
Related
- AI is coming for the laptop class
You will own nothing and be starving
Automation is the source of all affluence.
Absent labor-saving devices, most of us would still dedicate our lives to extracting nutrition from the soil, forever susceptible to starvation due to adverse weather conditions. Robots are why most Americans can retire for the night with a satisfied appetite in a temperature-regulated home, watching an endless stream of audio-visual content on demand — all without the need to face sunburns or serious injuries.
However, automation has also frequently upended people’s lives. Granted, robots usually benefit the majority of workers eventually. But in the intervening period, some jobless individuals die impoverished.
Nonetheless, in every previous wave of automation, it was at least hypothetically achievable for those who lost their jobs to secure higher-paying positions by improving their skills. Even as machines rendered human labor uncompetitive in certain areas, they encouraged the discovery of new commercially valuable pursuits. When the steel plow removed field laborers from farms, they could discover new work in factories. When industrial robots pushed manufacturing employees off the assembly line, they could — theoretically, as the saying suggests — “learn programming.” Far from depleting a limited supply of jobs, technological advancements generated additional and improved roles.
But there exists no guarantee that this will always be true. No scientific principle prevents silicon from enabling logical thought as effectively as brain tissue. Nor does any aspect of the universe ensure that human hands will consistently be superior at folding towels (or any similar detailed task) than mechanical ones. And the quantity of tasks that machines cannot accomplish has, actually, been rapidly decreasing.
Six years prior, the globe’s most advanced AI models struggled to compose a fragment of Python code or a clear paragraph. Today, they are capable of earning top marks in constitutional law and exceeding 99.5 percent of individuals on specific coding assignments.
Advancement in robotics has been less rapid. Machines find it simpler to excel in calculus than to assemble Ikea furniture. But the more proficient robots become at handling data, the more comprehensively they will be capable of outperforming people in the physical realm — a fact emphasized by current progress in driverless cars.
“The majority of us are presently concentrating on machines having the ability to execute all cognitive tasks,” University of Virginia economist Anton Korinek conveyed to me. Yet, in his judgment, one of the key lessons of the last couple of years is that “if we equip our existing robots with enhanced brains, they can function considerably better.” Consequently, Korinek believes that machines may “soon” be equipped to surpass humans in most “cognitive and physical labor.”
None of this signifies that AGI is just two years away, as certain industry advocates (and pessimists) have been predicting. But it does present some justification for believing that economically fully competent machines are plausible.
And there are reasons to infer that its arrival might be inevitable over a sufficiently lengthy period.
Humans, for instance, did not evolve to optimally generate worth in a contemporary market economy. We cannot perform tasks for extended durations without experiencing hunger, fatigue, ennui, or lust. We struggle to accept mandates unless shown some degree of esteem. Fundamentally, excluding cyborg augmentation, the efficiency of our minds is restricted by our biology, similar to the strength of our limbs.
Machines’ capabilities are less inherently constrained. Indeed, given considerable progress in AI, hardware, and energy technologies, it is possible that robots could conduct a week’s worth of any person’s work while using fewer resources than it would require to keep that individual alive.
This has prompted some in the AI sector to forecast that their technology will ultimately drive human wages below subsistence levels. In that scenario, after briefly rescuing humans from lives devoted to cultivating land and pursuing animals for nourishment, technological advancements would have reverted most humans to that prehistoric state.
Or, at a minimum, it would achieve this — unless elites opted to share the gains of AGI with the masses. But it is uncertain whether we should anticipate such redistribution, in a reality where an ordinary individual’s labor is not worth the expenditure of their sustenance.
Related
- Why can’t anyone agree on how dangerous AI will be?
How AI could break the engine of social progress
America today is a strikingly unequal nation, in which the powerful exploit and disregard the vulnerable in multiple ways.
It is also among the most fair-minded civilizations to have ever existed.
Since the commencement of agriculture, the majority of people have resided in severely stratified communities, where privileged groups accumulated both resources and statutory rights. Typically, at the summit of these societal structures stood aristocrats, who received income without exertion; at the base, enslaved persons, who labored without recompense. In certain cases, the former held nearly absolute power over the latter, even to the point of killing them.
Ordinary individuals, meanwhile, had limited legal authority, wielded no political clout, and tended to pay higher effective tax rates than the affluent, while receiving a meager segment of economic production and restricted public services beyond military protection.
Compared to ancient Rome or China, modern industrial democracies bear resemblance to socialist heavens. Even in the notoriously uneven United States, all inhabitants enjoy explicit equality before the law and a basic level of existence within the economy: The working class is entitled to dietary support and government-sponsored healthcare coverage, the elderly or infirm to guaranteed income, and the young to 13 years of no-cost education. Meanwhile, the average middle-class household pays a diminished tax rate than the average wealthy one. Numerous employers feel economically forced to supply their employees with paid family leave, health benefits, and other incentives — and all businesses are statutorily obliged to pay a minimum wage, maintain secure operating conditions, and permit unionization
We can build a highly automated economy in which everyone can flourish. … We just need to develop healthy political and economic institutions.
What clarifies the comparative compassion of numerous modern societies? Some would point to the unbiased concepts spread by universal religions, Enlightenment ideology, and social democratic philosophies. And such doctrines have certainly been relevant.
Yet the dissemination of unbiased concepts in recent centuries doesn’t merely reflect their inherent moral attraction but also their capacity to facilitate mutually beneficial cooperation. The termination of slavery did not merely render America more equitable; it also made the nation more economically successful. Labor markets comprised of independent employees are more productive than those filled by serfs or enslaved individuals. As a result, the emancipation of those tied to the land triggered a surge in both compensation for ordinary individuals and revenues for elites.
Similarly, in an environment of vigorous economic development — where regular workers’ hours are becoming increasingly valuable — the affluent have a vested interest in upholding political stability and labor peace, even at the expense of sacrificing a portion of power. Ceding to mass enfranchisement may have weakened elites’ control over administrations. But doing so also assisted in legitimizing organizations that granted capital holders with elevated and enhancing living standards. Enduring unions may have compelled business leaders to concede a higher wage expenditure, but doing so also protected them from the possibility of strikes. Funding public schooling may have obliged the privileged to contribute more taxes, but it also cultivated a more productive workforce.
This isn’t to suggest that material motivations automatically dictate history or human conduct. People crave prestige, purpose, and self-respect as intensely as they do prosperity. The quest for these other benefits can inspire economically secure individuals to champion progressive reforms, even if they have no clear material or political motivations to do so. Indeed, as economic progress satisfies numerous affluent individuals’ monetary aspirations, we would anticipate them to begin placing greater importance on post-material concerns — which, in numerous instances, will translate to advocating for greater social and economic parity.
Still, elites’ powerful impetus to preserve the economic collaboration and political allegiance of average workers undeniably simplified democratization. It is not coincidental that the recent two centuries of unbiased advancement has corresponded with a prominent rise in the market value of human labor — and consequently, the bargaining authority of human employees.
AGI threatens to reverse all these trends.
The future is…Congo?
For a glimpse of fully automated neofeudalism, consider the Democratic Republic of Congo.
Or at least that’s what the AI business owners Luke Drago and Rudolf Laine suggested earlier in the year in an essay titled “Defining the Intelligence Curse.”
As Drago and Laine note, Congo possesses abundant natural resources. The country contains significant petroleum deposits and approximately $24 trillion in mineral riches. Yet Congo ranks among the five most impoverished countries on the globe, with over 70 percent of its populace surviving on under $2.15 daily.
This conflicting blend of commodity riches and widespread destitution is not extraordinary. On the contrary, substantial resource endowments appear to correlate with slower economic progression and elevated corruption — an occurrence that economists and political scientists have named “the resource curse.”
Scholars have attributed this contradiction to numerous diverse elements. But Drago and Laine emphasize one in particular: Substantial commodity supplies can lessen elites’ incentives to elevate the productivity of average workers.
Vast energy or mineral deposits offer investors with a ready source of earnings — and countries, with a straightforward source of income. The former do not need to jeopardize capital on intricate production procedures to obtain returns, while the latter need not trouble themselves with the difficulties of forming effective tax collection agencies, the fiscal costs of cultivating a skilled workforce, or the political perils of nurturing an educated population and diversified economy with competing power hubs. Oligarchs and public representatives can merely enrich themselves with commodity incomes instead.
According to Drago and Laine, AGI could cultivate an even more severe manifestation of this phenomenon. Aided by super-intelligent robots, the theory suggests, capitalists won’t have to appease troublesome workers to generate revenue. And states won’t rely on average individuals for tax income. On the contrary, as machines consign the bulk of workers to permanent unemployment, governments will possess few funding sources beyond the surplus gains of corporations. The regular individual’s economic influence over public and private authorities will be extinguished. And states and businesses will have sparse material reason to invest in their schooling or well-being.
In autocratic petrostates, everyday people can still apply some pressure on elites through the implied risk of rebellion. As per some political experts, anxieties of provoking a popular insurrection limit the misuses of kleptocracies, urging some to finance considerable social benefits.
But Drago and Laine contend that, post-AGI, mass societies are likely to forfeit even this enduring constraint on tyranny. In a reality of super-intelligent drones and AI-operated surveillance, they worry, administrations will require the endorsement of inconsequentially few individuals to uphold an effective monopoly on violence.
Perhaps, the simple strength of enlightened beliefs or altruistic empathy will encourage the powerful to distribute wealth in such a society, anyway. But the conduct of numerous American tech millionaires and public figures today — when they still depend on everyday workers for their livelihood — does not spark substantial confidence.
We need to AGI-proof American democracy
If we can’t trust in the magnanimity of oligarchs, what can safeguard average individuals’ authority amidst AI advancement?
To economists such as Erik Brynjolfsson and Daron Acemoglu, part of the solution involves promoting the development of artificial intelligence that enhances labor rather than superseding it. AI instruments could plausibly increase wages by making employees more productive. Conceptually, these innovations could safeguard laborers’ economic leverage and promote a more just framework for advancement. Regrettably, in this interpretation, the tech sector’s infatuation with AGI has predisposed it to concentrate chiefly on labor-automating machines, rather than the labor-augmenting variety.
Brynjolfsson and Acemoglu maintain that the government aggravates this inclination by taxing capital at a lower rate than labor, thus incentivizing automation. But they also acknowledge a role for entrepreneurs to cultivate the labor-augmenting technologies that major laboratories are overlooking. This is what Drago and Laine have aimed to accomplish with their own startup.
If AGI ever becomes technically achievable, however, the financial benefits of embracing it will be overwhelming; Well-intentioned business owners are unlikely to turn it back. In that situation, our prime means of staving off techno-feudalism will be institutional transformation.
The more equitably we apportion capital ownership before AGI depreciates human labor, the less oligarchic a fully automated economy is prone to be. And the more democratically accountable we render our governance, the better average individuals’ prospects of constraining elites through the exercise of political authority, even as their economic leverage diminishes.
The “resource curse” is a frequent predicament. But it is not an unavoidable one. Norway is an unusually oil-rich country, with fossil fuels producing a quarter of governmental earnings. Yet it also ranks among the most prosperous and equal nations on the planet.
The Norwegians’ resilience to the resource curse has a straightforward clarification. By the point Norway discovered its substantial petroleum supplies in 1969, it had already established one of the globe’s most equitable social frameworks. When processed via the nation’s social democratic systems, the emergence of a vast new source of inanimate wealth did not generate windfall rents for plutocrats but instead a sovereign wealth fund that sponsors generous pensions for ordinary Norwegians.
If we can enhance our political economy, sophisticated AI could yield similarly expansive benefits to future Americans. “The very essence of technological progression is that the economy can accomplish more with less,” Korinek remarked during our discussion. “It signifies you can possess greater output, greater prosperity, greater goods and services that individuals can utilize. The central question is how those goods and services will be allocated. But the favorable aspect is that, theoretically, since we have greater abundance, everyone could potentially be made better off.”
We can cultivate a highly automated economy in which everyone can prosper — even the obsolete writers of sprawling, self-referential “explainer” journalism. We simply need to nurture effective political and economic structures.
Admittedly, at present, achieving that appears roughly as challenging as constructing a God from silicon.
Source: vox.com
