Hyperspecialization, Complexity and Progress

Do you remember those old times when it required only a “webmaster” to launch a website? This seems to be so far away and yet it was only 15 years ago. He was involved in design, integration, front and back-end development, server administration, communication, marketing and sometimes even content writing. Now I can give you a hundred different jobs for these roles. If we take a look at developers for example, we need to distinguish mobile developers (+ iOS / Android / Windows / Xamarin) from website developers, distinguish between front-end and back-end, themselves divided among a wide range of available languages. One year it’s the PHP, then the Ruby, then the Javascript, then all possible versions of JS frameworks. The specialization of the developer’s job is endless. There is something convenient about specializations : creating sub-universes of specialities reduce the tensions stemming from too many competitors. Everyone can thus become the best in his hyper-specialized field, everyone wins a medal. Pierre Bourdieu had already noted this in 1987.

As soon as the jurists have gained their autonomy from the princes, each has begun to divide the speciality so as to be the first in his village rather than to the second in Rome. Pierre Bourdieu, 1987

In this article we will apply the theory of decreasing marginal returns to new technologies in order to understand the relation between hyperspecialization, complexity and progress. This theory was developed by Adam Smith and David Ricardo among others. The law can be stated as follows: when increasing the quantity used by a factor, beyond a certain level, production increases less and less.

Let us clarify this notion of diminishing marginal returns with the example of natural resources. In England at the end of the Middle Ages, lands were deforested to provide fuel and agricultural space to a growing population. But wood was no longer sufficient for heating, cooking, and supply the demanding industry; A dependence on coal began gradually. Coal was an unattractive solution because it was dirty, polluting, limited in its geographical distribution, and much harder to obtain than wood. Thus coal mining in underground mines was more expensive than obtaining an equivalent amount of wood in calorific value. This is called decreasing marginal returns. We first gather the low-hanging fruit, before collecting — by necessity — those at the top of the tree. The effort required is therefore more important to obtain the same amount of extra fruit. We can identify this scheme in a wide range of areas. For example :

  • Hardware: making a quantum computer will cost much more than the first computer for a lesser impact on society than computing.
  • Science: Recent discoveries are much more expensive (NASA, LHC budget) than the research costs of scientists like Newton, Galileo, Einstein who brought us much more knowledge. (Or at least that had a greater impact individually)
  • Medicine: nano-robots will cost much more than penicillin to save the same number of lives.
  • Agriculture: the increasing intensity of the agricultural use of land is caused by an investment in labour which is very disproportionate to the yields obtained.

This can also be seen in new technologies. All the heart of Google was developed by the 2 founders, now Google has 60,000 employees. It’s not because they doubled their number of employees that Google gave us twice better results. Not to mentions Google spends astronomical sums, for lower returns per dollar invested than before. Hence, the value added for us is marginally decreasing in relation to the number of employees. Warning, there is no value judgment, I do not recommend Google to remain at a staff of 2. This is not something negative, it is simply a universal law that exists and has consequences.

To remain a leader, Google must diversify, fine-tune its algorithm, manage its many employees, integrate a bureaucracy, in short, become more complex. Once the company reaches a certain complexity, it needs specialized people because the nature and diversity of tasks have become too difficult to be dealt with by a small group of people. This is how the bureaucracy is created and how we go from a startup of a dozen people to firms of thousands of employees. This complexification requires increased specialization of employees in each of the company’s areas.

Fig 1. Technology and marginal productivity

I computed the productivity (in M $ per employee) discounted relative to the inflation of 3 big American company: Google, Amazon, Apple. Productivity is given by dividing the revenue of these companies by their number of employees, the data can easily be found on Statistica. It is clear that starting in 2010–2012, productivity began to stagnate and then to decline. They all have thus reached the stage of decreasing marginal productivity.

Fig 2. The S-Shape Curve Theory of Progress

The only way to break this dynamic is technical progress. Innovation makes it possible to change paradigm, to break the curve of diminishing marginal returns and to regain positive marginal productivity. 
As can be seen in FIG. 2, the S-Shape curve represents an exponential production gain following a new technology. You can see a plateau at the end of both curves, it is where the law of diminishing marginals returns is set up and the production no longer increases much. You have to wait for a new technology before you can see the production take off again. We are therefore in an endless race to progress. Yet, as we have seen before, things “simple” are done first, so — and given the current saturation of the world — there are only complicated problems left. Technological leaps will then become increasingly difficult to achieve.

The division of labour has a very distant origin and appears as early as the Neolithic with the development of agriculture and livestock. The more one divides the global work into a multitude of small specialized task the more one can hope for productivity gains. However, this productivity is subject to the law of decreasing marginal returns, so that new technologies have become necessary to sustain our economic growth. Given the complexity of the problems to be solved, hyperspecialization has become the preferred way of innovating.

In the next articles, we will see how hyperspecialization coupled with artificial intelligence could create a society in decline. We will also see how a cross-fertilization of domains of knowledge could be a solution to apprehend this phenomenon

If you liked this, click the💚 below so other people will see this here on Medium.

Ulric Musset