Engineering romantics fall in love with the work of Jorge Luis Borges early in their careers. Long after Douglas Hofstadter is forgotten for his own work in AI (which seems dated today), he will be remembered with gratitude for introducing Borges to generations of technologists.
Borges once wrote:
“I should define the baroque as that style which deliberately exhausts (or tries to exhaust) all its own possibilities and which borders on its own parody…I would say that the final stage of all styles is baroque when that style only too obviously exhibits or overdoes its own tricks.”
The baroque in Borges’ sense is self-consciously humorous. Borges’ own work in this sense is a baroque exploration of the processes of thought. As one critic (see the footnote on this page) noted, Borges writings “serve to dramatize the process of thought in the apprehension of truth.”
Unlike art, complex and mature technology (not all technology) is baroque without being self-conscious. At best there is a collective sensibility informing its design that can be called a baroque unconscious.
This post is a sequel of sorts to The Gollum Effect. You can read it stand-alone, but you will probably get more out of it if you read that first. Within the Lord of the Rings metaphor I developed in that post, “baroque unconscious” is basically my answer to the question, if extreme consumers are Gollums, who is Sauron?
This idea of a baroque unconscious helps clarify things about the phenomenon of technological refinement that have been bothering me for a while. In particular, it helps distinguish among three kinds of refinement in technological artifacts: refinement that is useful to the user, refinement (often exploitative) that is useful to somebody besides the user, and refinement that benefits nobody at all.
It is this last characteristic that interests me. Refinement that benefits nobody — anything that attracts the adjective overwrought — is what I attribute to the workings of the baroque unconscious. And I write this fully aware of the irony that this kind of post, might be viewed as overwrought analysis by some.
Interestingly though, viewed from this perspective, the other two kinds of apparently intentional refinement can be seen as opportunistic exploitation. They arise through manipulation of those elements of the workings of the baroque unconscious that happen to be consciously recognized.
In other words, I am arguing that the collective unconscious component in the evolution of technology is primary. The conscious component is peripheral.
Or to borrow another idea from art, it is technology for technology’s sake. And unlike in art, there is no primary artist.
The Baroque in Art
There is no such thing as the baroque unconscious in art.
When art exhausts its own possibilities unintentionally we generally characterize it as camp (what Susan Sontag aptly called “failed seriousness” in Notes on Camp). The baroque element in the work is evident to observers, even if the creator lacks the self-awareness to recognize it.
When art exhausts its own possibilities as a side-effect, while pursuing other objectives, we do not call it baroque. We call it either cynical or tasteless. The auteur theory of art applies well enough that if we cannot reasonably impute baroque intentions to the artist, we feel safe assuming that artist was aware of the baroque consequences of his/her decisions. Michael Bay’s Transformers movies (especially the last installment) are examples. They are both tasteless and cynical, but they are not campy or baroque.
Technology is generally more complex and collaborative than even the most collaborative kinds of art, such as movies. The process can create things that exhaust certain possibilities, with no single creator or observer being fully conscious of it. Yet, we cannot call such things campy, cynical or tasteless.
To understand this, suspend for a moment your default idea of what it means for something to be baroque. You are probably thinking of European architecture of a certain period with an exaggerated and visible sort of drama on the surface. That prototypical idea of the baroque is what we tend to apply, in unreconstructed form, to technology: clunky user interfaces and a degree of featuritis that has us groaning.
This is a narrow sense of the baroque. The original architectural instances served a specific function: to impress and intimidate commoners with a display of awe-inspiring grandeur (some art historians have argued that the original examples of baroque were therefore not baroque at all, but cynical). The exhaustion of possibilities in that kind of baroque is all on the surface.
But things can be baroque without being visibly so, depending on the audience for the original function. The key is that the governing aesthetic must seek to self-consciously exhaust its own possibilities.
Invisible, but still intentional baroque is particularly common in modern American pop culture. Most viewers of The Simpsons for instance, miss the bulk of the hidden pop-culture references in the show. A loyal subculture of fans devotedly mines these references and discusses them online. While this sort of thing is often cynical (deliberate creation of baroque plots to create addiction, as in the show Lost), in the case of The Simpsons, I suspect the writers genuinely seek to exhaust the possibilities of the artistic technique of reference, without annoying the mainstream audience.
The Baroque in Technology
In technology, Apple’s products border on the baroque in their exaggerated simplicity. Once the iPad achieves the edge-to-edge display and maximal technically feasible thinness for instance, it is hard to imagine how one would parody it — there is no room left for exaggeration in the physical form at least. Certain possibilities will have been exhausted.
This sort of intentional (and therefore artistic) baroque in technology, however, is not really what interests me. What fascinates me is technology that grows baroque without anyone consciously intending to exhaust any design possibilities. Social forces, such as the competitive pressures of an arms race, or the demands of extreme lead customers, don’t seem to be sufficient explanations.
Art is usually the outcome of a singular vision. But technology, even the auteur form of technology practiced by Steve Jobs, is deeply collectivist. Engineering real things is far too hard for one mind to impose a singular vision on all but the simplest of products. When a piece of technology appears to be the work of a single mind and possesses the dense layers of coherent complexity that can only be the product of a large team, it is evidence of a deep coherence in the team itself. In such a team, individuals trust the collective to the point that they feel comfortable narrowing their domain of conscious concern to their own work.
The baroque sensibility resides in the collective unconscious of the team that produces it. The baroque in the whole is greater than the sum of the baroque accounted for by the self-awareness of the many individuals.
Moderately obsessive-compulsive attention to detail at the level of individuals oblivious to larger purposes, eventually turns into baroque exhaustion of possibilities at the level of the whole product.
This brings us to the idea of refinement, and the question of when, why and how wrought keels over into overwrought.
Refinement and the Baroque
When I first started thinking about refinement, in the context of addictive consumption (as in, refined cocaine), I had examples such as American fast food in mind: precisely engineered concoctions of key refined substances (salt, sugar and fat) designed to cause addictive over-consumption.
The pathologies of consumerism can be traced to an entire universe of such refined goods. I offered the term gollumized to describe humans who end up being entirely defined by a pattern of such consumptive behavior, much like the character of Gollum in the Lord of the Rings, with his addictive, enslaving attachment to the One Ring: a highly refined, pure essence.
Something bothered me however, about the implicit equation of refinement with pathological addictive dependence on the one hand, and cynical exploitation on the other.
The refinement in the construction of something like the space shuttle does not seem pathological. It seems necessary.
A highly refined kitchen knife that plays a role in your creative self-expression as a chef seems somehow different from a McDonald’s hamburger or an expensive wine, both of which are consumption-addiction refined in their own ways.
Even with hamburgers, while acknowledging that they are effectively exploitative and addictive foods designed to enrich the food industry by ruining the health of consumers, it is clearly farfetched to believe that there is some vast conspiracy that includes every biochemist.
The idea that the creation and sale of such foods is more a matter of cynical opportunism is more reasonable. You could accuse the industry of carefully engineering high-fructose corn syrup as a way to make money off corn surpluses, but the industry didn’t create the necessary biochemistry knowledge or surplus-creating agricultural advances with the idea of eventually selling cheap and addictive burgers (for one thing, the evolutionary processes took longer than the lifetime of any individual involved in the story). You could say that the existence of HFCS is 10% intentional and 90% a consequence of the baroque unconscious driving food technology.
In other words, the existence of a Gollum does not imply the existence of a Gollumizer. Sauron in the The Lord of the Rings is at best a personification of the baroque unconscious (with Saruman being one of the cynical exploiters — an HFSC creator so to speak).
But let’s figure out what refinement in technology really means. Consider the following senses of the word refinement:
- Refinement as in purity or purification of substances: ore, oil, drugs, foods
- Refinement in the sense of highly developed and cultivated sensibilities, as in refined palate
- Refinement in the sense of elaborate sophistication of mature or declining cultures
- Refinement in the sense of detailed, attentive design in advanced technologies
- Refinement in the sense of an Apple product (or any other possibility-exhausting product aesthetic)
How do these different senses of the idea of refinement relate to each other and to the baroque? What distinguishes the space shuttle, quality kitchen knife from an iPad, an expensive wine, or a McDonald’s hamburger?
The Sword, the Nail and the Machine Gun
I found a key clue when Greg Rader decided (to my slight discomfort) to overload this sense of refinement with an economic meaning in his 2×2 model of types of economies.
In Greg’s model, the economic role of refinement is to make it easy to value artifacts in an impersonal way, in a cash economy. Unrefined artifacts get you attention or help build social capital in relationships. Refined artifacts help you earn money or participate in the gift economy.
But why should refinement lead to easier valuation and thence to exchange for money.
The crucial missing piece is the role of interchangeability in mass production. As Joseph Ellis writes in The Social History of the Machine Gun:
It was always theoretically possible to conceive of a gun that would spew out vast numbers of bullets or whatever in a short period of time…manufacturing techniques [were not] sufficiently well-advanced to allow individual craftsmen to work to the fractional tolerances demanded for every part of such a complex gun.
The key point here is often lost in discussions of industrialization that use Adam Smith’s simple example of a nail to highlight the division of labor aspect of industrial production. Nail manufacture illustrates the reductionist capacities of industrialization, but it is the integration capacity of industrialization that drives refinement.
The machine gun illustrates the dynamics of integration. It is a complex machine, and as such, liable to break down more easily. Reliability involves network effects within a complex artifact. Roughly speaking, in a design with no redundancy, the more parts you have, and the more complex and fast-moving the linkages among them, the less reliable the machine.
Unless you find an opposed network effect that can scale at least as fast, machines will get less reliable as they scale.
The opposed network effect that was discovered late in the industrial revolution was interchangeability. Interchangeability creates a network effect between artifacts. Crucially, they need not be functionally similar. They only need share a structural language. A machine gun can be cannibalized to repair a telescope for instance.
The significance of Ellis’ point about fractional tolerances has to do with replacement and cannibalization. Craftsmen are capable of very refined work, but the work tends to be unique. It involves fitting this hilt on this sword with great precision. You can get away with this because craft also tends to involve fewer parts, static linkages and performance regimes where breakdowns are infrequent.
With interchangeability comes the possibility of easy valuation, since it is possible to talk of supply and demand at the level of many non-unique parts that can be compared to each other. That helps connect the dots to Greg’s economic hypotheses.
But we still haven’t fingerprinted the essence of refinement itself.
Replacement and Repair
The first key threshold crossed on the road to industrialization was the replacement of human, animal and uncontrolled inanimate power (wind or water) with controlled inanimate power: coal and oil. Much of the attention in attempts to characterize industrialization is given over to the study of this threshold-crossing.
The second key threshold crossed was the shift from repair to replacement. When breakdowns became frequent enough that anticipatory manufacture of replacement parts became cheaper than reactive repair or replacement, the network effects of industrialization truly kicked in.
The network effects of reliability in a sword are not strong enough that you need to counteract them with interchangeability effects. In fact, much of the complexity in a sword may well be in baroque artistic elements that serve no purpose (a sword that loses a diamond from its hilt is still equally effective on the battlefield).
Even early industrial-age artifacts, do not have enough complexity and speed to really require interchangeability. This is one reason I find elaborate steampunk fantasies fundamentally uninteresting. They involve imagined machines that come across as laughably Rube Goldberg-esque precisely because they don’t comprehend reliability problems, and the methods actually created during the industrial age to mitigate them.
When you get to something like a machine gun though, where breakdowns are frequent and waiting for custom replacement parts is hugely expensive, you must meet absolute tolerances, so that any replacement part can replace any broken part (and equally crucially, so that two broken, complex assemblies can be cannibalized to produce at least one working assembly).
So we can conclude that:
- Refinement in craft based on relative tolerances leads to uniqueness.
- Refinement in manufacturing based on absolute tolerances leads to interchangeability.
From these two basic kinds of refinement, we get the five connotations of the word I listed earlier. This happens via the appearance of a refinement surplus.
The Refinement Surplus
Interchangeable parts based on absolute tolerances solve the reliability problem and then some. The network effects of interchangeability turn out to be stronger than the network effect of increasing unreliability in individual complex artifacts.
What’s more, since interchangeability limits the need for communication among collaborating makers, refinement of component technologies can progress much faster (as Adam Smith noted). This is what we call “specialization.” It happened in physical engineering before object-oriented programing ported the idea to software engineering.
You could say that work previously achieved by communication among makers is now achieved via communication among artifacts. This is most obvious with software objects, but the core idea is present even when you shift from a custom-made nut-bolt pair to a standardized pair that “communicates” via numerical absolute tolerances.
So interchangeability creates a social network of (say) machine guns. There are functional linkages within complex artifacts that make them useful, and substitution and reuse linkages between them that make them reliable (redundancy inside an artifact is merely a semantic distinction: think of it as carrying interchangeable spare parts inside the boundary of the artifact, with the capacity to automatically switch out broken parts). Interchangeability and standardization make every machine gun less unique, and more a part of a sort of hive-machine-gun beast.
Dramatic as this effect is, it pales in comparison to the effect of commonalities across the needs of different types of complex systems. This connects all complex artifacts into a giant social network. The One Machine.
A high-tolerance part can serve a low-tolerance function, but not vice versa. Economies of scale then kick in and dictate that many components become more refined than they need to be, for typical artifacts that make use of them. The result is that systems gradually get more refined than they functionally need to based on immediate intentions. The needs of a few artifacts drive the refinement levels in all technologies.
This creates a refinement surplus. Industrial technology, unlike craft work, runs a continuous refinement surplus. The surplus was initially triggered by the need for interchangeability to solve the reliability problem, but that turned out to be a case of using a sledgehammer to kill a fly.
Or so it might seem if you only look at individual artifacts. I’ll argue in a future post that once software and the Internet kick in, reliability problems can once again overtake what interchangeability can mitigate. As the One Machine gets increasingly interconnected, the unreliability network effect may overtake the interchangeability network effect, hence the fundamental Singularity-vs.-Collapse debate.
The possibilities represented by limiting refinement levels are always greater than the universe of artifacts in existence at any given time.
Exploitation of this refinement surplus is fundamentally what creates the predictable “growth” in industrial age Schumpeterian creative destruction. But it isn’t the intent to exploit that drives the evolution. It is a collective unconscious drive to exhaust possibilities and find limits, independent of any specific need.
The Platonic Baroque
The Lord of the Rings captures artistic anxieties about engineering: the “good” races create beautiful craft, the “evil” ones engineer ugly things.
Where LOTR goes wrong is in focusing on beauty in craft as the distinguishing factor (there is a line in The Hobbit which goes something like “the Goblins create many clever things, but few beautiful ones”).
In LOTR, evil engineering artifacts are crude, unrefined and possess little symmetry. Good ones made with craft are intricate, refined and highly symmetric.
This is obviously the exact opposite of what actually happens.
Open up a laptop and compare what you see to (say) a beautiful hand-crafted necklace. Not only is the inside of the laptop more intricate than the necklace, it is more intricate than you can even see. You would need electron microscopes to get a sense of how unbelievably intricate, refined and symmetric a laptop is.
The technological landscape is defined by two kinds of beauty. On the one hand, you have the possibility-exhausting conscious baroque artifacts that we view as “pushing the envelope.” Both the iPad and the space shuttle belong on this end of the spectrum. One contains chips at the limit of fabrication technology, the other contains materials that can handle enormous heat and cold, produce unimaginable levels of thrust, and so on.
On the other hand you have things that are not at the edge of technological capability, but manufactured out of component and process technologies created for those leading edge technologies. And I don’t just mean obviously over-engineered things like space pens that write upside down (which you can buy at NASA museums). I mean everything. Regular Bics included.
In this category, makers strive to exhaust the possibilities, but always lag behind. The surplus refinement potential shows up in the unnecessarily clean lines of modernism. Unused bits. Unbroken symmetries. Blank engineering canvases that expand faster than designers and technicians can paint.
The interaction of the two kinds of beauty is what creates the texture of the modern technological landscape. I call it platonic baroque. This may seem like a contradiction in terms, but bear with me for a moment.
The baroque unconscious is the force that drives technological evolution: a force whose potential increases faster than it can be exploited.
Recall that the baroque seeks to exhaust its own possibilities. It is a technical exercise in exploring process limits, not an exercise in expressing ideas or creating utility. But this process needs ideas to fuel it.
In the days when royalty and religion loomed large in the minds of creators, it was natural to exhaust possibilities by filling them up with the content of the mythology associated with the power and money that drove their work. It was natural to fill up blank walls with gargoyles and cherubs, popes and princes.
But when the power and money come from a force whose main characteristic is vast and featureless potential, the baroque aesthetic seeks to exhaust possibilities by expressing that emptiness with platonic forms.
So the Bauhaus chair is not a rejection of the baroque. The modernist designer merely seeks to build cathedrals to his new master: a vast emptiness of possibility within the refinement surplus. This possibility is the father of industrial invention, a restless, paternalist force that replaces necessity, the mother of craft-like invention.
I am tempted to explore that male/female symbolism further, but I’ll limit myself to one overwrought metaphor. This unexploited possibility that is the father of industrial invention is at once a Dark Lord and engineering Dark Matter.
Maker Addiction and Exponential Technology
Where there is surplus, it will be exploited. Possibility, rather than necessity, drives invention. When ideas for exploitation lag the potential to be exploited you get baroque unconscious design.
Why would somebody build something simply because it is possible?
Both craft and engineering are driven by an addiction to making. It does not matter whether needs or possibilities enable the making. Makers will make. What determines how fast they make is whether they are able to focus on their strengths or whether they are limited by their weaknesses.
This is the shift in maker psychology due to industrialization: from deliberative craft work limited by individual weaknesses, to reactive engineering work that is not limited in this way, thanks to specialization.
Need-driven making requires a focus on function and utility. Non-functional making in craft is easily recognized as artistic embellishment.
The idealized craftsman — and it was usually a he — was a deliberate and mindful creator. He made the whole, and he made the parts. When things broke, he made repairs or crafted new parts. Each whole was unique. When craftspeople collaborated on larger projects — stone-masons making blocks for cathedrals say — assembly itself became a craft that was limited by the skill of the best (if you look at the history of masonry, you can see an obvious and gradual progression from rough-hewn blocks carefully fitted together, to more refined blocks that look increasingly interchangeable in late pre-modern architecture).
In industrial artifacts based on interchangeability, however, the role of craftsman bifurcates into the twin roles of technician and engineer-designer (for now, we can safely conflate engineer and designer). Both are reactive roles where function and utility take a backseat to sheer maker addiction.
The technician reacts to component work defined in terms of absolute tolerances by pushing the boundaries of process capabilities and component quality with addictive urgency. I explored this earlier in my post, The Turpentine Effect (though I didn’t connect the dots until now). The result is Six Sigma, an explosion of process tools, and the dominance of an intrinsic and abstract notion of potential future value over an extrinsic and specific notion of realizable current value. Somebody will use this in the future beats nobody can use this right now. By and large, this trust is justified: increasing demands for refinement from the most demanding applications keep up with the possibilities.
In this process of reactive design, refinement in available components and processes starts to drive refinement levels in complete artifacts that have already been invented, and suggests new inventions. A positive feedback loop is set in motion: increasing component and process refinement overtakes application needs as individual artifacts mature, but then new applications emerge as pace-setters. Design bottlenecks migrate freely across the entire technological landscape, via the coupled technological web, instead of remaining confined within the design space of individual artifacts.
For those of you who are familiar with the S-curve models of technology maturation and disruption, imagine disruption S-curves bleeding across unrelated artifact categories via shared components and processes, creating an overall exponential technology evolution curve of the sort that both Singularity and Collapsonomics watchers like to obsess about, and that I will obsess about in future posts.
Across the fence from the technician, the engineer-designer loses mindfulness by shifting from deliberately dreaming up useful ideas to reacting to the possibilities of available component and process sophistication levels.
A perfect example is Moore’s Law: semiconductor companies began pushing fabrication technology to extremes before applications for the increased capability became clear.
On the other end we have Alan Kay’s reaction to Moore’s Law in the early 70s: the idea that computing should strive to “waste bits” in anticipation of decreasing cost. Computer design shifted from fundamentally deliberative before PARC to fundamentally reactive after.
Effects, Large and Small
So the net effect of maker addiction faced with refinement surplus is that existing artifacts get pulled into a baroque stage of their evolution and new artifacts appear to exploit possibilities rather than respond to necessities. I am not sure this is much better than Gollumizing consumption.
The One Machine gets increasingly integrated, and takes on an eerily coherent appearance due to uniform refinement levels and the operation of the platonic baroque aesthetic at the level of individual artifact design. Design bottlenecks drift around within this technological body politic, making it more coherent, more eerily platonic-baroque over time.
If the creation of unrealized refinement potential ever slows, and exploitation starts to catch up, you can expect the platonic baroque to become less platonic and more visibly overwrought. The blank canvas will start to fill up.
Thanks to this eerie collectively created aesthetic coherence, the One Machine takes on the appearance of subsuming intelligence and intentionality that suggests visions of a Singularity-AI to some. Whether this is a case of anthropomorphic projection onto a smooth facade beneath which unreliability-driven collapse lurks, or whether there is an emerging systemic intelligence to the process, is something I still haven’t made up my mind about. If you’ve been following my writing, you know that at the moment, I lean towards the collapse interpretation. Darwinian evolution as refined complexity created by a blind watchmaker is too much of a precedent to ignore.
At more mundane levels, the baroque unconscious creates a critical shift in the nature of engineering: the pull of under-exploited refinement surplus is so strong that nominally less useful things that exploit the surplus can diffuse far faster, and suck away resources, far faster than nominally more useful things that ignore it.
All you need is a human behavior with potential for escalating addiction. You can then move as fast as the refinement surplus will allow. I explored this idea in The Milo Criterion.
Ignoring this leads to the classic entrepreneurial mistake: attempting to build useful things instead of things that exploit refinement surplus. The most high-impact technologies of the day are almost never whatever the wisdom of the day identifies as the most potentially useful ones. They are the ones that can spread most rapidly through The One Machine, mopping up refinement surplus.
So the best and brightest flock to Facebook or Google, and cancer remains uncured. Again, I am not sure whether this a good thing or not. Perhaps from the perspective of the Dark Lord, optimizing the One Machine, now is simply not the right time to cure cancer. One day perhaps, the design bottlenecks will drift to that corner of the technological Web. Until then, we’ll have to content ourselves with doctors who tweet during surgeries and webcast the proceedings, but still cannot cure cancer.
I’ll stop here for now. This post has been something of a stream of conscious expression of my own baroque-unconscious addicted-maker tendencies. But then, I figure I can allow myself one of these self-indulgent posts every once in a while. Especially since my birthday is coming up in a couple of days.