Stability in the Age of AI

AI is coming to fruition: Why is it frightening?

The fear that AI will upend the economic order is widespread, and the question of whether it's justified is one I want to take seriously rather than dismiss or amplify. To do that, it helps to start with what we actually know.

I'll preface this by saying that we have no evidence to support or evidence against the fact that the adoption of AI is different or the same as any other technological revolution of the past. This includes, for example, the industrial revolution and the adoption of computers.

What we have seen in both of these cases is periods of widespread, massive societal upheaval followed by a new status quo; however, the question is not so much that humanity will never find a status quo (with the exception of extinction-level AI catastrophe) but rather, what would that status quo look like? After all, after the industrial revolution, it was not so much the case that all manual labor vanished, but rather that manual labor changed and machines were able to do a large amount of the largely automatable work. Similarly, with the adoption of computers, there were a number of jobs whose entire demands had changed with their adoption.

Essentially, this begs the question: What does societal stability look like in an age of artificial intelligence?

I'd like to begin by breaking down one of the most essential core underpinnings of a modern society, which is the idea of a stable income. The reason why people pay for housing, for example, is because they earn wages in a stable fashion that they can continue to regularly pay for a house, which provides stable income to the owners of that house or banks, which in turn can provide stable loans with regular interest. This also is the underpinning policy of insurance, taxes, budgeting, and stocks. Insurers, for example, take on risk with the assumption that the larger insurance fund will continue to sustain itself due to the stability of income of all its constituents. Similarly, the government budget is set with an expectation of taxes, which is calculated based on the incomes of some 300 million Americans.

Thus, the reason why the loss of so many jobs is so dangerous and causes so many downstream effects is because the idea of stable income is undermined. Without a well paying job, we do not have a system set up to provide a high, stable income, which makes loans harder to give because interest is therefore not reliable. Tax collection is not reliable. Budgeting is therefore also not reliable. Insurance becomes unreliable, and the many downstream effects talked about as a result of the loss of jobs become realized because of this crucial underpinning.

A Way Forward

For the sake of discussion, let us imagine a future where AI has replaced so many jobs and it is impossible to create a new stable status quo of wage labor for the majority of people. We instead raise the floor so that individual wage labor is no longer required for stability. This has been discussed at length and in far greater detail than I should like to write about, but in short, solutions look like universal basic income, free social programs, free health care, free housing, etc. Now the question arises, “what do people do instead?” Here, there are two paths forward.

In the first case, AI can literally do anything a human can. This has been discussed extensively since at least Keynes's 1930 essay “Economic Possibilities for our Grandchildren”, which predicted that solving the economic problem would force humanity to face its "real, permanent problem" of what to do with freedom from economic necessity. Simply put, the recourse I find most promising is a world of recreational achievement, like chess, Rubik's cube solving, or body building (but for every activity, both cognitive and physical).

The second case is one where humans still have something to contribute, namely that AI will only be effective at well-scoped, defined tasks, and not chaotic, divergent thinking tasks. In this world, traditional wage labor still disappears. This is more or less the job displacement fear that we are seeing at present, often disguised as being the first option.

But this doesn't have to be a case for instability.

Some Definitions Before We Get Started

I'll define traditional wage labor here to be labor in which an employee executes on a set of instructions to accomplish a task. Not much agency, and primarily convergent thinking.

I'll also define what being a “founder” is. To not confuse it with the Silicon Valley archetype, I will state that a founder is “a high-agency individual who is able to identify a gap and fill it.” (Metaphorically, of course, unless one was to invent a novel method of caulking). An employee within a large corporation who spurs a new initiative that solves a real problem is a founder. Similarly, a researcher is also a founder, filling gaps in scientific knowledge.

This definition is a vector, in the sense that it must have both magnitude and direction. You might notice that your team needs a slide deck made by tomorrow, so you go ahead and do it despite no one assigning you to that action. That is founder-like, just with minimal magnitude (and the initiative is appreciated nonetheless).

The Power Law

In a world where traditional wage labor is gone, we must shift to a system of primarily divergent thinking. Not working to accomplish a predefined task, but rather thinking of as many new tasks as possible, turning the chaos and hard-to-find needs of the real world into well-scoped problems that AI can then execute on.

To make this concrete, consider what work actually looks like in this system. A researcher notices that a particular intersection of two fields hasn't been explored, scopes a research question, and hands the literature review, data analysis, and initial drafting to AI. An designer develops a new aesthetic direction by combining unrelated traditions, then uses AI to produce variations at scale and iterate on what resonates. A community organizer notices that a specific group in their city lacks a particular service, designs the structure of an organization to provide it, and uses AI to handle the operational scaffolding. A founder in the traditional sense identifies an unmet need in a market, defines the product, and uses AI to build, test, and ship it. In all of these cases, the human contribution is the same: noticing the gap, defining its shape, and judging when the proposed solution actually fills it.

However, every idea is not going to be successful. Traditional wage labor is already scoped, and it is largely the job of a few higher-ups in organizations to be high-magnitude founders. In our new system, most individuals are doing lower-magnitude gap-filling work, the kind that quietly improves a team, a field, or a community. A smaller subset takes higher-magnitude shots, and the success rate of those shots is not high (any entrepreneur will tell you this). The power law describes what happens at the tail, where a small number of high-magnitude attempts produce outsized impact.

The implication is that economic activity in this system is fundamentally different in shape from wage labor. Under wage labor, output is roughly linear in hours worked. Under founder-economy work, output is heavily skewed. Most attempts produce modest value, a few produce enormous value, and the system as a whole produces innovation at a rate that the wage-labor system, with its narrower band of who gets to scope problems, cannot match.

A Different Type Of Floor

UBI on its own is not enough. It's a safety net that catches people when the wage market fails them, but doesn't give them opportunities for growth. A floor that only handles survival produces a population that can pay rent but cannot meaningfully take shots, as they lack the compute, the tools, the time, the networks, and the slack required to translate a spotted gap into a filled one.

A particular concern raised in some technology circles is what happens when AI comes to full fruition under a UBI-only regime. Those without pre-existing capital are stuck. Their UBI covers survival but provides no surplus to invest, no compute to build with, no time to spend on ventures that might fail. Meanwhile, those who entered the AI era with existing wealth can deploy it, take shots, and accumulate further. The result is a two-tier society; this is the current startup demographic problem (where the well-resourced can found more easily than the unresourced) made permanent and universal.

What we need instead is a floor that is also a runway.

Consider what a startup founder actually receives when they raise a seed round. It's not only money for food and housing, but also access to legal ownership, distribution channels, freedom to work on their venture, and most importantly, the social permission to work on something for a year with no realistic expectation of success.

So, let's generalize this.

My proposed runway would have a foundation that is often coupled with UBI, namely access to healthcare and education. I particularly highlight these because in a world without traditional wage labor, tying healthcare to employment is a moral wrongdoing. Similarly, quality education is necessary for innovation and gap-filling, and is a necessary public investment in order to provide a foundation for future founders. On top of this foundation, access to compute and intelligence is necessary, and must be treated at the level of importance of electricity and broadband. As for less centralized change, time to attempt things must be protected institutionally, to be able to attempt things without immediate progress or results. Culturally and institutionally, failure must become accepted and even encouraged.

I'm not simply proposing a more generous UBI, but rather centralized and distributed institutional and cultural change. Traditional safety nets and UBI ask, “how can we prevent people from destitution when traditional labor and societal frameworks fail them?”. The founder runway asks, “how can we enable everyone to participate in a founder-dominant economy where the primary activity is identifying and filling gaps?”

Money Money Money

The obvious problem is where the funding for this runway would come from. To fund UBI, there exist a small number of meaningful solutions, the majority of which boil down to higher corporate taxes or individual wealth taxes. To be clear, both are valuable propositions, but I will focus on a different (or perhaps modified) funding methodology that makes the runway possible.

If AI development truly replaces wage labor at the scale discussed, the economic surplus that previously flowed to workers as wages will go elsewhere: to AI labs, capital owners, and firms deploying AI. The underlying principle to generate funding for the runway is to take some portion of this economic surplus and return it to the population that the AI replaced. The exact mechanism is a policy choice: it may look like public ownership in foundational AI infrastructure, taxing AI generated revenue, compute dividends, or an AI sovereign wealth fund (like Norway's, but funded by AI rents rather than oil).

There is significant overlap between the firms that would be affected by broad AI-surplus capture and those that would be hit by individual wealth or corporate taxes. Many of the same names appear on both lists. But the two approaches are not equivalent.

A runway funded by discretionary taxation of individual winners is fragile. Winners use their wealth and political influence to reduce their own contribution, eroding the floor over time. This is the well-documented historical pattern of redistributive systems: they get whittled down by the people they most depend on. Instead, the runway should be funded through a dedicated, rule-based mechanism that isn’t left up to ordinary politics, like Social Security or the Federal Reserve, rather than discretionary annual spend. The same firms still contribute disproportionately, because their share of the surplus is disproportionate, but the system does not depend on any individual entity's continued willingness to pay.

So, What Are We Left With?

What emerges as a result is neither a society lacking stable economic activity nor one where wage labor is artificially manufactured for the sake of preserving human work.

Instead, survival and basic comfort are decoupled from economic activity, while cultural and societal growth is driven by a vastly greater population taking shots in every domain. Most of these attempts will fail. That is the nature of a power law and is not a flaw of the system but a feature of it. We will see an explosion of innovation in scientific research, art, institutional design, ventures, communities, and all other ways humans fill gaps. This is what societal stability could look like on the other side of the AI transition. It's not stability that comes from individuals earning predictable wages, but stability that comes from absorbing AI displacement and channeling it into something generative.