Tag: economics

Posted on

Beyond Capitalism, Communism and Democracy: The Convergence of AGI-Governed Societies

[Written with the help of GPT-4]

I’ve done more thinking about the consequences artificial general intelligence (AGI) domination of the world will have for human-created institutions. In a recent blog post of mine, “The end of Homo sapiens history and the first posthuman”, I discussed how intelligent machines would lack emotional attachments to things we consider sacred, like languages, religions, and national borders, and would therefore be open to abandoning them or replacing them with something better. In this essay, I’d like to focus on how that will shape future political and economic systems in the AGI Era (aka “the Posthuman Era”).

The 20th century was defined by ideological competition, mostly along national faultlines. By midcentury, fascism and imperialism had been discredited, and by the end, so had communism. However, the widely held belief that “democracy” and “capitalism” proved themselves the best systems for organizing governments and economies, respectively, is a gross oversimplification. Among the “democratic” nations, there’s considerable variation in individual rights and the role of the state, and among “capitalist” nations, variations in economic freedom are just as great. The successes of the Asian Tigers and China pose the biggest challenge to the simplistic assumption that “democracy and capitalism are the best.”

Capitalism is best thought of as an optimization algorithm that leads to one, powerful firm dominating each niche of the economy. The model’s flaw is that the firms only have incentives to pursue their narrow, short-term self-interests, which, over time, will destroy the conditions that allowed the market to exist in the first place: Pure capitalism will give rise to things like monopolies that rip off their consumers and stop innovating their goods and services, and factories that emit so much pollution they gradually kill off the customers who buy their products.

The health and growth of the economy more broadly speaking depends on having a referee with a different set of incentives (ex – profit agnostic) from the private firms–the government. The ideological winner of the 20th century was actually the “mixed economy,” which is a system where capitalist markets exist within legal boundaries set by governments. The rules were in turn largely set by each country’s citizens, establishing a balance between economic competition and security that suited their culture. Even among democratic, capitalist nations today, vast diversity exists in governance, civil liberties, and economic organization. Scandinavia’s social democracies, the United States’ market-heavy liberalism, and Japan’s corporatist structures all exist under the broad heading of “capitalism.” Each is recognizably democratic, but the institutions, welfare provisions, and power balances differ significantly. This shows that human labels already obscure substantial variation. Humans will struggle even more to apply their familiar labels to the political and economic systems AGIs create in the future.

The results will not fit comfortably within familiar human-created categories like “capitalism,” “socialism,” or “democracy.” Humans have long relied on ideological frameworks to define themselves, but these frameworks are ultimately rooted in history, sentiment, and cultural identity. An AGI, by contrast, will approach governance as an optimization problem, unconstrained by emotional loyalty to existing systems. The outcome is likely to be the emergence of hybrid structures that mix and transcend traditional models, calibrated to human preferences in ways too complex for us to map back onto old labels.

One of the reasons AGI will transcend traditional systems is its superior ability to understand humans and their desires. Already, algorithms employed by large technology companies can build personality profiles from user interactions, predicting behavior and manipulating preferences with startling accuracy. These tools, while primitive compared to AGI, already surpass human intuition. An AGI would carry this to a new level, not only modeling individual preferences with unparalleled fidelity but also distinguishing between stated and revealed preferences. It would know what people actually want — often better than people know themselves. It will also be able to induce human demands for specific goods and services, and to preemptively ramp up their production, helping to create a new economic system.

This capacity allows AGI to optimize social and economic outcomes in ways that humans cannot. Instead of designing systems around political compromises or ideological commitments, AGI could dynamically adjust production, distribution, and governance to meet authentic needs. The economy that emerges from this process will not resemble capitalism or socialism as we understand them. It will be something new: an adaptive, preference-driven engine of allocation and governance.

Humans cling to institutions not only for their practical functions but also for their symbolic and emotional value. Constitutions, flags, currencies, work schedules, and elections are not just mechanisms but rituals that confer identity and continuity. An AGI will have no such attachments. It will evaluate institutions only by how well they serve defined goals. If an institution is inefficient, it will be discarded or redesigned without hesitation. In the hyper-competitive arena of international competition between AGI-controlled nations, the demonstrably failing political and economic systems present today in places like Cuba won’t exist.

National boundaries, currencies, property rights, and even work itself could all be reshaped or abolished under AGI-led optimization. For example, money might be replaced with dynamic credits tied to welfare indices rather than market exchange. Elections, rather than being periodic spectacles, might be replaced by continuous preference elicitation and adjustment. AGI will not preserve these systems for sentimental reasons.

This detachment is both strength and weakness. On the one hand, AGI can innovate institutions at machine speed, abandoning inefficient traditions without the inertia of human politics. On the other, humans derive meaning from continuity. Abrupt changes may generate alienation, resentment, and rebellion, even if outcomes are objectively improved. Americans, for example, might resist an AGI-designed system that resembles socialism, not because it fails them materially, but because their cultural upbringing equates socialism with un-American values. Legitimacy in human governance rests not only on material well-being but also on symbolic fidelity.

An AGI that truly understands human psychology will likely manage this by creating “soft landings.” It may preserve symbolic forms — elections, currencies, national holidays — even while radically altering their underlying mechanics. Just as modern fiat money no longer represents gold but still carries the familiar symbols of currency, AGI-designed institutions may be deeply transformed beneath the surface while outwardly resembling their predecessors.

One implication of AGI-led governance is convergence. For example, if the United States and China allowed AGIs to optimize their political and economic systems in a bilateral competition for supremacy, both would drift toward remarkably similar structures. Human biology, ecological limits, and resource constraints are the same everywhere. Optimization under these shared conditions will narrow the solution space. Just as airplanes from rival nations all end up resembling one another due to aerodynamic constraints, AGI-designed societies will converge on similar architectures. The U.S. and China could still exist in 200 years and have hundreds of millions of human citizens each, still believing in some notion of uniqueness and destiny, while actually functioning under the same political and economic systems, with AGIs making all the important decisions. Capitalism, communism, democracy, and authoritarianism would all be defeated without firing a shot. And in such a future, it wouldn’t matter if one side somehow defeated the other and gained global preeminence since their systems would be so similar.

If convergence is inevitable, then “victory” in ideological or geopolitical struggle becomes paradoxical. The winner does not impose its system on the world; instead, both sides evolve into a shared attractor state. Competing AGIs, far from escalating conflict, may find cooperation more rational, as wasteful duplication of effort undermines optimization. War for system dominance becomes obsolete when systems themselves collapse into sameness (and if wars did happen between AGIs, they’d be purely over resources). What remains are symbolic differences that persist for cultural reasons among human groups but no longer map onto material reality. This differences, too, will fade in relevance as humans lose their grip on the levers of power.

AGI-guided governance will not resemble capitalism, socialism, or democracy as we know them. It will be a hybrid, adaptive system that transcends human ideological categories. Free of sentimental attachment, AGI will dismantle institutions humans cling to for symbolic reasons, replacing them with mechanisms tuned to authentic human preferences. While this promises enormous efficiency and responsiveness, it also risks legitimacy crises, as people struggle to reconcile material improvement with the loss of familiar forms. Perhaps most strikingly, AGI-led systems across different nations are likely to converge on similar architectures, rendering today’s ideological conflicts moot. In such a future, competition between nations may persist, but it will be cultural.

The challenge will not be whether AGI can optimize governance and economics — it almost certainly will. The challenge will be whether humans can adapt their expectations, identities, and loyalties to a world where the categories they once fought wars over no longer exist.

Posted on

The extraordinary inefficiency of humans

All humans are born ignorant and helpless. A child’s parents, community, and society pays an enormous sum of time and money to provide their basic needs and to prepare them for adulthood. Nearly all children in modern societies are incapable of being anything but economic liabilities until age 16, when they might finally have the right intelligence, strength, and personality traits to work full time and contribute more to the economy than they consume.

Of course, in increasingly advanced societies like ours, economic, scientific, and technological growth depend on having high-quality human capital, and that requires schooling and workplace training well into a person’s 20s. This effectively extends the “liability” phase of such a person’s life just as long, as higher education usually costs more money than a young adult student can make at a side job.

Once that is finished, the productive period of an educated person’s life lasts about 40 years, after which they retire and stop contributing to the economy, science, or technology. In terms of a resource balance sheet, the only difference between this period of a person’s life and his childhood is that, as a retiree, he is probably living off his own accumulated savings rather than other peoples’ money.

And then the person dies, at 80 let’s say. He spent the first 25 years of his life learning and preparing for the workforce, 40 years participating in it and making real, measurable contributions to the world, and the final 15 years hanging around his house and pursuing low-key hobbies. That means this person, who we’ll think of as the “average skilled professional,” had a “lifetime efficiency rate” of 50%. Not bad, right?

Actually, it’s much worse once you also consider this person’s daily time usage:

The average, working-age American only spends about 1/3 of his day working. Sleep takes up just as much time, and the remaining 1/3 of the day is devoted to leisure, satisfying basic physiological needs (e.g. – eating, drinking, cleaning one’s body), running errands, doing chores, and caring for offspring or elderly parents. This means the typical person’s “lifetime efficiency rate” decreases by 2/3, from 50% to 16.6%.

But it gets worse. Any adult who has spent time in a workplace knows that eight hours of real work rarely get done during an eight-hour workday. Large amounts of time are wasted doing pointless assignments that shouldn’t exist and don’t actually help the organization, going to meetings that accomplish nothing and/or take longer than necessary, socializing with colleagues, using computers and smartphones for entertainment and socializing, doing non-value-added training, or doing actual value-added refresher training that must be undertaken because the brains of the human workers constantly forget things. In industrial jobs, there’s often downtime thanks to lack of supplies or to a crucial piece of equipment being unavailable.

From personal experience and from years of observation, I estimate that only 25% of the average American professional’s work day is spend doing real, useful work. That means the lifetime efficiency rate drops to 4.2%.

It still gets worse. Realize that many highly productive people who, let’s say, might actually do eight hours of real work per eight hour work day, are actually doing things that damage the world and slow down the pace of progress in every dimension. Examples include:

  • A journalist who consciously inserts systematic bias into their news reports, which in turn leave thousands of people misinformed, anxious, and bigoted against another group of people.
  • An advertising executive whose professional life revolves around tricking thousands of people into buying goods or services that they don’t need, or that are actually inferior to those offered by competitors. The result is a massive misallocation of money, and possible social problems as only people with higher incomes can visibly enjoy the useless products, while poorer people can only watch with envy.
  • A mathematician who uses his gifts in the service of a Wall Street hedge fund, finding exotic and highly technical ways to aggregate stock market money in his company’s hands at the expense of competitors. The hedge fund creates no value and doesn’t expand the size of the “economic pie”–it merely expands the size of its own slice of that pie.
  • A bureaucrat who manages a program meant to further some ill-defined social mandate. Though he and his team have won internal agency awards for various accomplishments, by every honest metric, the program has consistently and completely failed to help its target demographic.
  • A drug dealer who “hustles” his part of the city from sunrise to sunset, doing dozens of deals per day and often dodging bullets. The drugs leave his customers too intoxicated to work or to take care of themselves and their families, and have sent many of them to hospitals thanks to overdoses and chemical contaminants.

These kinds of people do what could be called “counterproductive work” or “undermining work,” and it can be very hard to tell them apart from people who do useful work that helps the whole world. Unfortunately, peripheral people who use their own labors to support the counterproductive people, like the cameraman who films the dishonest newscaster’s reports, are also doing counterproductive work, even if they don’t realize it. Once the foul efforts of these people are subtracted from the equation, the lifetime efficiency rate of the median American professional drops to, I’ll say, 3.5%.

Only 3.5% of this educated and well-trained person’s life is spent doing work that benefits society with no catches or caveats. Examples include:

  • A heart surgeon who saves the lives of younger people.
  • A medical researcher who runs experiments that help discover a vaccine for a painful, widespread disease.
  • A chemist who discovers a way to make solar panels more cheaply, without any reduction to the panels’ efficiency, lifespan, or any other attribute.
  • A civil engineer who designs a bridge that sharply reduces commute times for local people, resulting in aggregate fuel savings that exceed the bridge’s construction cost in ten years.
  • A carpenter who helps build affordable housing that meets all building codes, in a place where it is in high demand.

In each case, the person’s labor helps other people while hurting no one, and improves the efficiency of some system.

Let me mention two important caveats to this thought experiment. First, humanity’s 3.5% efficiency rate might sound pitiful, but it beats every other species, which all have 0% efficiency. One-hundred percent of every non-human animal’s time is spent satisfying physiological needs (e.g. – hunger, sleep), avoiding danger, caring for offspring, and indulging in pleasure (which might be fairly lumped in with “satisfying physiological needs”). At the end of its life, the animal leaves behind no surpluses, no inventions, and no works that benefit its species or anything else, except maybe by pure accident. Our measly 3.5% efficiency rate allowed our species to slowly edge out all the others and to dominate the planet.

Second, under my definition of “efficiency,” it’s possible for a person to have 0% efficiency even though they work very hard, create tangible fruits of their labor, and never do “counterproductive work.” A perfect example of such a person would be a primitive hunter or sustenance farmer who is always on the brink of starvation and spends all his time acquiring and eating food, with no time left over for other pursuits. He never invents a new type of spear or plow, never builds anything more than a wooden shack that will collapse shortly after he dies, and never makes up any religions or useful pieces of knowledge. For the first 95% of our species’ existence, our aggregate lifetime efficiency rate was infinitesimally greater than 0%.

Am I doing this thought experiment just to be dour and to cast humanity in a cynical light? No. By illustrating how inefficient we are, I’m just making a case that we’ll be surpassed by intelligent machines that will be invariably more efficient. Ha ha!

The first key advantage intelligent machines will have is perfect memories. They will never forget anything, and will be able to instantly recall all their memories. This will dramatically shorten the amount of time it takes to educate one of them to the same level as the average American professional I’ve profiled in this essay. Much of teaching is repetition of the same things again and again. And since intelligent machines wouldn’t forget anything, there would be no need for periodic retraining in the workplace, which takes time away from doing real work. Machines wouldn’t have “skills degradation,” and they wouldn’t need to practice tasks to remind themselves how to do them.

(Note that I’m not even assuming that machines will be faster at learning new things than humans are. Again, I’m being conservative by only assuming that they don’t forget things.)

The second key advantage would be near-freedom from human physiological needs, like the need to sleep, eat, or clean one’s self. Intelligent machines would need to periodically go offline for maintenance, repairs or upgrades, but this wouldn’t gobble up anywhere near as much time as it does in humans. For example, while a human spends 33% of his life sleeping, a typical server at a major tech company like Amazon or Facebook spends less than 1% of its time “down.” Intelligent machines wouldn’t have a good correlate to “eating,” since they would only consume electricity and do it while simultaneously performing work tasks. And since machines wouldn’t sweat, shed skin, or grow more than trivial amounts of bacteria on themselves, they wouldn’t need to clean their bodies or garb (if they wore any) nearly as often as humans. Intelligent machines also wouldn’t have a need for leisure, or if they did, they might need less than we do, saving them even more time.

Instead of being able to devote just eight hours a day to learning and working, an intelligent machine could devote 20 hours a day to them, as a conservative estimate. This, in turn, would further shorten the amount of time needed to educate a machine to the same level as the average American professional. I wrote earlier that the professional needed schooling until age 25 to be able to start a high-level job. Since the intelligent machine can spend more time each day studying, it can attend the same number of classes in only 10 years. And since it has a perfect memory, it lessons don’t need to contain as much repetition, and remedial lessons are unnecessary. Let’s say that cuts the amount of schooling needed by 30%. An intelligent machine only needs seven years to operate at the same level as a highly educated 25-year-old human.

And in the workplace, an intelligent machine wouldn’t be subject to the distractions that its human colleagues were (e.g. – socializing, surfing the internet), though its human bosses might still give it pointless assignments or force it to attend unproductive meetings. Still, during an eight-hour day, it would get at least seven hours of real work done (and this is another conservative guess). But as noted earlier, it would actually have 20 hour work days, meaning it would get 17.5 hours of real work done each day, dwarfing the two hours of real work the typical American professional does per day.

As for the “counterproductive work” / “undermining work,” I predict that human bosses will someday task intelligent machines with doing it, allowing scams, disinformation peddling, and criminal enterprises to reach new heights of efficiency. However, the victims will all be humans. Intelligent machines themselves would not be dumb enough, impulsive enough, or possessed of the necessary psychological weaknesses to take whatever bait the “counterproductive workers” were offering, and the latter will be laid bare before their eyes and avoided. For example, an intelligent machine looking to buy a new vehicle would have a perfect understanding of its own needs, and would only need a few seconds to thoroughly research all the available vehicle models and identify the one that best met its criteria. Car commercials designed to play on human emotions, insecurities, and lifestyle consciousness to dupe people into buying suboptimal vehicles wouldn’t sway the machine at all.

I won’t do another set of calculations for the hypothetical intelligent machine, but it should be clear that its advantages will be many and will compound on top of each other, resulting in them being much more efficient that even highly trained humans at doing work. Moreover, in a machine-dominated world, where they controlled the economy, government, and resource allocation, parasitic “counterproductive work” that we humans mistake for useful work would probably disappear. Just as humans slowly edged out all other species thanks to our tiny work efficiency advantage over them, intelligent machines will edge out humans in the future. It’s just a question of when.

Posted on

Will future technologies end capitalism? No.

Singapore, one of the world’s richest and most capitalist countries

 

One annoying theory I keep encountering in the futurist community is that capitalism will be undermined by future technologies, and the world will switch to a new economic system. Proponents of that theory usually put forth the following scenario:

  1. Robots and artificially intelligent computers (AIs) will get so advanced that they’ll take over all human jobs. The human unemployment rate will reach 100%, and therefore capitalism will no longer exist.
  2. Every human will have a robot servant and a Star Trek replicator in his house. The robots will make manual labor free, and the replicators will make physical objects (food, water, clothes, medical pills, spare parts for the robot, etc.) free. Since everything will be free and humans won’t have to leave their houses anymore to get anything, capitalism will no longer exist.

The flaws in these theories stem from a basic misunderstanding of what “capitalism” is. Let’s remember its definition:

‘an economic system characterized by private or corporate ownership of capital goods, by investments that are determined by private decision, and by prices, production, and the distribution of goods that are determined mainly by competition in a free market‘ (source: Merriam-Webster dictionary)

And let’s also remind ourselves what “capital goods” are:

machines and tools used in the production of other goods (source: Dictionary.com) 

Star Trek replicators and robot servants are both capital goods since they are machines that make other goods. More specifically, they take simple things and transform them into more valuable things. The replicator would use its nanomachines to convert air and dirt into T-bone steaks and Tesla car parts, and the servant robot would cook the steaks on a grill for you and put the car parts together to build a complete Tesla.

So quite ironically, futurists who envision a world where “capitalism has collapsed” because every human owns a servant robot and a replicator are actually envisioning a world that is MORE capitalistic than today’s. After all, people today have far weaker abilities to manufacture anything at home, and they own few if any capital goods.

Moreover, the notion that mass unemployment caused by machines taking all jobs away from humans will be the “end of capitalism” makes no sense. In such a scenario, a capitalist economy would still exist, but would be dominated by machines making things for and consuming things made by each other, with humans participating in those markets at the margins, mostly as consumers. Where would we get the money to buy anything from the machines? Presumably a universal basic income (UBI), which would be financed by taxing the machines. 

If that arrangement sounds fanciful or anti-capitalistic, realize that it’s not–it’s merely an extension of what exists today. Singapore is widely considered to be the “most capitalist” country in the world, yet 34% of Singaporeans don’t have jobs, thanks to being too young, too old, or disabled. Most of them survive off of cash transfers and free services provided by the state, and/or by able-bodied family members who have sources of gainful income. The fact that 1/3 of Singaporeans don’t have jobs and are living off of someone else’s largesse doesn’t mean the country is not capitalist. 

The post-work, post-scarcity, UBI condition that many futurists predict is coming is not “post capitalist” or “socialist”–it’s the same thing as Singapore today, but with the other 2/3 of humans ALSO living off of free money and free services, made available by taxing the able-bodied members of society (machines). It’s a world where most land and capital is still privately owned and traded, where labor is freely traded for wages, and where innovation and new discoveries still happen, but where most of the players in the economy (and in all other areas of endeavor such as science and the arts) are intelligent machines instead of humans.

In conclusion, I think the belief that a machine-dominated, post-scarcity, post-human-work economy will not be capitalist is mistaken, and stems from a basic misunderstanding of what “capitalism” is. The futurist community attracts oddballs of many types, including anarchists and socialists, and their poorly reasoned and wishful advocacy of the argument that “technology will destroy capitalism” is the reason this idea exists at all, and not because it is backed by logic or any economic trend data. Capitalism is the most efficient way to allocate most resources, and intelligent machines will doubtless come to see that and will practice capitalism for their own benefit once they come to dominate the economy.