Machine age musings on algorithmic growth theory

by
Peter Radford
My articles My siteMy booksMy videos
Follow on:

from Peter Radford

Don’t mind me.  I am just thinking out aloud…

That we live in a Machine Age is indisputable.  Our lifestyles depend entirely upon the mediation of machines.  Without them modernity collapses back to whatever existed in the prior ages and we surrender most of what we currently cherish.

And it is important to use the phrase “machine age” because other phrases such as Industrial Age and so on limit us.  Some say we are now entering a Digital Age, but this too is a mistake.  All that is changing is the nature of our machines.  The Machine Age lives on.

So what is a machine?

It is a piece of information that allows us to channel energy and concentrate it such that we amplify the output we expect from that application of energy.  In this case the information is embodied in a plan, design, or layout of a particular physical structure through which we harness the energy.  The nature of that information has changed mightily throughout the Machine Age, ranging from the design of steam engines at one end to the design of supercomputers at the other.  At its core, though, the notion remains the same.  The application of information to the harnessing of energy is the basis of economic activity.  We might even say that we don’t simply harness energy through the use of machines: we also harvest it.  Or, at least, we use that aspect of energy available to do work.  And an economy is simply the totality of work we need to do to provide whatever it is we feel we want — whether those “wants” are essential or luxurious is irrelevant to this discussion.  That’s a moral not a machinery question.

Viewed this way a design is, of course, simply an algorithm we apply to energy to accomplish our goals.  So we might as easily argue that we live in an algorithmic age.  But all designs of all tools are algorithmic.  Using information to structure some physical substrate for the purposes of improving our productivity is a timeless phenomenon in human history.  We need a more specific appellation to divide modern society from prior ages, and, using the phrase Machine Age provides that specificity. 

Economists have been trapped in their old nomenclature and thus unable to renovate their theories to accommodate the change in machinery.  The original machinery question, which arrived in its modern form in chapter 31 of the third edition Ricardo’s “Principles” focused on the social impact of the arrival of modern machinery.  The obvious question concerned the displacement of laborers from the work that was now presumed to be done by machines.  The differentiation between “work” and “labor” was essential to discussions of what became to be known as technological unemployment.  And, a few years later, as the deleterious effects on living standards of the laboring class became obvious other, more political, issues added themselves to the overall Machinery Question.  

These political questions so intruded into the discussions that whenever economists theorized about the impact of machinery and discussed growth they referred to combinations of capital and labor as if they were the inputs to the process of economic work.  Neither is.  Information and energy are.  Labor and capital are political labels used to enable a debate about control of the way in which work is done.  They are not inputs, however, to the actual work being done.  A laborer possesses a variety of attributes necessary to the accomplishment of economic work.  Muscle power was always the dominant throughout the early millennia of human existence.  And that was augmented by various skills learned along the way to amplify the amount of work that the muscle power could produce.  So labor is a melange of inputs, not a single, input.  Using it as if it were homogenous is an error that continues to echo throughout economic theories of growth — notably in the existence of something called ”total factor productivity” which is a vague veil drawn over the ignorance produced by not using the correct input categories to start with.  

A re-creation of economics without the categories of “labor” and “capital” would open up discussion useful to moving our understanding of the future of economic work forward.  Those categories can then be applied in the subsequent political discussions of how the produced of that work is distributed, which is where class based categories, which is what they are, provide the necessary framework for such discussions.

One example of why going down this path is helpful is in the current debate over the apparent slowing of growth.

When we conceive of all machines as algorithms, and when we consequently look at growth as the product of an ever larger set of algorithms being applied to our environment in order to extract ever greater volumes of produce, we can examine the nature of the algorithms themselves as a potential source of a slowdown in growth.  Put differently, we can theorize that algorithms range in their inherent complexity and that those of a more simple nature are those first exploited.  Simple machines doing simple tasks are invented before their more complex successors,  This, in turn, implies that our exploration of the entire set of algorithms — those discovered and those awaiting discovery — moves along a frontier of increasing complexity.  Each step deeper into the set produces greater value because of the denser combination of information and energy available as a result of that greater complexity.  

However, it is the capability or complexity our existing machinery that provides both the tools necessary to explore the set of algorithms more deeply, and which also limits the speed of our progress.  I suggest that the current apparent slow down in discovery, in growth, and in productivity is due to our having arrived, at least temporarily, at a point where the limitations of our contemporary machinery outweigh the exploratory value it has.  Recent advances in the use of supercomputers to delve into protein designs is a good example of this phenomenon.  The enormous cost of assembling the supercomputers needed to push us further into the set of algorithms suggests that, for now, further exploration will be much more difficult than before.  We should therefore expect growth to slow down until we find new ways of building the machines we need to accelerate exploration once again.

This is a simple example of why we need to look at economic work differently.  I would, as an addendum, suggest that it is the role of business to exploit the newly discovered algorithms in order to bring the value within them to the consuming public.  Business does this by standardizing, organizing, and replicating the energy/information combinations of those algorithms.

One last note: the first great stage of the Machine Age was focused on the physical consequences of the application of algorithms.  It was heavily material and dominated by the harvesting and use of energy.  The next great stage will see the emphasis shift towards the other input to economic work.  Information, rather than energy, will be the source of greater value.  In which case the overall material content of growth will diminish in favor of a greater informational or experiential content.  

Unless economists change their understanding of the inputs into economic work — along the lines laid out above — we will underestimate the wealth being created.  We will mis-measure what we are produce simply because we mis-measure what we put into the production of that output.  Measurements and theories are contextual. With the shift in emphasis underway from energy to information as being more important, older measures need updating.