Introduction

Technology is an incredibly important part of the modern world. In fact, it is arguably the single greatest factor in shaping the world we live in. Especially when we consider the magnitude of its impacts over the last ~500 years¹:

But for all its importance, I would argue we tend to have very little collective knowledge about it. This often sounds strange to people who rightly point out that we know a lot about a lot of technologies. This is true at the individual level. We have an incredible knowledge of the intricate workings of many individual technologies. But our generalised understanding of what technology is and how it evolves is considerably less developed. Consider the following question: what is technology? Do you have a good definition? 

Most people don’t and neither did I until I read Brian Arthur’s excellent book — The Nature of Technology. In it he attempts to articulate a generalised theory of technology, answering the following questions:

1. What is technology?

2. How do novel technologies evolve?

He presents a compelling case which I will explore below.

What is technology

Technology pervades every aspect of our lives. From the houses we live in to the cars we drive, the roads we drive on, and the offices we work in. It is quite hard to imagine escaping technology for even a second. Imagine you went into the forest somewhere, alone and without any devices. Have you escaped technology? Not if you are still wearing clothes — the primitive technology used to shield our bodies from the weather. The realisation of just how pervasive technology is makes the question all the more important - what exactly is technology?

Brian argues that technology is ‘a means to fulfil a purpose’. Thinking through that - do our examples above fit nicely within this framing? I would argue yes. You don’t need to do too much work to realise that a car, a road, a building or even clothes present a means to fulfil a purpose. However this framing is quite broad and doesn’t tell us much about how to think about technology — where it comes from, how it evolves and the limits of its reach. I have often heard people say that the CCP’s governance system for China is a technology. This definition would support that but it doesn’t give me much else to work with.

In thinking about properties of technology, the book argues that there are broadly 3 properties shared by all technology:

1. Combination: All technologies are fresh combinations of what already exists

2. Recursion: Technologies are built from sub-components assembled together, which are themselves built from sub-components, all the way down to their elemental base. In this was each technology is recursive

3. Phenomena: All technologies leverage some phenomenon or natural regularity

Combination

Brian argues for what he calls combinatorial evolution — the process by which early technologies form using existing primitive technologies as components. These new technologies subsequently become building blocks for further new technologies. As a result, he argues, over time there is an ever greater supply of base components from which new technologies can form. In this sense technology creates itself.

Recursion

Technologies are also recursive in that their components are organised into a central assembly, with sub-assemblies supporting the primary function. Each of the sub-assemblies itself is organised in the same way. This continues down to the elemental base. Brian gives the example of the aircraft fuselage, which is a sub-assembly of the F-35 jet, which is itself a sub-assembly of the broader technology — the squadron of F-35s. A squadron of F-35s in turn is a sub-assembly of an aircraft carrier, and so on. In the opposite direction, you could follow the fuselage down further and further, examining each sub-component which is at some point the primary technology needed to solve a problem, but in this context is merely a part of the whole. In this way, technology has no ‘characteristic’ level — all technologies are available to become a sub-component of a higher-level technology in the future.

Phenomenon

All technologies are born from phenomena. As we examine any technology, we see that its core purpose is to harness some underlying phenomena or natural regularity in order to achieve a means. Radar and MRI harness the reflection of magnetic waves and nuclear magnetic resonance. Oil refining is based on the phenomena that different components or fractions of vaporised crude oil condense at different temperatures. Rocketry is based on a number of phenomena, including that producing a large amount of heat in a given direction generates thrust in the opposite direction. Early humans leveraged phenomena in the natural habitat — the sharpness of obsidian and the momentum of stones in motion.

Once we realise this to be true, the value of basic research becomes more apparent. Investment in discovering and unlocking new knowledge in the form of harnessable phenomena is clearly beneficial. Every new phenomena potentially unlocks multiple new technologies, each of which has the potential to generate reinforcing demand (which we will discuss further below). Every new phenomena or regularity is added to the ever-growing pool from which inventors and entrepreneurs can pull as they attempt to link what is known to what is needed by society.

Joel Mokyr makes a similar point in his excellent book A Culture of Growth when speaking to why the steam engine could not have been created in China during the 18th century, despite being similarly advanced in many relevant areas of their society. The answer lies in the existence of an epistemic base from which a would-be inventor needed to pull from. The European culture which promoted the conjecture and criticism of ideas led to the advancement of many scientific theories. One of which was Evangelista Torricelli — a student of Galileo — who first proposed that there was an atmosphere surrounding the earth. Thomas Newcomen's early steam engine was an atmospheric engine which condensed steam drawn into the engine cylinder, creating a partial vacuum which allowed the atmospheric pressure to push the piston. Without knowledge of the atmosphere, which was highly unlikely to have existed in a society that demonised the criticism of long-held beliefs, it is almost unthinkable that one could come to hold the knowledge required to build such a machine.

Arthur points out that one implication of this is that if we were to take our technologies to a location where the underlying natural regularities were different, they would need to be rethought. A simple example is space, where the absence of one of the most common and well-known regularities — gravity — requires an alternative method for doing almost everything, including a task as simple as drinking water.

So where does this all lead us? Arthur summarises that it suggests a new and improved definition of technology: technology is one or more phenomena captured and put to use. This is a description of technology at its most intrinsic level, since what makes a technology work is the core principle upon which it is built.

Where do ‘social technologies’ fit in

One of the questions I had when reading this book was: where do things like monetary systems, legal systems, and systems of government fit in. As you might have already realised, these all fit the initial broader definition of ‘a means to fulfil a purpose’. As we have just discovered however, a more complete definition refers to the concept of harnessing some phenomenon. The conclusion Arthur draws is that all of these things are technologies, but the underlying phenomena they harness is behavioural rather than physical. The monetary system leverages the social phenomenon that we value things that are scarce (e.g., gold) and that we trust a system when we believe other people trust the system (e.g., fiat money). These often feel like less of a technology than something leveraging physical regularities, but that is more to do with the concrete nature of physical phenomenon vs the abstract nature of behavioural principles.

What does this all mean?

Perhaps this feels obvious but all of this tells us that technologies are invented. They are brought into reality by pulling together sub-components from within a domain, and doing so recursively to solve every problem that is met along the path to completion. At each level some phenomenon is leveraged in order to solve the problems required to will the new technology into existence. Arthur presents the analogy of a programming language, where each individual technology is to the domain of its origin as a computer program is to its language. In this sense the inventor can be said to instantiate their new technology from within the domain that they are working. And when they’re done, that technology becomes yet another sub-component within the domain, another primitive that can be leveraged by the next inventor looking to instantiate their idea into existence. 

How do novel technologies evolve; what is innovation

This understanding of technology allows us to investigate the question of how novel technologies evolve with some more structure than is typical. We have a shared language and understanding of technology from which to reason. From Arthur:

Innovations in history may often be improvements in a given technology—a better way to architect domes, a more efficient steam engine. But the significant ones are new domainings. They are the expressing of a given purpose in a different set of components, as when the provision of power changed from being expressed in waterwheel technology to being expressed in steam technology.

What Arthur is arguing is that there are two forms of technological ‘innovation’ that are possible:

1. Improvements in the use of a given phenomena. We can think of this as a more efficient means to fulfil a purpose. Or, using our more refined definition: an improvement in the efficiency with which we harness and put to use a given phenomena

2. ‘Re-domaining’ of a given purpose: We can think of this as a new means to fulfil a purpose, or the harness and use of new phenomena to achieve our purpose

His argument is that the second type — re-domaining — is the true form of innovation. Again, from Arthur:

Consider: In the 1970s computer printing was carried out by line-printers, essentially an electronic typing machine with a set of fixed characters. With the coming of the laser printer, computers printed by directing a laser to “paint” text on a xerographic drum, a different principle. In the 1920s, aircraft were powered by a piston-and-propeller arrangement. With the coming of the turbojet, they were powered by gas turbine engines using reactive thrust, a different principle. In the 1940s, arithmetic calculation was carried out by electromechanical means. With the coming of the computer, it was accomplished by electronic relay circuits, a different principle. In all these cases a new technology came into being—the laser printer, the turbojet, the computer—from a new or different base principle. A change in principle then separates out invention—the process by which radically novel technologies arise—from standard engineering. It also allows us to draw crucial distinctions between mere improvement and real origination. We can say that the Boeing 747 is a development of the 707 technology, not an invention. It improves an existing technology but uses no overall new principle. And we can say that Watt’s steam engine is an improvement of Newcomen’s. It provides for a new component—a separate condenser—but uses no new principle.

This is by no means consensus. Edmund Phelps’ view, expressed in his book Mass Flourishing, is that inventions like Newcomen’s steam engine are overrated. He argues that the constant improvements — like that of Watt’s much improved steam engine — generate the true value of technological innovation. While it seems true that most innovation is type 1 (improvements to an existing phenomena), the truly transformative innovations largely appear to be type 2. If we think about some of the most significant innovations in the last 500 years — cars, planes, new forms of energy, and computing — each of them was created as a new means to fulfil a need, and each was instantiated from a new domain, using new phenomena. Further, without new phenomena to harness, any form of mass-flourishing-style incrementalism would eventually reach diminishing returns. Put another way, the existing phenomena are a fixed factor in the economy, and like any fixed factor will eventually reach diminishing returns unless we continue to build upon it with more discovery.

One potential challenge I had with Arthur's type 1 vs type 2 definition of innovation was where innovation involving software would sit. At first it might feel like all software is one domain, leveraging a consistent set of phenomena. This implies that no ‘true’ innovation happens in the field of software, just improvement. However, I think the way to think about this is that true innovation occurs when a given purpose can be re-domained to the world of software. This again feels consistent with reality. Most of the transformative power of software has been in which problems we can begin to solve with computers, essentially leveraging a different set of phenomena (those used in computing) to achieve many of the things we did more manually. If we think about the tech giants in our society, each of them solves a problem that was previously done without computers — searching for information, commerce, socialising, and communicating. While they were not always the first to conceive of using software to solve their respective problems, they were the eventual winners by virtue of having some mix of the best execution and the right timing.

How technology begets more technology

Arthur is a techno-optimist, believing that the ever-growing pool of technologies provides a greater variety of sub-components from which the next generation of inventors can mix-and-match to bring their ideas to fruition. He also believes that there are other mechanisms by which technology is self-reinforcing. In total, he suggests 4 methods by which this happens:

1. Growing the pool of sub-components: as discussed above

2. Creating new demand niches: every time we create a new technology, there is the potential for new needs to arise. Once we discover how to diagnose diabetes, the need for insulin arises. 

3. Servicing the technology: each new technology sets up an opportunity to support it, whether through manufacturing, distributing, or maintaining it. Arthu gives the example of the automobile, which created many ancillary needs: assembly line manufacturing, paved roads, refined gasoline. Gasoline in turn leads to refineries, importing crude oil, and oil exploration technology. We see again here a hint of the recursiveness inherent in technology

4. Solving the problems inherent in technology: technology often has unintended consequences or unlocks one door which leads to 3 more that we need to invent our way through. In the early 1700s Britain was in need of more coal, so they started digging deeper mines. The ability to mine further underground was itself a technology. It created a problem — at these depths the mines were regularly flooded with water. This problem led to the invention of Newcomen’s steam engine, which was primarily used to pump water out of mines, helping to expand the coal industry

There are many more examples for each of these mechanisms, but the important takeaway is this: technological innovation is self-perpetuating.

Summary

What are the implications of Arthur’s theory? The primary one that springs to mind is to do with phenomena. If phenomena are the building blocks of all technology, then we would expect a burst of new inventions some period after a new field of phenomenon is discovered. If this is true, we might expect there to be an early cluster of inventions leveraging the newly discovered regularities. Perhaps this explains the burst of invention in the 18th century, as early scientific pioneers like Galileo and Newton observed and unlocked the first wave of phenomena which could be leveraged by inventors. 

What then, have we seen from the discovery of the phenomena underpinning computation? Many argue that computers have done little in the way of innovation outside of communication technology. What appears promising to me here is the recent developments in fields like biology, with DeepMind’s AlphaFold. If we believe that true innovation comes through ‘re-domaining’ an existing need, perhaps the issue was not inherent in computers, but the productive use cases merely lagged their invention. What if we simply needed to reach a critical threshold of sophistication — through inventions such as neural nets and the advancement of Moore’s Law — in order for computers to begin to ‘re-domain’ existing needs. Today, computers are beginning to play a larger role in science: predicting the structure of proteins, accelerating clinical trials through simulation, enabling scientists to see and track individual cells, and allowing us to control devices with our mind. I am hopeful that we have reached a threshold of computing power, adoption, and sufficient tooling to unlock the next generation of scientific discovery.

---

(1) I understand that growth here could also be driven by more productive social practices such as increased trade, but even those are largely enabled by the development of technology such as ships