Architecture: Biological Form and Artificial Intelligence.
Nikos A. Salingaros (*) and Kenneth G. Masden II (**)
University of Texas at San Antonio
(*) Department of Mathematics
(**) College of Architecture
A revised version of this paper, with illustrations, is published in The Structurist, No. 45/46 (2006), pages 54-61.
Abstract: This paper utilizes ideas from biological life to examine the architectural condition in which we live. We reveal how modern-day designs have come to operate in exclusion of any connection between the built environment and the primary animating properties of living structure, i.e. (i) organized-complexity, (ii) metabolism, (iii) replication, (iv) adaptation, (v) intervention, (vi) situatedness, and (vii) connectivity. Architecture today for the most part seems empty and lifeless, devoid of the requisite innate information necessary to engage sentient human beings in their everyday lives. Unintelligible in its form and application, architecture as a built form no longer carries with it the power to affect the lived experience. Drawing analogies from living structure and artificial intelligence, we find the promise of a new direction for architecture in the 21st century. Looking to modern robotic science and technology, a strong correlation can be made between biologically-driven functions of living structure and the adaptive processes that once gave form to architecture, and which now serve to steer the Mars Explorer.
Architects today act in one important aspect very differently from most other human beings; and indeed, from other organisms that define the biosphere. In the business of their day, architects set forth ideas that shape the living experience for others, ideas that take form in the world around us. Surprisingly however, many contemporary architects and their work are no longer situated in the world they serve. During their training, an architect's internal representation of the real world is replaced by an abstract, formal set of images. Although this practice pre-dates our own time, it is now exacerbated by the increasing reliance of architects upon computer-based images, not only for design, but more importantly, as a preferred alternative to the physical world. Yet, these images do not fully represent reality, but rather reveal fictitious objects that don't really exist.
Increasingly, architects are distancing their work from the physical world with representations of a false, artificial reality. It is in this process of an internalized vision that contemporary expressions of architecture are rendered. And it is these expressions that now give form to the built environment as manifested in the mind's-eye of the architect. Understanding how our world, body, and mind work together to inform our soul of its existence exposes the problems inherent with such visions.
2. FROM THE NATURAL TO THE UNNATURAL.
Although architecture has embodied a variety of different designs and styles throughout the ages, the most successful buildings and urban environments have an essential commonality with living forms, i.e. material properties and an assembled nature. It is important, however, to distinguish between superficial resemblance, which can lead to dysfunctional and inhuman buildings, and an approach based upon a genuine understanding of life processes.
Historically, building design evolved through the natural occurrences and processes of the earth and the structural principles of the physical world. Imbued with a deep understanding of human needs and activities, traditional methods of design and construction revealed an honest (real) expression of the built environment. As human beings came to master their natural environment, they began to extend their designs beyond the physical limits imposed by form and materials. Seeking to advance their architectonic expressions, master-builders raised their great cathedrals from the earth, reaching higher and higher.
Unable to transcend human existence, yet still innately compelled by the need to overcome the limitations of the materiality of building, the study of architecture began to develop independently of its natural environment. Formalized within the condition of academic studies, architecture soon became the intellectualized property of the University. This set into motion a process that would ultimately render architecture as an artificial and abstract expression of man's disconnected philosophical and ideological ponderings.
Architects today continue to fool themselves into believing that philosophy or ideology can substitute for a cogent understanding of the natural processes of the earth, and the structural principles of the physical world. Even though all great architectural works of the past were derived from some aspect of nature, or perhaps in spite of this, a peculiar choice of philosophy is misused in contemporary architecture to supplant nature. By rejecting natural and human mechanisms, architecture has oriented itself away from essential principles of physical structure, toward an aestheticized and internalized expression of the built environment. This is not meant as a critique of any specific style. It is more of an observation about the processes involved in contemporary architectural thinking.
In an effort to better understand how architecture is fundamentally grounded in the natural world, we need to delve further into biology. Curiously enough, many of the twentieth century's pioneering architects have been strongly influenced by the same properties of living structure that we discuss here. Nevertheless, they only had a cursory understanding of the scientific basis of this body of knowledge. As a result, the built applications did not fully realize the intention. To make things worse, thoughtless imitation of such innovative prototypes reduced these ideas to superficial expressions, which ultimately gave way to one or more fashionable styles.
3. PROPERTIES OF LIVING STRUCTURE.
Living structure is known to satisfy several natural properties such as: organized-complexity (information storage); metabolism (energy use); replication (self-reproduction); adaptation (the organism changes itself to better profit from its environment); intervention (the organism changes its environment); situatedness (embedded in the world through sensors); and connectivity (information processing). In biological entities, all processes usually occur together, but theoretically, these are separate concepts.
Following Christopher Alexander, 1 we associate biological and architectural order with the organization of complexity, which represents the compression of information. An ordered structure has to be complex, yet it is also ordered because it has a large number of correlations that lead to an overall coherence. In architectural examples, correlations arise as visual symmetries and connections, which are easily perceived. 2 Life, whether biological, artificial, or architectural, results from the physical concentration of information. A noncomplex structure, on the other hand, requires little mathematical information to create, leading to simplistic structures without any internal differentiations. The world of rectangular building blocks that characterizes industrial architecture and urbanism is mathematically empty. Many architects perceive a superficial "ordering" in this empty world because of alignment and lack of distracting substructure. Seeking uniformity in this way, however, can be seen as a misreading of the actualities of order.
One particular insight of Stephen Wolfram is illuminating, because it surprisingly links uniformity with randomness. Wolfram points out that uniformity of structure is not simple, but is instead the result of intentionally directed processes:
"But in nature uniformity often seems to be associated with quite complex microscopic behavior. Most often what happens is that on a small scale a system exhibits randomness, but on a larger scale this randomness averages out to leave apparent uniformity ...". 3
Here we have a perceptive statement of how uniformity arises from randomness (i.e., disorganization). The implications for design are significant, since uniformity is thereby linked not to simplicity or order, but to disorganization.
An animal or plant metabolizes in order to maintain its organized-complexity. The complex physical structure is there to permit metabolism, replication, and connectivity. Metabolism is a process by which existing sources of order are absorbed, and disorder is shed, so that the organism maintains its structural organization. In the case of a developing organism, such as an embryo or young animal, the entity metabolizes at the same time as it increases its organized-complexity until it reaches some optimal stable plateau. Towards the end of the organism's natural lifespan, metabolism fails to maintain its organized-complexity at the optimal plateau for different reasons, which signals the onset of aging. Metabolism maintains the single individual, whereas replication maintains the design (i.e., template of structural information) after the individual dies.
The act of weathering and repair, therefore, can make a building more alive. This might shed some light on Japanese building tradition, in which some holy shrines are entirely rebuilt in the exact manner every few decades. There develops a psychological bonding between human beings and a structure that shows fractal patterns with weathering (but not if it becomes ugly or falls apart). In this analogy, minimalist, non-weathering structures do not metabolize. We are thus questioning the drive towards sleek building surfaces and geometries that oppose natural processes, and suggest that older techniques that accommodated the inevitable weathering are in fact more adaptive.
Replication is often considered as the main characteristic of living structure. Organisms reproduce by making copies of themselves. Nevertheless, it is possible to have a replicating structure that does not metabolize, as for example a virus. 4 It is also possible to have an entity that metabolizes but cannot replicate; there are examples in animals such as mules -- those exceptional typologies cannot propagate directly.
The simplest non-metabolizing templates (viruses) replicate more readily than animals with a higher complexity, because the latter's investment in metabolic and connective systems raises their organized-complexity. 4 Among metabolizing organisms, again those with a lower degree of organized-complexity (e.g., bacteria) replicate much more readily than higher animals, and are thus more abundant.
There are several different types of adaptation: an organism adapts to its environment by responding on the short term, and also the genotype (i.e., its DNA) adapts in the long term by evolving so as to better profit from existing or changing environmental situations. Short-term adaptation depends on connectivity to the environment -- being situated. Long-term adaptation follows a Darwinian selection process that culls portions of a population that are marginally worse adapted. Subsequently, survivors breed to define a new population having more of the positive adaptive trait.
We argue here for an architecture of adaptation, and criticize contemporary trends as being fundamentally non-adaptive. The reason that a non-adaptive architecture was able to develop is that the selection process among buildings and architectural styles is not as direct as selection among organisms. 2 Selection in architecture is driven by forces external to the natural process of adaptation, i.e. fashion, opinion, and politics.
Another way that organisms can act (when they are capable of doing so) is to change their environment to the organism's advantage. This is in some ways the opposite of adaptation. Nevertheless, the interventive practices that have survived evolutionary (natural) selection always appear as combined adaptive/interventive applications. Animals build nests; beavers build dams; a squid ejects ink to help it escape from a predator; certain plants inject chemicals around them that prevent competing plants from growing; etc. Humans are champions at this: we applied our intelligence for clothing, shelter, hunting, and agriculture, which give us an unbeatable advantage over all other animals.
Traditional architecture and urbanism concisely represent both adaptation and intervention. However, since about the middle of the twentieth century human constructions have become primarily intervention, with little or no attention paid to adaptation.
A living organism is naturally embedded in the world, interacting directly with it via direct sensory mechanisms. External feedback from internal sensors dictates the organism's behavior: recognition and pursuit of a food source; recognition and reaction to an environmental threat; fight or flight when faced with an aggressor; etc. An organism is situated in its environment, and is constantly sensing the state of the environment. Although one of the properties of living structure, this property is best understood from research in robotics rather than biology. 5,6
Situatedness depends upon the existence of sensory mechanisms that provide information about the world, and those, in turn, require a connective framework. The opposite of being situated is to exhibit behavior that is decided on the basis of abstract descriptions. We are not aware of any lower organisms that can do this -- it logically appears to be a capability of animals with sufficient neural development for memory storage. An organism cannot form and act on an internal representation of the world unless it has sufficient capacity to store it as memory.
In biology, correlations arise as connective mechanisms. These include structural ones, such as plant stems, animal bones, arteries, and ligaments; and informational ones such as in hormonal fields, nerves, eyes, and photosensitive surfaces on leaves. All of these are prime examples of organized-complexity, and each instance employs a complex physical network to perform a connective task.
An embryo develops by repeatedly splitting cells, so that its growth is obviously bottom-up, guided by genetic instructions in the DNA. Nevertheless, Alexander argues that embryonic development is impossible without a global control that keeps the growth from getting out of hand. 1 Whether this is due to a process of iteration in which each component helps to support and guide the development of other components, or to hormonal fields, what is important is that a global communication occurs. Each component (cell) of the embryo communicates chemically with the entire embryo existing at that time, so that each cell checks its position and future growth. In this way, embryonic cells develop either into muscle tissue or brain tissue, depending on their relative position at a particular time in the process.
A living system is one that acquires and actively uses information. 7 Information transfer takes many different forms in biology. Hormonal and nervous systems in animals are essential for interacting with the external world, and also for communicating internally within the organism. Stored genetic information encodes templates that permit the replication of individual cells, which replace worn-out cells in the body on a regular basis. For example, all except brain cells are routinely replaced in a mammal. Inherited information (across generations) is also stored in the brain, enabling all the instinctive behavior routines that permit animals to function. As we move up the evolutionary ladder, information and its processing plays an increasingly central role in life. The higher mammals are capable of learning, which is made possible by information storage mechanisms.
Human beings have evolved the ability to process information, both immediate (which is embedded in our environment) and stored (as mental images). Architects, in the process of distancing their work from actuality, have begun to rely more on stored information than immediate information. 2 The ability to store an artificial representation can go awry when such an artificial image replaces the real world. This ultimately leads to a state of disconnection, where a human being connects with some senses and emotions, but not with others. Many contemporary architects, in their efforts to impose abstract architectural and urban solutions on the environment, do just that.
4. ARCHITECTURE AND BIOLOGICAL PROCESSES.
Using science and technology constructively and humanely we can begin to sense the intimate connection between living structure and architecture. 1 We believe there is a direct analogy that can be drawn between metabolism in biological entities, and the process of maintaining complex structure (form) in non-biological ones. Buildings as non-natural artificial entities require varying degrees of repair by human beings after being built. Does the act of repair make the structure more alive? Does the process of repair in some way constitute a form of metabolism?
Considering human dominance of the world, and our physiological dependence on the physical structures we build around us, we can assume that there is an inherent necessity for buildings to replicate (even though they are inanimate entities). Often we see the replication of form in the built environment considered as a predicate of place, i.e. through indigenous localized forms. The replicating form is something that works within the limits of the material systems available in a certain region, and responds to local climatic conditions. So within a specific region very similar forms replicate and adapt to the programmatic differences and varying site conditions.
What are the forces that affect the survival of specific architectural templates? For example, building glass-walled high-rise buildings in both hot and cold climates is disastrous from an energy point of view. And yet, large rectangular buildings were universally adopted as an early twentieth-century design typology. This and other industrial examples are nonfunctional, but are copied from templates that have no relevance to human needs. There is a contradiction here with biological replication.
Non-adaptive forces in the built environment (dominant in a culture of architectural media-hegemony) give form to replicating structures around the world. Architectural and urban structures that simply replicate instead of growing out of very explicit local needs follow the architect's internal visual template that was developed generically, and not adaptively. This seems to be the crucial disconnection. 4
In deepening the biological analogy, Freeman Dyson identifies metabolism with the emergence of proteins (analogous to physical structure), and replication with the emergence of nucleic acids (analogous to a reproducible design typology). 8 We wish to identify connectivity with the emergence of complex sensory organs and communicative pathways in biological structure. Thus, connectivity is a much higher system function than either metabolism or replication, and makes possible adaptation, intervention, and situatedness in organisms. We are convinced that the architectural analogues of these properties are essential for a human built environment.
Situatedness is necessary for several of the other properties to occur. An architect who is not situated can respond neither to context, environment, nor the physicality of form. The architecture that comes out of this precondition turns out to lack connectivity and thus the ability to adapt.
5. INFORMATIONAL PROCESSES, ROBOTS, AND ADAPTATION.
Adaptation in living structure (forms) takes many different expressions at different scales. Internal adaptation balances temperature and chemical gradients leading to homeostasis and osmosis-regulation. This is essential to maintain the components of each organism in good working order. Next, we see adaptation to the physical environment, such as turning towards sunlight, reacting to temperature and threats, and adjusting thickness of fur for different seasons. Some reactions are immediate, whereas others adapt to longer-term environmental changes. Those adaptive mechanisms are essential for the survival of the individual organism. Finally, species adaptation occurs via natural selection. When an organism's physical mechanisms cannot cope with changing external conditions, some variants of a species die off; leaving those that might already have a slightly better adaptation. By evolving through survival, a species gradually changes its physical characteristics.
An architecture of adaptation must follow certain rules. The design process should consist of a large number of steps, so that feedback can influence the final product. This method is well known in adaptive software design, where the goal is left more loosely defined so that the designer can concentrate on the sequence of steps in its development. 9 Every design decision must be guided by its affect on the whole as it exists at that instant. This means that a design must communicate with each of its components (usually in the mind of the designer). A design has to develop by paying attention to the whole (e. g. landscape, surrounding buildings, historical culture) at each stage. 1 That is possible only if communicative systems exist that link the small scale to the large scale, and allow for feedback. 1,2 As there is no "nervous" or "hormonal" system on a building site or drawing board, the human brain must convey these connections through its own decisions.
Adaptive design is the opposite of generic or formalistic design. Adopting an adaptive design method imbibes the built environment with life, and at the same time replaces the lifeless industrial forms that have replicated uncontrollably since the early modernist era. Only a global conception establishes a framework of communication among all its constituent parts. This is not a formalistic method that imposes a preconceived order; but instead a mechanism for connectivity. The connective and feedback network then enables the form to evolve by conceptualizing the whole, just as in an embryo, rather than by the accretion of isolated parts. 1
Understanding basic life processes eventually leads us to questions about higher life forms, such as what constitutes intelligence. Biologists do not have all the answers to how we interact with our environment. For further insight, we turn to robotics engineers and experts in artificial intelligence. Laboratory robots provide an analogy of the sophisticated systems for processing environmental information that distinguish intelligent animals from other life forms.
It was assumed in the past that intelligent action occurred in three steps: information input; comparison with a stored representation; followed by a decision for action. Robots that work in a different way proved this hypothesis wrong. Rodney Brooks created robots that have no internal representation of the world: they use their environment directly for all their decisions. 5 The external world already contains all the information needed to reach decisions. Enormous computational resources, better used elsewhere, are saved by not duplicating this information inside memory. Built along these principles, Brooks' Mars Explorer robot was very successful at navigating the planet's rough terrain. 6 We can learn a lesson from this. Life forms are probably using intelligence without internal representation to connect with, and navigate their environment. Brooks argues that it would be hugely inefficient to do otherwise. 5
Robots that work with the alternative three-step process have also been built, but they get stuck on (and are severely limited by) their internal representation; they are incredibly slow to react, and cannot recognize novel forms. The reason is simple to see. Any internal representation of the world has to be abstract, formal, and highly simplified. In the robotics world, that means a universe made up only of monochromatic regular solids such as cubes, cylinders, and pyramids. Even so, robots built in this manner still get confused if an object in their abstract geometrical environment is not perfectly aligned, or if it casts a shadow. 6
An organism or robot that bases its behavior on direct sensory contact with the world is situated. By contrast, a robot or organism whose behavior is based upon abstract descriptions (internally stored information) is not situated, and thus cannot be said to satisfy all of the components of living structure. We argue that contemporary architects are no longer situated in the world, and are thus fundamentally different in their perception of the built environment from other human beings. True, this is not an inheritable genetic trait, but is only the result of a form of psychological conditioning that often occurs in today's media-driven culture. Nevertheless, the architectural profession has ensured the continuation of this trait across generations, by conditioning students to interact with the world via abstract images instead of via their direct sensory perception.
We again turn to navigation for an analogy of situatedness. Driving a car requires continuous sensory input and an interpretation of the immediate environment. While an abstract road map may guide the overall journey, the countless intermediate decisions are based on being situated in the physical road network, responding to every variation of the environment. Taking this temporal example, we propose its spatial analogue as the appropriate model for an adaptive architecture, in which design is sensitive to forms, needs, and surroundings at every stage.
6. ARCHITECTURE AND INTELLIGENCE.
This discussion touches upon the question of how discovered information is stored by an intelligent system that uses its physical environment as its primary representation. Such information is stored externally, by imprinting it on the environment. Human beings have many different options for how to do this. Representing information as structure can take many forms, including calligraphy, representational ornament, geometrical ornament, with the built object increasing in size up to architectural and urban spaces. In a fundamental sense, therefore, the traditional informationally-rich built environment can be interpreted as an information storage system, in and of itself. By contrast, the contemporary built environment and its architecture fails as an informational storage device. It is strictly an attempt at externalizing abstract mental images deficient in organized-complexity.
The type of information that is preserved in the traditional built environment is organized-complexity: precisely the type of information that defines living systems themselves. Thus, the traditional built environment consists of evolved and discovered solutions (schemata) that make our life easier and more meaningful.
Ollivier Dyens raises further questions about a stored internal representation, when that representation is entirely artificial. This is crucial to contemporary architecture, since increasingly, architects' world-views are formed by computer images. More than just formal systems, computer images substitute for a world that never existed. According to Dyens:
"Digital imaging technology suggests models of life based on a completely different representational style, one that is founded not on reproduction but on production ... With digital images, there are no primary, original moments to which we can point and say: 'This image is an analogy of that thing or that object'. On the contrary, digital images are, at once, worlds and models of worlds. Since digital images are not tied to exterior dynamics and do not extend toward exterior phenomena -- such as concepts, referents, ideas, etc. -- representation becomes an abyss, imploding into an endless collection of possible meanings." 10
An organism that exists in a symbolic abstracted domain is not totally alive, since there is nothing to ground it to the real world. It is more like a computer, executing an algorithm but not participating in the external world. This entity resides partially or entirely within its own model of an artificial world. One may go further and suggest that such an organism is not intelligent. As stated by Brooks:
"It is hard to draw the line at what is intelligence, and what is environmental interaction. In a sense it does not really matter which is which, as all intelligent systems must be situated in some world or other if they are to be useful entities. The key idea from intelligence is: 'Intelligence is determined by the dynamics of interaction with the world'." 5
In a direct analogy with the two types of robots mentioned above (those that work with and without an internal representation), architects have been imprinted with an artificial internal representation of the world that is irrelevant to physical reality. They judge the built environment by comparing it to this internal representation. By contrast, non-indoctrinated people (which includes everyone else) react directly with their environment, using their own senses and intuitions to make decisions about architecture. Based on the computational mechanism by which people respond to their environment, we have identified two distinct entities: psychologically-conditioned individuals (i.e. many of today's architects), and biologically-driven individuals (i.e. the majority of the rest of us).
Indigenous architecture represents human intelligence supported by memory (stored information) embedded in an external representation. Forms, spaces, textures and materials have evolved adaptively, in a way to maximally connect to the user. People build things that help them live and at the same time give back emotional satisfaction. Design decisions are based on direct perception and interaction with the form.
Formal architecture, by contrast, is more often than not an imposition of formal rules or external images, and buildings are judged by how closely they conform to some internal ideal stored in the architect's memory. While this method is fine for certain situations, that process is by definition not humanly adaptive. An internal representation is independent of reality; it is not checked and is thus prone to corruption, and can eventually replace the real world. Many an architect's judgment is driven by an artificial world picture that has nothing to do with human systems of perception. Architectural education has been geared towards the goal of generating this internal world picture (a representation of an artificial world) that is at odds with the physical world.
This analysis helps to explain the amazing disagreement between architects and everyday people about what constitutes a "good building". This difference of opinion is well known for domestic residences, where the tastes of architects are often diametrically opposite to those of everyone else. Another example is a recent debate about new buildings that maintain the spirit of Thomas Jefferson's original buildings on the campus of the University of Virginia. The faculty of its architecture school condemned the contextually conceived -- i.e., adaptive, hence traditional-looking -- buildings in the harshest possible terms. By contrast, everyone else, from students, to faculty, to the university trustees and visitors to the campus, loves them. What is probably happening is that everyone else is responding viscerally with those forms (by connecting to them in an immediate sense), whereas the reactions of architecture faculty come from comparing them to a stored internal representation of an artificial world.
Just as biological systems and natural processes give form to all living structures, so too these formative devices once rendered architecture and the built environment replete with life. Animated within the processes of everyday existence, architecture served humankind as an adaptive necessity, and as a conveyor of information -- one which worked to attach and engage human beings with their everyday environments. With the onslaught of industrialization, which sponsored an extended physical condition, the natural relationship between human beings and their environment has been severely compromised. We argue for an architecture of adaptation, one structured within the limits of natural processes and imbibed with informational content, such as needed to facilitate a more immediate (contextual) determination of architectural form.
1. Christopher Alexander, The Nature of Order: Books One to Four (Berkeley, California: Center for Environmental Structure, 2002-2005).
2. Nikos A. Salingaros, A Theory of Architecture (Solingen, Germany: Umbau-Verlag, 2006), Chapter 1.
3. Stephen Wolfram, A New Kind of Science (Champaign, Illinois: Wolfram Media Inc., 2001): 353.
4. Nikos A. Salingaros, Anti-Architecture and Deconstruction (Solingen, Germany: Umbau-Verlag, 2004), page 142.
5. Rodney A. Brooks, Cambrian Intelligence (Cambridge, Massachusetts: MIT Press, 1999).
6. Rodney A. Brooks, Flesh and Machines (New York: Pantheon Books, 2002).
7. Freeman Dyson, "Is Life Analog or Digital?", Edge <www.edge.org> (2001).
8. Freeman Dyson, Origins of Life, Revised Edition (Cambridge: Cambridge University Press, 1999).
9. James A. Highsmith, Adaptive Software Development (New York: Dorset House Publishing, 2000).
10. Ollivier Dyens, Metal and Flesh (Cambridge, Massachusetts: MIT Press, 2001): 87-88.