Isolated narratives of progress
How analyses focusing on growth malaise often miss the point when it comes to technological progress
[Argues the Great Stagnation is not Great, or strictly a Stagnation in the technological sense, but is rather what progress looks like when it’s the sum of multiple subsector S-curves rising at varying speeds. It seems to pause while their interactions with each other aren’t strong enough yet to create a positive feedback loop. Necessarily speculative, though held with a reasonable degree of credence.]
The literature on the Great Stagnation grows and grows. Most recently Jason Crawford also wrote an article called Technological stagnation: Why I came around where he argues that the problem is real, joining an esteemed pantheon who have articulated this very thesis. And there are counterpoints where people point out that you can't say there's no progress, because look at genomics, computation, communication. And even others who argue that even in the world of atoms there's progress, steady and unspectacular progress, during the supposed dark age we're in.
Jason's articulation of this problem, paraphrasing Peter Thiel, is that in our current paradigm of technological progress, in the third industrial revolution, we're seeing progress only in information technology and nowhere else. When you compare it against the previous eras, we see progress in energy, transportation, medicine and manufacturing (in his schema). The thesis that's articulated in the piece suggests that we have S curves for industries that have been stalled in multiple domains. But my contention is that there's a further nuance here.
It's become straightforward these days to say that our effectiveness in pushing the S-curve upwards in sectors like energy and biology has stalled. But if we want to actually address the problem we need to have an opinion and an argument about why the S curves stalled and never took off again. I've written in the past about the phenomenon as part of the great polarisation and called it a punctuated equilibrium theory of progress.
So this is going to be an exploration on why this seeming productivity slowdown actually happened within certain sectors. I call it the S-curve dependence theory, or "standing on shoulder of giants" for the cheap seats. To start, let's use Jason's preferred taxonomy of technology/ industry. He suggests that within the three Industrial Revolutions, the progress we made spans the following categories:
This isn't a bad categorisation. We could legitimately argue whether Material Science should be added to the list, or whether we should combine Manufacturing and Energy in some places. We could argue if Agriculture should even be there. But while all of those could still be true, we'll still see the same trends of decreasing productivity in some categories vs others. So let's stick with this list as a good enough guide, and not worry overmuch about the exact boundary drawing that would make this accurate or more tractable.
The three revolutions for what it's worth were:
1700s-mid 1800s - comprising mechanisation and steam engine
Mid 1800s to mid 1900s - comprising chemistry, electromagnetism and microbiology
Mid 1900s to now - it's primarily been around electronics and communications
Various other sources all say there have been similar numbers (Salesforce Trailhead, IED, We Forum, even Wikipedia) so I think this is a better boundary-drawing conclusion. While it isn't as clear a delineation as we would like perhaps, it's nevertheless at a high enough degree of abstraction that it's instructive.
There are two important aspects to remember when we look at the three revolutions described above - 1) that they weren't as contiguous as it would seem from the timeline, and there were clear slowdowns punctuated amidst them, and 2) the social development underscoring standardisation, economics surrounding production techniques and just pure knowhow transfer were also massive catalysts for industrialising the change.
Now, let's have a look at what it means that the progress in each of them has flattened at the top of an S curve. Vaclav Smil, the eminent interdisciplinary professor from the University of Manitoba, wrote eloquently about the industrial revolution that created the modern world, and called this the "Age of Synergy".
So let's have a look at the movements category by category. If you're really bored by the historical montage that's coming up below, because it's too long or too short, believe me that I feel your pain and feel free to scroll to point II below.
Energy: The previous kick out of the S curve in energy came about from the invention of the steam engine, courtesy James Watt and others, and emergent understanding of thermodynamics. It also saw the emergence of oil as an energy source, improvements in iron making processes including puddling process. It also saw the emergence of machines to take advantage of that abundant energy source.
What did it take to make the machines come about? In the second industrial revolution, it required some combination of ability to refine petroleum and advances in basic physics enabling the construction of generators and motors.
In the second industrial revolution, the Bessemer process also helped make better and cheaper steel. This helped make the electrical system work better. There was also the discovery of iron making, puddling process and hot blast. Thermodynamic theory advanced at a rapid pace.
Machine tools got built and systematised and started becoming common, enabling high degrees of systematisation and mass manufacturing. There was the discovery of the light bulb too that happened, also helped along not just by energy breakthroughs but also improved materials design!
Not to mention, Faraday's work on electromagnetic fields helped things along quite a bit. Maxwell's unification of electromagnetic theory meant that there was significant advances that happened in the fundamental realm.
Manufacturing: The first industrial revolution came with the steam engine. The engine came about from advances in basic physics and chemistry. There were also the invention of machine tools, starting with the boring machine by John Wilkinson. The successive machines enabled the construction of interchangeable parts, and made mass production feasible. From a Northwestern paper:
In the half century following the invention of the fundamental machine tools the machine industry became the largest industrial sector of the U.S. economy, by value added.
Power looms were built up over a century from de Gennes onwards. There was the invention of flying shuttles and spinning jennies to speed things along. Energy dense coke got found as a replacement for charcoal. There were significant advances in chemistry including production of soda ash and bleaching powder. The invention of steam engine pushed the industry along fast, as a great way to supply power, and it itself depended on steel and materials.
The second industrial revolution saw new materials being developed that helped with manufacturing. The Bessemer process, courtesy Henry Bessemer, sped up the creation of steel from weeks to 30 minutes, and completely disrupted the process. This helped make cheaper and better steel, and reduced the cost of steel by 80%.
This affected construction too, and helped bring about the skyscraper. If you want a framework lattice that could handle a giant, tall and heavy hunk of concrete, then you need steel girders.
In the third industrial revolution we have made these processes more efficient and computerised them. But there's not been any step change in our ability to produce more, produce better or even cheaper. We do have smaller plants, but not decentralised enough that it's ushered in a new era.
Agriculture: The first industrial revolution still thought of natural fertilisers as key, and helped kickstart wars over guano. and saltpeter (also useful in gunpowder).
The second industrial revolution helped bring fertilisers and pesticides to fore, changing agriculture beyond recognition. The growth of both of those came from advances in chemistry. Bear in mind these weren't "nice to have" inventions, but rather literal life-savers as they helped prevent the Malthusian apocalypse that many people foretold.
The Haber-Bosch process, discovered by its eponymous inventors, helped create vast quantities of ammonia, the ingredient in fertiliser (and in a historic twist, also useful in gunpowder). Along with advances in chemistry to help with superphosphates and ammonium sulphate, this transformed agriculture as we know it through pesticides and fertiliser.
Transportation: The first industrial revolution came with the steam engine, helped along by James Watt and co. That helped the making of steam ships and locomotives and brought the world closer together. Energy dense coke replaced charcoal and helped push the locomotive train onwards. Cast iron helped, and later the invention of steel. The increase in coal mining technique sophistication and purification processes helped make this happen.
The growth of the railroad network and the steel industry were tightly coupled, with each increase in supply creating demand in the other, in a perfect virtuous circle.
The second industrial revolution brought with it the combustion engine. Actually it started with the electric engine and then moved to combustion. That kick-started the automobile filled world we live in, and the reason why Ford Prefect chose his name.
The building of locomotives was almost entirely a result of the availability of high quality, cheap, steel. Without it the cost of rails would've been 5x what it was!
Information: The second industrial revolution got its first major advances in information technology through better electronic communication. This period saw the invention of the telegraph, the telephone, the radio and ended with the television. Not bad for a century or so. The discoveries were all highly interlinked with previous discoveries made in physics, including discoveries by Oersted, Volta, Ampere, Faraday and Henry, the great physicists of the 19th century.
This went into hyperdrive later on with the discovery of the electron late in the 19th century. Maxwell's equations created a unified theory around electromagnetism. But most interestingly, to help fulfil Theodore Vail's vision of making AT&T work coast to coast, Frank Jewett, who was in charge of transmission engineering, realised that he could use the "electron beams" that the physicists had been studying, and asked his friend in University of Chicago to send one of his top students. That's how the vacuum-tube amplifier was built and commercialised. It also kickstarted Bell's tradition of paying top PhD physicists to come and do communications research!
The next phase of pushing the communications revolutions in the 1920s onwards relied on further discoveries made in the materials era. They got a way to solve noise vacuum tubes through better magnetic materials used in transformers and loading coils. There was significant research into the ways in which materials like alkaline-earth oxides or permalloy could help make the communication transmission less noisy.
I could probably quote the Physics and communications industry paper ad nauseum, but suffice it to say that the communications industry got a boost from pretty much every side. Chemistry and materials science that helped lay the foundation for the actual bits to move around, core physics to create the theoretical foundations and also to allay the engineering roadblocks, and the discovery of new fundamental forces in the universe to help think about new paradigms of information transmission.
And all of this without even starting in on Shannon and how he kickstarted an entire industry! Or how computers essentially rode that wave to help affect pretty much everything we see and feel around us.
Medicine: The first industrial revolution had the discovery of vaccines, starting with the technique discovered by Edward Jenner in 1796 for smallpox.
The second industrial revolution had the discovery of the germ theory of disease, and with it a dramatic reduction in disease. There were antibiotics discovered in the latter part of this era. And vaccines took a giant step forward with Louis Pasteur who after establishing the germ theory, not content to rest on his laurels, also helped create artificial vaccines.
The third industrial revolution had the discovery of DNA in its timeline at the start. But most excitingly it's the era that's seeing the furthest progress today. It take the vaccines, for instance, to its logical conclusion by saying you don't put the live virus in your body (like in the first industrial revolution), or attenuated virus in your body (like in the second and to-date the third industrial revolution), but rather use only specific parts of the virus to generate antibodies against it. First this was done by reengineering some cell like baker's yeast, but now it can be done through RNA directly.
Ok that was a fair bit of history. And as with most things historical, there's more variables than a story to stick them onto easily. I skipped most of what's important and tried to hit some of the major notes.
So let's have a look at the technological and scientific dependencies we relied on to get each of the breakthroughs that we looked at. This is the summary table version.
(If you want to mess with the table - google doc link here.)
Moreover, when we look at all the major breakthroughs that happened in the world together, a couple of trends appear:
Each invention was highly interdependent on and coevolved with previous research that spanned basic science, materials, economic knowhow and more
The discoveries and inventions were all pushed forward inexorably by at least one of the inputs becoming exponentially cheaper - e.g., steel, ammonia, rail and more.
So when we ask "did the IT revolution increase our productivity" we're asking the wrong question. Or at least a question that's at an insufficient degree of abstraction. The question should be how many different sectors is the IT revolution touching, and how many other technologies are also changing which can affect the ecosystem and coevolve.
The interdependence amongst the innovations leads me to believe that what we see and call the Great Stagnation is anything but. It's a Punctuation.
Jason summarises the problem we're seeing in the Great Technological Stagnation theory as the fact that we're just not seeing progress across all six categories. As he summarised it:
3 in IR1: mechanization, steam power, the locomotive
5 in IR2: oil + internal combustion, electric power, electronic communications, industrial chemistry, germ theory
1 in IR3 (so far): computing + digital communications
But the fact that we're only seeing one major category changing, even apart from the fact that boundaries around categories are incredibly porous and hard to draw neatly, is not a function of the effort we're putting into the other categories.
The fact of the matter is that the general purpose technologies that we built, and which had disproportionate impact across all categories, they're few and far between. Once we built steel, it impacted everything. One we invented electricity and the laws of electromagnetism, it affected everything. But getting to the benefits of both also required advances in chemistry, in manufacturing the right compounds, in thermodynamics, in automobiles and motors, and so much more, not to mention the advances in humanity-organisation-techniques that we were forced to invent along the way. They all reinforced each other in a positive feedback loop. Advances in one is what led to advances in the other.
Inventions and innovations are random. When at the earlier stages of an S curve the growth is slow. Later it's steep, and then it's slow again. But the inflection points aren't predictable. We can't forecast breakthroughs. Sometimes it can take much more time and effort before one realises whether you're affecting a paradigm shift à la Kuhn.
For instance, let's say the next S curve we need to innovate on energy requires, say, breakthroughs in information, materials and transportation. Each one of those is built up of another several innovations that also have to happen.
As the dependencies get larger, the possible combinations that can be explored in the idea-space also get larger. This model I’ve been playing with with serves as an inspiration. And as the dependencies get larger, you have to wait occasionally for each part to get ready and come up to the fun part of the S curve.
The chain of reasoning is as follows:
Progress overall depends on progress in sectors A, B, C, D, E
Progress in these sectors depend on insights within their subsectors + synthesis of insights across sectors
As the number of sectors grows this is a harder problem - the idea surface grows quickly
Synthesis becomes an even more important skill, helping collapse the idea-space by synthesising new fields
Once we have a manageable number of synthesised new insights, if there's talent clusters available, we can see positive feedback loops take off and generate growth for everyone
The good news is that the dependencies don't just keep going up as the factorial of the possible number of elements. They can also collapse, as happened with the world of electromagnetism, or neuroscience. But we still need the process to work itself through.
With just the two postulates that we saw above on the trends we see behind each breakthrough, for us to see broad economic growth we would need to see multiple strands coming together.
As I've argued before, the TFP stagnation is a real phenomenon, but it hides in its easily explainable form an idea that there's a monocausal view of the cause too. There might not be. So let's posit the following and connect the dots:
TFP is stagnating and with it parts of the GDP
TFP is a way of assessing an overall productivity measure which is the sum of measures that arise from multiple technologies acting in concert within the economy
These technologies don't all rise and fall together - they progress at varying rates, they have breakthroughs at varying times, and their progress and ‘rate of breakthroughs’ are highly interdependent with each other
Therefore our null hypothesis should be that overall growth will seem to have punctuated by erratic periods of seeming stasis, a punctuated equilibrium model of growth
This can be true even if it’s the case that we’re over-regulated, we’re under resourced, we’re plagued by the coordination tax of management and bureaucracy, and we’re mired in politics everywhere.
Nintil's argument that there is no great technical stagnation is another aspect of exactly this phenomenon. He analyses individual pieces of technological progress and shows how most of them continue to see steady progress in ways that count despite the seeming flatlining of TFP measurements.
And that makes sense.
The trouble is that the function that combines some of those technologies to create the next breakthrough doesn't seem to be happening. And without that there are no positive feedback loops either.
Sometimes this can feel fatalistic. Because the answer to when we might be able to get the TFP rates back up and aggregate productivity to rise is that we need to wait for said conditions to arise across multiple technology sectors, which feels a tad tautological.
Jason ends with an exhortation that we should also focus on atoms, joules and cells in addition to bits. But we already are. There's roadblocks for sure - bureaucratic, regulatory, and more. I've written about a few - on how hierarchies suck and how as we grow larger organisations and ecosystems they create a complex set of tradeoffs.
But we're not at a standstill. We're in a world that sees plenty of isolated pieces of growth and not enough synthesis.
Is this similar to the calls for more polymaths or interdisciplinary studies? Yes, in part, with two caveats. I'm not sure whether aiming for more interdisciplinary research is enough to make interdisciplinary breakthroughs happen, but it can't hurt! And usually the research being interdisciplinary is not the object, it's that certain goals can only be attained by the medium of interdisciplinary research. And depending on how far down the rabbit hole you go, all research is interdisciplinary research. The former requires funding, the latter requires a vision.
If we assume that the inputs are fixed and the same, just that their efficiency increases as we learn more, then we get the exponential growth mode that Robin Hanson modelled out. This shows how there might be gradual transitions between the modes of growth, but the "stagnation" while waiting for the next phase of growth is essentially unpredictable. It's as if we're waiting for someone to make the next breakthrough, and once they do, voila! Or you're waiting for the (randomly distributed) next valley to get discovered so that the wave of progress can fill it.
(As I see it the death knell for this thesis is definitional. Sufficiently fine grained definitions of Innovations, of breakthroughs and of historic antecedents, will make it seem as if we're creating narratives by drawing arbitrary boundaries around complex systems. And yet, in the absence of that were left floundering in an epistemic fog.)
The call to make progress our yardstick by Tyler Cowen and Patrick Collison is essentially to solve for this. It's not literally to make the planes fly faster. But to be able to draw upon the isolated insights within particular sciences and to use those technologies to create new virtuous loops. So instead of asking "why are we not coming up with new insights" we should be focusing on "how can we connect these disparate insights from multiple sciences to actually do something."
To do this we need to focus on two parts:
Focus more on connecting the pieces of knowledge amongst the various advances that's happening now - ideally focus on getting at least one technology to undergo an explosive supply glut
Focus on creating 'talent clusters' - places where talent can mix in an interdisciplinary fashion and see if they can't find ways to use what emerges from Point 1, and create a virtuous feedback loop
We can already see the initial shoots of this working. The mRNA treatments we’re so awed by was the result of a decade+ search that technology finally caught up with to make possible, atop an even older theoretical basis. Much of the advances in batteries or solar cells also find their roots in computation and material science taking advantage of multiple modern tools. Step changing in farming efficiency through verticalisation takes its cue from AI and robotics. VTOLs are being built as we speak, as is supersonic aircrafts, with new materials and the aid of computational simulations.
So what we need to do is more tactical. Help folks take risks in trying something new (fund people not projects maybe from Nature, and from Nintil). It's by chipping away at the barnacles that we'll smooth that transition. We don't need more Manhattan Projects for energy and chemistry. We shouldn't be focused on silver bullets of various general purpose technologies to lift us up. Rather, we need more people to try, research and experiment. That's what we should be encouraging.
It seems odd to end an essay about why we're in a Great Stagnation with the thesis that it's all going to be okay. But that's where I find myself. We need to still do the hard yards of finding places where technologies can intersect and create feedback loops, and we are! Nanomaterials find usage in medicine and industry. Computing finds application in biotechnology. Material science is integral to sending up rockets. These are syntheses that are taking place without direct intervention or guidance. Let's try and encourage more of this. If nothing else it'll be a Hansonian way of looking for that next fertile valley.
Update - a tl;dr summary from discussing the article and its implications.
Overall growth and progress (especially economic growth) comes from multiple fields simultaneously progressing, integrated with each other in positive feedback loops (like steel and railroad)
This is usually led by genius-clusters who can take advantage of the confluence and make cool things happen
For this to happen though, the individual logistic curves in each industry has to be sufficiently advanced, and they have to reach a maturity where they can reach out and overlap with other S curves
It's like groves of trees growing together to form a broader canopy - groups of trees are the industries, and canopy is the overall economic growth levelling up
Sometimes it takes a while for all trees to reach the same height
Sometimes overgrowth in a few trees makes the ground more fertile, and speed up the growth of other trees
But until there's a full canopy the level set won't be complete
The speed of each individual logistic curve growth isn't fixed but highly variable, and pace of breakthroughs aren't predictable
The rate of discovering new S-curves is dependent on number of recombinations of existing S-curves
As time goes on, the number of industries where we need such progress to happen is also changing
It increases as the number of industries increase and more things are specialised
It decreases as some technologies collapse together to become one (electromagnetism, neuroscience)
Our overarching approach to help increase the rate of progress therefore has to revolve around the identification and creation of more clusters and enabling better cross-pollination of ideas between sciences