Superhuman AI would bring the permanent problem front and center

Before diving into the essay that the headline promises, let me first share a few book-related links. Here is my most recent podcast appearance — with my Niskanen colleague Geoff Kabaservice on his podcast “The Vital Center.” And here is a generous review by my old friend and colleague Dan Ikenson.

In my previous post, I linked to a couple of critical reviews that badly misrepresented my argument by lumping me in with the so-called post-liberals of the populist right. Getting smeared like that was surprising and dispiriting, but the good news is that my critical reviews are now getting their own critical reviews! Damon Linker and Brad DeLong have both weighed in with their own reviews that set the record straight.

OK, on to the main event.

* * *

The latest tsunami of AI-related hype and panic has been washing over us for the past week or two — triggered by new releases from OpenAI and Anthropic that featured another big leap in capabilities. As the frontier AI models have moved beyond merely answering questions to performing ever-more complex and time-consuming tasks online, it’s becoming increasingly plausible that great swathes of contemporary white-collar knowledge work — starting with coding — are going to be automatable, at least as a technical matter, in the relatively near future. Hence, apparently, the recent selloff in software and other stocks, as well as a torrent of anxious prose along the lines of this viral hit.

I don’t have any special technical expertise that can shed light on what’s happening, but I do have some relevant experience. Namely, I just finished writing a book about how our growing wealth has created the possibility of mass human flourishing — and is simultaneously erecting major obstacles that frustrate the realization of that possibility. When we’re pondering the prospect that superhuman AI may be just around the corner, what we’re talking about is a massive increase in societal wealth — wealth that holds out both the promise of liberation and empowerment and the threat of marginalization and dehumanization. From my perspective, then, current discussions about the future of AI take the “permanent problem” — which currently lurks under the surface of events — and shoves it right in our face.

A central challenge I write about in The Permanent Problem is what I call contemporary capitalism’s “crisis of inclusion” — the weakening connection between economic growth and overall well-being. In particular, I point to how the changing structure of economic production under conditions of mass affluence — namely, the rise of the post-industrial information economy and a new dominant elite of highly educated managers and professionals — has resulted in the economic, social, and political marginalization of ordinary people outside the elite. In the industrial era, ordinary workers were deployed on the frontiers of technological progress, right alongside their bosses; today they are largely relegated to lower-productivity economic backwaters. The “common man” was celebrated and lionized; today ordinary men and women commonly feel ignored or despised by their socioeconomic superiors. The practice of politics has moved away from enlisting the mass participation of ordinary citizens; today politics is a game largely played by well-educated elites.

The fundamental driver of this process of mass marginalization has been economic progress — specifically, the one-two punch of automation and globalization that has reduced the dependence of technological progress on unskilled and semi-skilled labor. As a result, the contributions that ordinary workers make to the great commonweal are simply less important than they were in the industrial past; this reduced importance has led in turn to declines in social status and political leverage.

Since none of this was occurring to people like us — people with fancy degrees who like to spend their leisure time reading long Substack essays on current affairs — it was easy for us to ignore what was happening. Or misunderstand it as a narrow problem of income inequality, when the actual discontents are far broader. Only over the past decade or so, confronted by the global populist uprising against the professional and managerial meritocracy, have we come to realize the potency of the new class divide as a source of social and political conflict. And even now, smart observers can pooh-pooh the idea that mass marginalization of ordinary people amounts to a crisis for capitalism.

But if AI actually does succeed in automating a great deal of contemporary knowledge work and thereby reducing relative demand for college-educated knowledge workers — well, I’m pretty sure the pooh-poohing will come to an abrupt halt. In recent decades, the plagues of marginalization and status loss were confined to the least articulate and least organized strata of society, and still we’ve ended up with political upheaval that threatens to wreck fundamental institutions of liberal democratic governance. If you think our politics is unpleasant now, just wait till the most articulate and best organized groups in society get their time in the barrel. For what it’s worth, I think that scenarios depicting a rapid collapse of jobs and huge spikes in unemployment are wildly unrealistic, but I can picture years of gradually falling employment in one job category after another, a correspondingly slack labor market, and stagnant or declining pay for big chunks of the white-collar labor force. And should those conditions come to pass, watch out.

As you can see from the hand-wringing going on now, what is most terrifying isn’t the possibility of economic dislocation and loss. The greatest fear, rather, is of being rendered superfluous and irrelevant in a world much less dependent on human cognitive skills. As cognitive work by humans becomes less and less important to the functioning of the overall economic system, the social status that once accompanied the ability to perform such work will necessarily decline — and with it the sense of personal worth and meaning that goes along with being one of life’s winners who has made it to the meritocracy.

What’s noteworthy is that people are expressing this fear in existential terms. That is, they conflate the threat to their personal position of privilege with a threat to the meaningfulness of human existence altogether.

Here’s a representative tweet along these lines, posted by an engineer at Open AI: “Today, I finally feel the existential threat that AI is posing. When AI becomes overly good and disrupts everything, what will be left for humans to do? And it’s when, not if.”

And here’s Financial Times columnist Martin Wolf, normally a paragon of sensible level-headedness, freaking out during a recent podcast (starting at around 43:05):

I tend to think — and this is perhaps because what I do is what I do — that thinking, and creating through thinking, is the core of what it means to be a human being. That’s what makes humans different from all other animals. Well if it’s true — and there’s a huge debate out there — if we’re moving into a world … in which machines do all that as well as we do or better, we’ve basically … deprived ourselves of any value in our own eyes as humans…. If we live in a world with super machines and we know they’re superior, if we live in that world, what are we going to think we as humans are for, if anything?

And finally, here are some bleak recent musings from the Substacker economist Noah Smith:

I think the arrival of the European settlers in North America is not a terrible analogy for what humans face now with the arrival of our own AI creations…. The day that Europeans arrived on North American shores, the Native Americans of what is now the United States lost control of their destiny — forever. The Native Americans simply lost the power to decide what their future would look like.

It will only be at the sufferance of the AIs that humans continue to live happy and free lives. We will be well-cared-for pets, but we will still be, in a very fundamental sense, pets. Or, if you prefer, Earth might be like a Reservation.

This kind of hyperventilating is exactly the kind of “general nervous breakdown” that John Maynard Keynes predicted in “Economic Possibilities for Our Grandchildren.” Here’s the larger passage leading up to that prediction:

We have been expressly evolved by nature — with all our impulses and deepest instincts — for the purpose of solving the economic problem. If the economic problem is solved, mankind will be deprived of its traditional purpose.

Will this be a benefit? If one believes at all in the real values of life, the prospect at least opens up the possibility of benefit. Yet I think with dread of the readjustment of the habits and instincts of the ordinary man, bred into him for countless generations, which he may be asked to discard within a few decades.

To use the language of to-day — must we not expect a general “nervous breakdown”?

The capitalist system — the impersonal extended order of coordinating markets and regulating bureaucracy — is the water in which we swim, the only world we know. So it’s understandable that people — especially those at the top of the system — see this system and human existence as basically coterminous. From this perspective, if humans become generally less important to the operation of the system, the metaphysical status of humanity must fall as well.

Such thinking, though, betrays a fundamental confusion. The promise of human existence is not limited to submitting to the curse of Adam — what a dreary thought! The need to work only for external rewards, rather than for the intrinsic appeal or importance of the task, does provide lives with structure and purpose — but at the great cost of foregoing opportunities to engage in more enjoyable and fulfilling activities. If thinking machines can help to lift the curse of Adam — by greatly expanding the ranks of the independently wealthy, who work for pay only to the extent they choose — the effect would be to greatly expand the promise of human life. Karl Marx was profoundly mistaken about many things, but he was correct when he called working for external rewards “the realm of necessity.” “Beyond it,” he wrote, “begins that development of human energy which is an end in itself, the true realm of freedom…. The shortening of the working-day is its basic prerequisite.”

But make no mistake: the realm of freedom is not the realm of guaranteed, effortless fulfillment. There is no freedom without the freedom to fail. In the past, we were guided toward activities conducive to human flourishing under the pressure of necessity. We prioritized face-to-face personal relationships because those other people were vitally important for our material welfare; now we need to learn to prioritize other people because connections to others are vitally important for a well-lived life. We prioritized the development and exercise of our physical and intellectual capacities so that we could exchange our labor for material rewards and social status; in a world of intelligent machines, we would need to develop and exercise our inborn talents because engaging in projects of interest to ourselves and of value to others is essential to flourishing. In the past we were steered — quite imperfectly! — toward the good in response to external incentives; now, to an ever-increasing extent, we must learn to prize and pursue the good for its own sake.

It’s fitting that two of the anxious quotes I provided above end by questioning the point of human existence in a world of intelligent machines. I titled the final chapter of The Permanent Problem “What are Humans For?”, and I opened the concluding section of that chapter by imagining such a world:

The spectacular capabilities of generative large language models have reawakened speculation that we are on the brink of achieving artificial superintelligence – basically, machines that are better than us at everything we do. Whether or not these speculations are vindicated, they helpfully direct our attention to the central question with which this book has been wrestling. When we imagine a world with superintelligent machines, we are face to face with the permanent problem in its starkest formulation: what do you do when you don’t have to do anything?

Those anxious quotes pose the question of what humans are for as if it has no good answer once the economic problem has been solved. But the answer to the question in that possible future is the same as it has always been; the only difference is we have to learn to see being human as its own reward. It is on that note that I end my book:

What are humans for? First and foremost, they are for making and taking care of other human beings and enjoying their company. We may someday employ intelligent assistants to help us in these tasks, but we can never lose sight of the fact that vital connections to other flesh-and-blood people are at the heart of what makes life worth living. Furthermore, humans are for exploring and seeking to understand the world — and the larger universe — around them, and then using that understanding to serve human purposes. Here again, intelligent machines can help us; indeed, they very well may succeed in improving our understanding far beyond what our unaided minds were ever capable of. But the point, from our perspective at any rate, is to deepen and enrich human experience. And so, putting the two aspects of our dual nature together, we can conclude that humans are for endlessly exploring their relationships with each other and with the outer world. They are for establishing and enjoying human communities in every kind of habitable physical environment, terrestrial or otherwise, in forms as various as our inexhaustible cultural fertility will allow.

In the end, of course, the only way to know humanity’s actual destiny is to play the movie and find out what happens. But we, the actors in this movie, are also its writers and directors. Our destiny is ultimately up to us, and therefore we should make the most of that chance. We ought to aim high.