Technological Progress vs. Diminishing Returns

After progress. Image by Stable Diffusion 2.1.

The world of modern economic growth and rapid, continuous technological advance feels natural to us, as it’s the only world we’ve ever known. But of course it’s anything but natural: it’s only been around for a couple of centuries, while Homo sapiens has been on the scene for a couple of thousand centuries. So now, as the engines of growth and progress are sputtering, it’s appropriate to wonder: how long can this go on?

From my survey of the contributors to dynamism’s decline, I’ve come to the conclusion that the path of progress lies under a foot or two of molasses. Because of thermostatic cultural responses to mass affluence and the elimination of institutional competition, moving forward is going to be slower and messier than otherwise would be the case. But difficult progress is still possible progress. What I’m most concerned about is this: somewhere down the road, does the path of progress terminate in a brick wall?

In my last essay, I concluded that the future of technological advance looks much more robust than the future of measured economic growth — that is, of ongoing increases in aggregate GDP. But for both, the most plausible candidate for a firm limit on future advance lies in the combination of innovative activity’s diminishing returns and the looming end and possible reversal of population growth. If a worker-hour of innovative activity yields less and less progress over time, and the total number of worker-hours is flat or declining, how can progress escape this squeeze?

Let’s assume, for the sake of argument, that the concerted efforts of governments around the world to encourage couples to have more kids are unavailing. Financial incentives for new births and subsidies for childrearing succeed here and there in boosting birth rates, but not by enough. By the end of the 21st century, in this scenario, world population has already been shrinking for decades.

Recall, from my last essay, the economist Charles Jones’s grim assessment of the outlook under such circumstances: “When population growth is negative, both endogenous and semi-endogenous growth models produce what we call the Empty Planet result: knowledge and living standards stagnate for a population that gradually vanishes.”

Is there no escape?

Well, one possibility is to pull a Cadet Kirk and fiddle with the settings on this Kobayashi Maru simulation. Even if monetary incentives fail to maintain global fertility at above-replacement levels, population stagnation and decline might still be avoided.

I can see a couple of deus ex machina possibilities. The first is the development of anti-aging therapies that considerably extend the human life span. Widespread use of these therapies would reduce death rates, perhaps to the point that they fell below birth rates and the population started growing again. Furthermore, if people were living dramatically longer lives — let’s say the prime-age working years, once 25-54, expanded to 25-84 or 25-114 (that is, they doubled or tripled) — the twenty-something years devoted to raising kids would now constitute a much smaller fraction of people’s working lives. With this considerable drop in the relative cost of having children, we might expect a corresponding rise in the demand for kids and thus the fertility rate.

Meanwhile, there’s another new technology that could completely scramble current fertility dynamics, and it seems like less of a stretch than the conquest of aging. I’m talking about artificial wombs — devices capable of gestating babies from conception to delivery. Removing pregnancy and childbirth — and all their associated pains, inconveniences, and health risks — from the equation when weighing whether or not to have another child would significantly shift the calculus so that more couples decided in the affirmative.

Beyond that, more ominously, artificial wombs raise the possibility of reproduction outside the bonds of marriage, love, and sex. One can imagine the emergence of a reproduction industry, with either for-profit enterprises or charitable foundations that create and raise children — perhaps for a cut of the tax revenues they ultimately generate. Such a prospect strikes me as dystopian — but then again, so does the so-called Empty Planet result. Perhaps we’ll be forced to pick our poison.

It is noteworthy that, with the advent of female legal equality and birth control, the rate of reproduction — upon which hinges the growth and even continued existence of the larger society — now depends on decisions by women, and usually couples, to make huge personal commitments for intensely personal reasons. And there’s absolutely nothing about the contours of those choices that draws any distinction between having two or fewer children on the one hand versus more than two on the other – notwithstanding the fact that this is basically a life-or-death distinction as far as the future of the larger society is concerned. This hardly seems like a stable arrangement for the long term, so the possibility that alternative arrangements will be explored should not be discounted.

But let’s remember, we’re ultimately concerned about population growth because it’s a means to individual welfare — first of all, by creating more individuals; and beyond that, by encouraging economic growth and technological progress. If population stagnation or decline casts a shadow over future individual welfare because of its effects on growth and progress, doesn’t limitless population growth cast a shadow of its own?

Here we’ve entered the murky domain of population ethics, where we face what seems to me to be a genuinely vexing question: if our goal is to “live wisely and agreeable and well,” what’s the right total number of humans? Now, if the question is “what’s the right total number of humans in the Universe?”, I think the answer, in principle at least, is clear: the more, the merrier. With effectively limitless territory to spread out in and resources to use and consume, there’s no necessary upper limit on the number of people capable of living full and fulfilling lives. But if the question is “what’s the right total number of people on this one planet?”, things get much more complicated.

I’m emphatically on the other side of the anti-Promethean backlash, and I believe that we haven’t yet reached the limit on the number of people who can live wisely and agreeably and well together. With achievable advances in energy and food production, we can make possible materially abundant lives for everyone while arresting and reversing the great damage we’ve inflicted on the natural world. But even so, it seems to me like there has to be some upper limit on the total number of people who are capable of flourishing in a flourishing natural world.

At the very least, I don’t see how we could maintain the rate of population growth that obtained during the 20th century — that is, during the period when economic growth and technological progress were at their most robust. The world’s population quadrupled between 1900 and 2000 — does anybody really imagine that such a pace is sustainable indefinitely? In a couple hundred years, we’d be at 128 billion people, and it seems unlikely that the colonization of space and other worlds would have progressed to the point that a significant fraction of that total could live off-planet.

Even if technology had reached the point that 100+ billion people could live materially comfortable lives without wrecking the natural world — and that’s a pretty big if — I’m far from sure that such a world would compare favorably to one with a considerably smaller population. (In other words, I reject the repugnant conclusion.) For one thing, such a world would have to be characterized by extreme inequality with respect to physically scarce or “positional” goods. Access to such goods — visiting historical sites, enjoying natural beauty, living in the most beautiful cities, eating at the most fashionable restaurants — would have to be severely rationed, with only a tiny elite enjoying things that are now open to a sizeable global middle class. And given that, the richer we get, the more we seem to care about relative status and positional goods as opposed to absolute material wellbeing, the Crowded Planet scenario looks to me like another flavor of dystopia.

Meanwhile, even 20th century rates of population growth do not appear to be sufficient to save us from the squeeze created by diminishing returns. In their much-discussed paper “Are Ideas Getting Harder to Find?”, Nicholas Bloom, Charles Jones, John van Reenen, and Michael Webb estimate that total U.S. research effort (the number of full-time-equivalent workers engaged in R&D) rose by around 4 percent a year between the 1930s and the 2000s — but that the productivity of research effort was declining by more than 5 percent a year. Hence, even with a huge run-up in the number of scientists, researchers, and engineers, productivity growth fell over the course of the 20th century. And since U.S. population was growing just a little more than 1 percent a year during this time, the 4 percent annual increase in research effort reflected a progressive — and ultimately unsustainable — mobilization of labor into innovative activity. Accordingly, even if global population growth continued as before, we would still sooner or later hit a ceiling where we can no longer increase the share of our work force engaged in innovative activity. The real significance of the slowdown and possible end or reversal of population growth is that it accelerates the timetable for hitting this ceiling.

OK, so let’s call off the search for possible ways to keep population growth going and squarely face the issue: does the fact of diminishing returns mean that technological progress is ultimately doomed? And remember, because of low fertility, the stagnation that is gaining on us is closer than it appears.

I see one possibility for deliverance by another deus ex machina, as well as a number of stopgaps to forestall the coming of stagnation while we await our technological miracle. Let’s start with a look at the stopgaps.

First, innovation to date has depended overwhelmingly on progress made in the United States and other rich countries. To maintain the global intensification of research effort needed to keep innovation rolling, we will need the rest of the world —China, India, and especially Africa, as that is where most future population growth will be concentrated — to up their game and contribute more to technological advance.

Recognizing, though, that the leading countries will continue to lead, it is important that they continue to grow through immigration. The world’s centers of innovative activity can continue to enjoy sprightly population growth for a long time even in the face of a shrinking global population. Again, though, simply throwing open the borders isn’t enough: a concerted effort must be made to integrate new arrivals into social and economic life and ensure they are making contributions at the highest levels. If such integration could be mastered, ongoing migration could concentrate the world’s population in its most productive places while allowing extensive rewilding elsewhere.

Beyond continued development in less developed countries and accelerated immigration into rich ones, we need global initiatives for talent search and development to find the “missing Einsteins” otherwise at risk of squandering their immense potential. Meritocracy has gotten a bad rap as highly educated elites have moved to entrench their position and privileges through self-dealing and opportunity-hoarding, but the initial impulse to broaden the elite by including the best and brightest from the rest of society remains fundamentally sound. In a world of stagnant or shrinking population, the globalization of meritocracy is imperative.

Finally, we need to address the other side of the equation. All of the stopgaps above focus on trying to keep total research effort rising even in the face of zero or negative population growth. But we also need to be working on slowing the depreciation rate for research effort. Exhaustion of low-hanging fruit dictates that research productivity tends to decline over time, but we might be able to buy time by reducing the slope of that decline. There is considerable interest these days in finding new ways to organize science to get more bang for the buck; this interest needs active encouragement and exploration.

Stopgaps, of course, aren’t permanent solutions; at best they can buy some extra time. To defeat the natural limits to technological progress and keep diminishing returns permanently at bay, we will need a technological miracle – one that, it just so happens, many people now believe is imminent.

I’m talking here about the possibility of artificial general intelligence — or at least the continued rapid development of what we now call artificial intelligence technologies. With mechanization, we have been automating the work of human hands and backs for a couple of centuries; with computers, we have been automating a growing range of cognitive tasks for decades now. And with further advances in AI, what we are now hoping to accomplish is the full or substantial automation of the cognitive tasks associated with scientific discovery, invention, and innovation.

If that could be achieved, the constraint imposed by population growth on the possible total number of researchers, and thus on the pace of innovation, is eliminated: we would now be able to manufacture as many researchers as wanted. And if robot technology continues to advance along with AI, at some point perhaps AI could reproduce themselves. In such a fully automated economy, or at least substantially automated economy, productivity rates could soar.

The scenario of a fully automated economy — or at least one where AIs are able to automate their own advancement and manufacture — is often referred to as the “singularity,” a kind of techno-optimistic millennium in which economic output can double in weeks. This kind of growth explosion has always struck me as a pipe dream: perhaps it’s a failure of imagination on my part, but it’s just too over the top for me to put much stock in. My skepticism has been reinforced, though, by the work of the indispensable Charles Jones, who with Philippe Aghion and Ben Jones coauthored a fascinating 2019 paper on AI and the future of growth (see also this essay by Matt Clancy for an illuminating explication of the Aghion/Jones/Jones paper). Their work shows that, even when innovation is highly automated, as long as it is not completely automated a Baumol’s “cost disease” dynamic kicks in: the share of the automated parts of the economy shrinks as productivity growth races ahead of demand and relative prices decline, and the share of the less productive, human-operated economy expands. In their words, “growth is determined not by what we are good at but rather by what is essential and yet hard to improve.” Under these conditions, we could still liberate innovation from the population constraint but the singularity would remain confined to science fiction.

How then to wrap up my assessment of the future outlook for technological progress? The headline conclusion has to be that the path of least resistance leads to a continued slowdown in technological advance. Various hostile cultural and institutional forces are arrayed against it, adding additional frictions to a process that is naturally getting harder over time.

Given how much I lean on the existence of that natural constraint on progress — namely, diminishing returns to innovative activity — your confidence in my assessment should be discounted by the degree of your skepticism that such a natural constraint exists. And here I should note that the conclusion by Aghion, Jones, and Jones that ideas are getting harder to find is not without its critics — see here, here, and here. There are the usual objections to how productivity is measured, other technical objections to how research effort is measured, objections to the relevance of individual case studies (maybe research productivity is declining in specific fields but new, fast-growing fields are regularly opening up), and further objections to various methodological assumptions. Maybe ideas aren’t getting harder to find, but instead the organization of science and R&D is getting worse (that’s a valid possible critique of Aghion, Jones, and Jones, but for the growth outlook what matters is that overall research productivity is falling, not the reasons why). Or maybe the truly relevant measure of research productivity isn’t falling: maybe Big Science is a government-funded gravy train that attracts a great number of deadwood careerists while the cream of the crop that actually accounts for progress is as productive as before.

Having changed my mind about many things over the years, I am open to the possibility that I’m wrong here and that the walls aren’t closing in on the future of progress. I very much hope I’m wrong! But the idea that we first discover what is in plain sight, and that to keep discovering requires progressively more effort, seems intuitively obvious to me; diminishing returns are ubiquitous in economic life, so we should expect to see them in the pursuit of innovation as well. Meanwhile, the experience of modern economic growth — in particular, the juxtaposition of massive increases in innovative activity and steady or declining growth in productivity — points strongly to the existence of diminishing returns.

So I believe that the path of least resistance leads to stagnation. But the path of least resistance doesn’t have to be the path we take. There is much that we can do to improve our capacity for innovation and utilize that capacity more effectively. And there are real prospects for transformative technological breakthroughs that can catapult us to new levels of wellbeing.

We should be realistic. We should recognize what we’re up against. But there is no cause for despair, and much reason to look forward to what the future brings.