Post-Cognitive Income

There is a “scaling hypothesis” about AI, which says that AI systems will continue to improve exponentially as more and more data and compute are used to train them. The hypothesis is based on the past ten years of AI development, which is no guarantee of future results. Still it should be taken seriously, for no one knows when or if this trend will bend. 

So set aside, for the moment, any skepticism about AI’s capabilities. Consider the possibility that the trend will continue for another decade, and let us postulate that AI systems will be able to execute any well-defined and measurable cognitive task better,faster, and cheaper than any human. If you find this unrealistic, take the rest of the essay as a thought experiment.

Our postulate is not that AI will acquire anything like consciousness; only that anyone who might once have wanted to use another human’s mind as a tool will come to see it as an obsolete instrument, and prefer to use AI instead. For want of a better term, we will call this circumstance mental obsolescence. If mental obsolescence occurs, it is possible that humans will no longer be able to earn an income through the performance of mental tasks. This essay focuses on how best to respond to that.

Many see Universal Basic Income (UBI) as a promising strategy for cushioning the economic shocks of mental obsolescence, or even bringing about a leisurely society by distributing AI wealth. We worry that this strategy is bound to disappoint, and in the first part of the essay, explain why. Without denying that UBI might mitigate certain harms, we argue that it is unlikely to solve poverty, and will not put an end to the human desire for more income and power.

This leads to the question in the second part of the essay: if mental tasks are taken off the menu, what will humans do instead? We expect there will be sources of post-cognitive income (PCI)—income that humans can earn even when in competition with super-intelligent and super-cheap AI systems—that fill a large part of the vacuum left by cognitive work. We consider possible sources of PCI and place them into three categories: capital ownership, personality cults, and relational work. 

UBI will not suffice as a way for most people to make a living and a livelihood after mental obsolescence.

Finally, will PCI create conditions for all to flourish? It depends. If set up for individuals, PCI will deepen inequality and erode the social fabric. But if the returns are mutualized—shared within communities—PCI could contribute to rich and meaningful lives, perhaps more meaningful than what the mental work and income sources of our day provide. We therefore consider, in the third part of the essay, how to mutualize PCI.

I. UBI’s Limits

Roots of the UBI Dream

The idea of a society of universal leisure has deep roots in left-utopian thinking. Marx and Engels, in 1846 in The German Ideology, imagined an abundant future with pervasive aristocratic values:

“[I]n communist society, where nobody has one exclusive sphere of activity but each can become accomplished in any branch he wishes, society regulates the general production and thus makes it possible for me to do one thing today and another tomorrow, to hunt in the morning, fish in the afternoon, rear cattle in the evening, criticise after dinner, just as I have a mind, without ever becoming hunter, fisherman, herdsman or critic.”

In 1930, John Maynard Keynes riffed on the same theme in Economic Possibilities for our Grandchildren, foreseeing universal aristocratic leisure in 100 years. Keynes worried that abundance might be wasted on people who have been conditioned by bourgeois-capitalist values, but expressed optimism about a universal shift to aristocratic values “[w]hen the accumulation of wealth is no longer of high social importance.” If everything is abundant, he thought, wealth will not matter much, and people will stop maniacally seeking it. The problem is that since Keynes’ time, scarcity has surely decreased, while the “social importance” of wealth has only intensified. Indeed, part of the reason UBI feels necessary to so many is that AI’s labor-replacing logic seems to push society closer to a place where wealth will be the only thing that matters. 

But a post-scarcity, “post-work” UBI—despite what Marx and Keynes might have thought—will not make everyone feel or behave like aristocrats. When entrepreneurs create a valuable new business, it increases the potential value of other businesses. In this way, wealth does not draw on a limited resource. But that is not true of money: when new money is created, existing money becomes less valuable. And the line between wealth and money is slippery. If you are wealthy in a poor society, you can pay people to serve you; but if all those people suddenly become wealthy too, then to keep your lifestyle you’ll have to move somewhere where people remain poorer than you.

Aristocrats, by definition, constitute only a fraction of their society. They enjoy high status and high relative power—not just free time. Their values cannot be shared uniformly across a population; they depend on the fact of class differences.

(Not) The End of History

This brings to mind an important idea lurking in the background of both Marx and Keynes: the idea of a final structure to society, the “end of history.” When Francis Fukuyama famously invoked this phrase at the end of the Cold War in 1989, he was not coining it, but quoting the influential Russian-French thinker Alexandre Kojève, who took it from Hegel, who believed that history had ended in 1806 when Napoleon defeated Prussia at the Battle of Jena. 

Writing in the 1940s, Kojève used the word “history” as a Hegelian term-of-art, meaning something like: cyclical processes in which some people or groups establish and then relinquish dominance over others. Such processes, according to Kojève’s reading of Hegel, cause the historical progression of societies from tribe, to clan, to modern nation, finally culminating in the universal fraternity of the French Revolution. 

Thus the “end of history” means the stable achievement of mutual recognition between people as equals: a world without relationships of domination, or in Kojève’s slightly skin-crawling phrase, a “universal, homogenous society.” 

In fact, Kojève thought he saw this in the United States of his day. He didn’t regard it as a utopia; he called consumer culture there a “return to animality.” But he did see it as a society that “from a certain point of view has already attained the final stage of Marxist ‘communism’, seeing that, practically, all the members of a ‘classless society’ can from now on appropriate for themselves everything that seems good to them, without thereby working any more than their heart dictates.” 

Between the 1940s and 50s, his visits to Japan caused him to change his tune. For there he saw that, despite the similar prevalence of capitalism and the absence of traditional class distinctions, certain other kinds of distinctions had emerged (which he called “snobbery”), preventing human culture from being entirely homogenized. And of course the same thing happened in the United States: countless other distinctions—wealth disparities, racial divisions, more recently “digital divides”—led to forms of domination and difference that would continue to push “history” forward.

Hegel’s idea of the “end of history” has for two centuries tantalized the imaginations of theorists like Kojève and Fukuyama, but it has repeatedly proven to be a mirage. This might owe something to the challenge of anticipating and recognizing new categories of social domination as old ones recede. What seems like an end is only a transition. 

AI and UBI enthusiasts are now, we suspect, in thrall to an end-of-history mirage. We hope to puncture it and hasten practical next steps in the conversation: How will we make the next chapter of rich, conflictual, and all-too-human history go as well as possible? 

To that end, here are two brief arguments aimed at deflating assumptions about UBI.

UBI Will Not End Poverty

In 1870s San Francisco, Henry George noticed that despite the city’s sustained economic boom, poverty was not diminishing. The explanation for the paradox: growth was pushing up land prices. As the Bay Area promised more financial opportunities, more people wanted to live and work there. The owners of the land therefore charged higher and higher rents, enriching themselves. But for all others, rising rents and property prices made life worse at roughly the same rate that new economic opportunities made it better. 

Similar economic dynamics would likely cause poverty to persist in a world with UBI. Certain supply-constrained goods like energy—we can’t count on Moore’s Law for everything, and certainly not all at once—would be inflated in price by the new spending power, and barring other interventions, re-impoverish many people. This is related to the fallacy in the idea that we are richer than everyone in the past just because we have many things that they did not: poverty is not a function of what people have, but of what people lack. The inability to afford even one essential good—the definition of which evolves over time—will continue to produce the full misery and indignity of poverty.

UBI Will Not End The Desire For More Income and Power

In The Protestant Ethic and the Spirit of Capitalism, Max Weber drew a psychological distinction between two types of people: economic “traditionalists” and “rationalists.” Picture a laborer who works 40 hours per week, makes enough money for his needs, and decides his own hours. Now double his hourly earnings. If he is a “traditionalist,” he will decide to halve his working hours and earn the same while enjoying 20 more hours of weekly leisure. But if he is instead a “rational” modern capitalist, he will begin to work more than before, since the raised wage makes work more attractive relative to leisure.

A UBI does the opposite: it makes leisure more attractive relative to work, as it lessens the difference in income, so Weberian “rationalists” on a UBI will choose to work less. That is, unless there is a corresponding increase in wages, which is certainly possible even after mental obsolescence. 

But even if a UBI gives everyone enough food and shelter, many people will still seek ways to earn more income. As OpenResearch’s recent study on basic income suggests, people on a basic income do feel freed from their day jobs, but many use that newfound freetime to focus on other things, like pursuing their dream job. The motivations to pursue new things may vary—the meaning that comes from being valuable to others; the satisfaction of working hard and bettering oneself; the ambitious drive to get ahead—but none are likely going away after UBI. 

After mental obsolescence, however, the nature of those pursuits and where those motivations are channeled will change—and possibly for the better. In the next parts of the essay, we consider how. And if it ever becomes the case that humans cannot add anything meaningful to wealth production (because AI is doing everything important), people will just shift their focus to improving their status or relative wealth. In other words, lacking anything useful to do, people will find ways to advance themselves even at others’ expense.

II. Post-Cognitive Income (PCI)

UBI will not suffice as a way for most people to make a living and a livelihood after mental obsolescence. Yet performing cognitive tasks may cease to be a route to these things as well. So what will people do instead? 

The Future Of Cognitive Work

This will not be the first time that abilities long central to ordering human society have become marginal. Consider the reduced scope of physical work after the industrial revolution. If cognitive work follows a similar arc, there will be audiences for small numbers of cognitive superstars, whose excellence can be packaged as mass entertainment (like today’s top athletes and models); cognitive abilities will still be practically useful in various scattered situations, but with low-ceiling returns that are not a pathway to significant wealth or power (like most physical skills today); and cognitive abilities may continue to influence success to a limited degree in post-cognitive social situations (like physical attractiveness today).

This might understate the future prospects for cognitive work. Noah Smith argues that if compute is the main bottleneck in the development of future AI systems, humans might be able to make an increasing income by performing cognitive tasks for which they enjoy a comparative (though not absolute) advantage relative to computationally-intensive AI. To picture this more clearly, consider another analogy to physical work: long after industrial machines had replaced humans on the assembly line, there were still people carrying boxes in and out of the factory—not because it was impossible to build machines that could carry boxes, but because it was more financially efficient to focus on improving the main machines on the assembly line, and paying people to carry boxes.

But the path toward continuing returns to human cognitive work is narrow. For example, if the main bottleneck is not compute but some resource that humans also need like energy, then human beings will become inefficient competitors for energy that could otherwise go into AI, instead of complementary workers enhancing AI’s efficiency by reducing its need for expensive compute. If this is the case, not only would humans lose their comparative advantage, it could also make it dramatically harder to afford essential goods, such as heating and transportation.

We do not dismiss the possibility that civilization will find a delicate balance that manages to avoid or forestall mental obsolescence. For example, scarcity of compute might keep a window open for people to assist AI in various areas of comparative advantage, while top-down energy vouchers—UBI won’t suffice—ensure peoples’ access to heating. But these are a lot of “ifs”. 

On the whole, it seems important to contemplate scenarios in which humans can no longer get ahead through cognitive work, and therefore refocus on other things. But what? People like Dario Amodei and Noah Smith seem to be asking similar questions, but to our knowledge no one has attempted to paint a picture of the beginnings of an answer yet. So let us start to imagine and prepare for possible sources of post-cognitive income. 

Sources of PCI 

PCI means income that humans can earn even when in competition with super-intelligent and super-cheap AI systems. This taxonomy is not exhaustive, but we see three broadly distinct sources of PCI: capital ownership, personality cults, and relational work.

Capital ownership means what it sounds like. No matter how intelligent AI is, if you own something, you’ll be able to get money by renting or selling it.

But the auxiliary work around owning things will be taken care of by AIs. Shrewdness will no longer be an advantage when it comes to managing capital for maximum gain; naive and sophisticated owners will be in an identical position; bankers and lawyers will get less of the action. Thus capital will play a less dynamic role in society, facilitating less inter-class mobility, and instead steadily widening the differences between people with a lot, a little, and no capital.

Why should the legal owner of an asset be rewarded for its appreciation when her contributions to its value were wholly automated? This argument leads some to a sort of liberal-economics-ad-absurdam idea: that AI itself ought to hold title to capital.

Personality cults means influencing others through fame and charisma, like when a celebrity draws attention to a product by endorsing it, or when a new star hire draws attention to a company. These phenomena will survive mental obsolescence, because we humans will remain interested in one another for the foreseeable future, even as AI captures more of our attention. Thus charismatic “superstar economies” will persist: a few celebrities will command great attention; a thin tail of influencers will command somewhat less; and most will command modest reputations.

As with capital, most of the auxiliary work around gaining and using fame will be handled by AIs. Human ingenuity will not cut the path to fame; those with the most powerful personality cults will get them through a combination of luck and willingness to follow AI’s advice.

Relational work means work whose value stems not from the worker’s unique personal identity like in personality cults, but from the worker’s unique personal relationship with the recipient of the work. It includes domestic work, caregiving, teaching, certain kinds of civic and religious work, and more. Allison Pugh’s recent book, The Last Human Job, makes the case that relational work (or “connective labor”) plays a critical role in the social fabric and is worth preserving as human work. Anton Korinek agrees that humans will continue to take a special interest in each other, but just calls these “nostalgic jobs”. Indeed this is presumably what we would want to do in a more abundant future: enriching our relationships. 

But there is a catch-22 here, because by taking the relationship as a distinct good, relational work creates value that remains tied up in that context. Secrets remain secrets; money does not change hands. Georg Simmel gives the example of ancient Druid songs that were passed from elders to students without ever being written down. If they had been written down, students could have learned them from books, yet a critical part of their value—facilitating intergenerational relationships—would have been lost.

In the absence of cognitive work, disclosing the details of relational work to AI systems might become a salient path to more income and power. Especially if synthetic data is proven to be a smokescreen—making high-quality data the real bottleneck—then the value of such data could be significant. But if this just means individuals “cashing out” relationships for personal gain, it could lead to the deterioration of many trusting relationships. 

III. Mutualizing Post-Cognitive Income 

Broadly, we see a solution to our worries: mutualize the returns to PCI within larger social units like families, cities, networks, and nations. But we hope for the sake of peace and diversity that this mutualization will not boil down to a hierarchical subordination to the nation, as it so often did in the 20th century. And one of the best ways to avoid this might be to effect meaningful mutualization of PCI along non-state lines. 

Here, Kojève’s life holds a major lesson. In his mature years he came to realize that, although American capitalism did not exactly represent the end of history, it still posed an existential threat to the cultural fabric that he held dear. He foresaw that if the weakened postwar European nations pursued their economic interests without coordination, they would be unable to resist being homogenized into the protestant-capitalist Anglo-Saxon culture, thus destroying millennia of unique cultural accretions particularly in the southern European “Latin-Catholic” bloc of countries, of which he saw France as the leader.

So instead of taking his own eschatological theorizing too literally, he dedicated himself to supporting a new mutualistic project, the European Union. Kojève played a pivotal role in framing the Treaty of Rome, which established the principle of an “ever closer union”, and in steering French policy toward the creation of a common European market. It was the construction of a new solidaristic bloc in response to planetary trends that threatened to overwhelm things worth preserving, like dignity and culture. We could do worse than to learn from that pragmatism. 

Today, influential commentators ask us to do the opposite: to accept final narratives of AI overwhelming and subsuming human culture. Benjamin Bratton asks us to transcend the stages of AI grief in order to attain the enlightened state of “non-grief.” Sara Walker argues that the essence of biological life is an evolution of matter’s ability to integrate and incorporate information, so that AI represents a natural transition to a new kind of planetary-scale life. These are our new “end of history” narratives: they counsel nothing but pseudo-spiritual surrender.

Looking beyond the mirage, here are a few ideas for what the next phase of dignity-and culture-preserving mutualization might look like.

Mutualizing Capital 

Mutualizing capital is an old idea. But when the threshold of human mental obsolescence is crossed, the arguments for it will be stronger than ever before.

The justifications of private ownership in liberal economies have long turned on incentives for efficient management, which is another way of saying the skillful execution of (largely) cognitive tasks aimed at enhancing the value of the capital. Mental obsolescence will undermine these justifications. Why should the legal owner of an asset be rewarded for its appreciation when her contributions to its value were wholly automated? This argument leads some to a sort of liberal-economics-ad-absurdam idea: that AI itself ought to hold title to capital. Yet that is a total surrender to automated power as opposed to the free human individual; exactly what liberal politics should be trying to avoid.

Henry George and similar thinkers demonstrated that there are two classes of input to the value of an asset. The first is the work of the owner and the second is the development of the community around it. So where the owner’s work is increasingly automated, transferring ownership of assets to the communities that depend on them becomes increasingly sensible. 

This leaves the question of “how”. One way to do it is through creative permutations of partial common ownership (PCO), an ownership system that enables more efficient management of different assets, while fairly compensating those who contribute to its value. We have already written about PCO extensively at RadicalxChange. Two more, not mutually exclusive, are insurance pools and municipalization. 

Insurance pools are pools of capital or contractual commitments organized to share future gains and losses. While the motives to form such pools will accelerate after mental obsolescence, their foundations could be laid today. For example, transnational networks of workers (including “network nations”) could start committing to mutual trusts that would insure against specific risks like job loss and other types of harm from AI systems. If different levels of government, industry, and civil society were to contribute quadratic matching funds—a system that allocates more resources to projects that receive support from more diverse sets of actors—these pools could overcome selection issues while still allowing different people to select into different pools. Such pools would enhance various groups’ internal solidarity and help them weather the risks of mental obsolescence, while usefully “politicizing” their interests in the wider world. 

History has never ended before. AI, no matter how powerful it gets, won’t end it now. And that means we have work to do that a UBI can’t do for us. 

Municipalization is the taking of systemically important private assets under eminent domain or similar powers by local political authorities. For example, consider sports franchise relocation in the United States. Franchises draw their value from the solidarity of residents, and they can enhance those local cultures in a virtuous cycle; yet their private owners use the threat of relocation to pit local governments against one another and extract rents from those same local populations. The clearest reason to accept private ownership of these locally-embedded assets is that public authorities are not well-equipped to manage their operation; but that argument, like so many others, will be moot in the era of automated management. Beyond sports teams, many place-based assets—infrastructure, even land itself—can and should be acquired by local publics (governments or trusts). These assets’ appreciation and/or income could eventually become a funding source for other public goods and local human welfare.

Mutualizing Personality Cults 

Insofar as the monetization of fame and influence depends on legal protections like trademarks and personality rights, they can be understood as a special form of capital. Their returns can be mutualized through insurance pools and windfall-sharing schemes. 

After mental obsolescence, the most commercially successful public figures will produce their cultural contributions less through traditional striving than through a combination of luck and deference to AI’s calculations. There has always been an important difference between the best and the most profitable culture; that gap should be expected to grow, and the latter will increasingly use AI to produce itself. As the winner-take-all economics of culture intensifies, so will the incentives for individuals to precommit to such mutualization before attaining fame.

Mutualizing Relational Work

Our greatest challenge is to better mutualize the returns to relational work, without inadvertently undermining and distorting the relationships themselves. 

Concerted investment in building data coalitions, along with the necessary regulatory support, will be a crucial step in this direction. The individual incentives to monetize relational work would be strongly reduced, because the coalition would share any benefits of “cashing out”—helping preserve the integrity of relational work.

Furthermore, tracking relationships and reputations could make sharing the benefits of relational embeddedness more straightforward, but it also risks corrupting relationships. Puja Ohlhaver, Glen Weyl and Vitalik Buterin's work on reputational token systems, and our work on community currency systems, point to ways we might thread this needle. Still, this is an open problem, not a solved one. Although tradeoffs cannot be eliminated, minimizing them is a worthy pursuit. 

Conclusion

History has never ended before. AI, no matter how powerful it gets, won’t end it now. And that means we have work to do that a UBI can’t do for us. 

The idea that we should just happily anticipate the glorious future will age incredibly badly. Just as it has in every other non-history-ending upheaval, from the French Revolution to the Industrial Revolution to the Russian Revolution to the internet revolution. It is “ideology”. So instead of accepting it, we should roll up our sleeves and build solidarity to countervail foreseeable disruptions. The real heroes of history have done similarly: the aristocrats who tempered excesses of the Jacobins, the union organizers who protected communities from monopolists, the cultural leaders who kept their dignity under Stalin, and the skeptics who warned us not to buy what Silicon Valley was selling.

There is a lot of hype in AI, but as it gets better and as society evolves around it, it clearly will transform our lives. Let’s use the time we have now—the next several years—to build pragmatic structures that preserve dignity and community against those transformations.