Abstract

Three forms of creativity are exemplified in biology and studied in ALife. Combinational creativity exists as the first step in genetic algorithms. Exploratory creativity is seen in models using cellular automata or evolutionary programs. Transformational creativity can result from evolutionary programming. Even radically novel forms can do so, given input from outside the program itself. Transformational creativity appears also in reaction-diffusion models of morphogenesis. That there are limits to biological creativity is suggested by ALife work bearing on instances of biological impossibility.

1 Introduction

In its broadest sense, creativity is the ability to generate new forms. Those forms include psychological or biological phenomena. Thus, creativity can be understood in both those contexts. This article—in which a previous account of creativity [3] is broadened to cover biology—shows how the two contexts differ, and how work in ALife can throw light on both.

(Some novel forms, namely chemical elements, have evolved in the stars. Others emerge as patterns in inanimate solutions. Social-collective creativity is key to human cultures. And many discussions of creativity appeal to theology. Those contexts are ignored here—but see [9].)

The most widely used sense of the term applies to human thought and activity. Understood in this way, creativity is the ability to generate ideas and/or artefacts that are new, surprising, and valuable. Let's call this psychological creativity, for short.

Biological creativity, by contrast, is the ability to generate new cells, organs, organisms, or species. This includes the (phylogenetic) creativity of evolution, the (ontogenetic) creativity of morphogenesis, and the (autopoietic) creation of individual cells.

Creativity (both psychological and biological) is often featured in ALife. And that's true whether we choose to consider ALife as art or as science.

On the one hand, the various types of computer art [8] include some that employ techniques widely used within ALife. Moreover, some (not all) of the artists concerned make a point of stressing the relationship between their artworks and living things [54, 30]. In their eyes, much of the aesthetic interest and value depends on this particular context.

(A few go even further, seeing their art as part of the search for strong ALife, or life-in-cyberspace. For instance, one claims that it requires us “to consider the power of technology to create life, rather than simply represent it” and that the audience is confronted by “the artificial creation of life and living systems” [49, p. 103]. Even if strong ALife is in fact impossible [2], their art can't be properly understood without realizing that this is its intended significance.)

On the other hand, ALife as science is frequently concerned with creativity too. For it often models biological evolution or development, both of which involve the generation of new forms.

Sections 2 and 3 enlarge on the definitions given above, and distinguish three types of creativity (combinational, exploratory, and transformational). Some ALife examples of each are addressed in Sections 456. Section 7 argues that biological self-organization in general, and morphogenesis in particular, counts as creativity. Finally, Section 8 suggests that some limits to biological creativity can be studied by ALife.

2 Novelty and Value

Creativity, as defined above, is the ability to produce new forms. And creative ideas/artefacts, besides being new, are also surprising and valuable. Each of these three criteria can be understood in different ways.

“New,” in this context, can be understood as referring either to the individual (person or organism) concerned or to the preceding history.

In psychological creativity, the novel idea may be new only for a particular person: Others may have had it countless times already. Alternatively, the idea may be new, so far as we know (and we may be mistaken), to the whole of human history. These two senses of “new” define what we may call P-creativity and H-creativity, respectively (P for psychological, H for historical). When speaking of creativity in art or science, we are usually more interested in H-creativity. But since H-creativity is a special case of P-creativity, to understand how it is possible we need first to understand how P-creativity is possible.

The same applies to biological creativity. In phylogenesis, H-new species (and H-new biological mechanisms) crop up from time to time. The creativity seen in morphogenesis, by contrast, is analogous to P-creativity—here more accurately termed I-creativity (I for individual). There have been trillions of cases of a chick ovum developing into a blastula, or an acorn into an oak. Despite their familiarity, such phenomena remain highly intriguing. How is it possible for yet another oak to be formed from a single undifferentiated cell?

For present purposes, we may assume that the identification of something as new (and especially as H-new) is unproblematic. In truth, however, it is not [4, p. 1.iii.f–g]. One difficulty, of course, is finding the facts. In locating psychological creativity, we want to know who said or did what, and when. In considering biological creativity, we want to know when and where a certain structure first occurred in phylogeny, or when a certain type of cell first differentiates in the embryo.

A deeper difficulty concerns just how to recognize or identify particular novelties. In attributing scientific creativity, how should we compare a novel but very general insight—or even a light-hearted joke—with a mathematically precise analysis and/or a careful experimental demonstration of the point in question? (Consider, for instance, the discovery of buckyballs in chemistry [21, 22, 44] or of backpropagation in connectionist AI [4, p. 12.vi.b–d].) Or how should we compare full-blown Cubism with its first intimation in a very early, perhaps even tentative and incoherent, sketch by an experimenting Paul Cezanne? Analogously, should biologists regard a patch of transparent epidermis as a primitive lens? That is only one of the many difficulties in identifying H-novelty in the evolution of vision [24]. All biological structures are affected in some way by photons: Is this proto-vision?

However, creativity is not mere novelty. A creative idea, for example, also has to be valuable. (This point is sometimes contested [39, 45]. But most accounts take value to be an essential criterion [34, p. 4].)

There are many types of value (e.g., beauty, scientific interest, musical harmony, usefulness). Value is assigned by judgments endorsed by sociocultural groups: from experts to peers, from academicians to the avant garde, and even including the fickle followers of fashionable celebrities. These values differ, and can change. So disputes about whether a certain idea is properly called “creative” may be based not only on disagreements about whether it satisfies this or that particular value, but also on whether that aspect should indeed be accepted as valuable.

It's normally assumed that the originator of the creative idea recognizes it as being valuable. Sometimes, this isn't so. (Johannes Kepler initially described his thoughts about non-circular planetary orbits as “a cartload of dung” [23, p. 217].) But in general, the originator sees, or soon comes to see, that the novel idea is worth pursuing. Indeed, they very often devote further time to consciously developing and improving their new idea, with specific value criteria in mind.

These points about the role of value in creative thought do not transfer smoothly to biology. Certainly, we normally think of evolution as creating valuable H-new forms, and of morphogenesis as creating valuable I-new structures. That's because we regard life itself as valuable, and also because we regard many living things as beautiful and/or exquisitely functional. But the sense in which the process of natural selection is “creative” is significantly different from that in which artists and scientists are.

Biology in general doesn't generate either ideas or artefacts. Nor does it involve conscious planning or self-monitoring, as much (though not all) human creativity does. And there is—at most—only one value involved: the fitness function of reproductive success.

3 The Three Types of Creativity

The preceding section showed that the words “new” and “valuable,” when used in discussing creativity, have more than one meaning. Now, let's consider “surprising.”

The term “surprising,” here, covers three different sorts of surprise. Each applies on two levels: phenomenological and procedural. In other words, the different sorts of surprise that we experience on encountering a creative idea correspond to different mechanisms for producing novelty. Those mechanisms mark three kinds of creativity: combinational, exploratory, and transformational.

(These are analytical distinctions between different psychological/biological processes. They are not intended as overall descriptions of the resulting forms. A single form—a painting, for instance, or an embryo—may have been generated by creative processes of all three types. In such cases, we need to ask which type of creative process has generated this or that particular aspect of the final form.)

The first sort of surprise arises when some unusual and unexpected combination of events happens. Very broadly speaking, it's a matter of statistics, not of style or structure.

Examples in art include visual collage and much poetic imagery—for instance, the unlikely combination of knitting and sleep in Macbeth's invocation: “Sleep, that knits up the raveled sleeve of care.” Scientific examples include William Harvey's description of the heart as a pump, and the Rutherford-Bohr account of the atom as a solar system. Examples in the biosphere include novel mutations in the genome.

The second sort of surprise, by contrast, is based in structure. It arises when something unexpected happens that is recognizable nevertheless as, in a sense, “more of the same.” That is, the novelty fits into some previously accepted style of thinking: a new benzene derivative, perhaps, or another Impressionist painting. Or, in the biological realm, another Dalmatian puppy, or maybe an unusual-colored rose.

These new instances are generated by the same structural constraints, or stylistic rules, that produced the earlier ones. The particular novelty may be surprising, but it can be seen—often, immediately—as something that was always a possibility. Before it was instantiated it was already waiting in the wings, for it was part of the space of possibilities defined by the familiar (conceptual or biological) constraints concerned.

This type of novelty exemplifies exploratory creativity. Where psychological creativity is involved, the possibilities—and, sometimes, the limiting boundaries—of the space may even be explored consciously and deliberately by the artist/scientist involved. In biological creativity, that doesn't happen. But the mutations and crossovers occurring in the genome can be seen as mini-explorations of the space of biological possibilities concerned. These are implicitly defined by the genome itself, together with the physics and chemistry that constrain gene expression, copying, and mutating.

The third sort of surprise, considered as a mode of experience, is deeper than the other two. It's the amazement prompted when a new form (idea, artefact, organism) arises that seems to be not just new but impossible. It couldn't have happened, we feel—and yet it did.

This impossibilist surprise is our phenomenological response to novelties that are generated by transformational creativity. Here, one or more of the defining constraints of the possibility space is itself altered, in a more or less fundamental way, so that structures that were strictly impossible before become possible—and, by hypothesis, instantiated.

Psychological examples include non-Euclidean geometry (wherein an explicit axiom is deliberately dropped), and the switch from string molecules to ring molecules in originating aromatic chemistry. Biological examples include gastric-brooding frogs (where the digestive juices are not secreted, to protect the young: See Section 8), and the differentiation of organs in a developing embryo.

4 Combinational Creativity in ALife

Combinational creativity rarely features in ALife as a standalone form of novelty generation whose results are considered valuable in themselves. (Exceptions include artworks generated partly by random access to the Internet, such as Christa Sommerer and Laurent Mignonneau's interactive installation The Living Room (2001).) But it often occurs as a first step on the road to exploratory or transformational creativity. In other words, combinational creativity within ALife usually exists as point mutations or crossovers defined by genetic algorithms (GAs).

Typically, the individual GA-driven changes are random. But they can be constrained in certain ways. Indeed, ALife workers using GAs have to decide beforehand what general types of mutation they will allow, and this will lead to certain sorts of novel combination rather than others.

For example, in his evolutionary image-generation program, Karl Sims [41] allowed mutations whereby one whole mini-program (capable of generating an entire image) could be nested inside another. This could lead to deep structural changes in the code, and hugely surprising changes in the resulting images. The evolutionary artist William Latham, by contrast, usually allowed only relatively superficial mutations—such as substituting one numeral in a mini-program for another [48]. As a result, the images originated at the various generations all shared a distinct family resemblance.

5 Exploratory Creativity in ALife

One example of exploratory creativity in ALife has just been mentioned: the use of GAs to explore a space of possibilities having aesthetic value. Sometimes, however, the interest lies rather in results that are analogous to biology.

A very early instance of the latter type was Thomas Ray's model Tierra [35, 36]. The biologically significant results, which were not all expected by Ray himself, included punctuated evolution (phenotypic saltations, despite underlying genetic continuity); the emergence of parasites; the growth of resistance to those parasites; and the development of hyperparasites that could overcome that resistance. That example is one of many showing that exploratory creativity can provide surprises. In general, the possibilities inherent within any given GA system may be largely hidden to the researchers concerned—as the development of parasites was to Ray.

In psychological creativity too, exploration of an accepted style—of music, painting, chemistry, etc.—can generate unexpected and highly valued results, even masterpieces. So exploratory creativity should not be underrated.

Indeed, even those artists or scientists who do manage to transform the current style of thinking normally then focus on exploring the newly transformed style, not on seeking a further transformation. This is apparent in retrospective exhibitions of painters wherein the canvases are arranged chronologically. Pablo Picasso and Alan Turing are two of the very few people who achieved successful stylistic transformation more than once. (Turing initiated a new style of thinking about computation, and also a new—and very different—way of thinking about morphogenesis.)

Another computational technique widely used in ALife is cellular automata, or CAs. A CA is a set of rules that define the possible changes in the individual cells of a matrix. Often, every cell is bound by identical rules. However, the rules may vary across the matrix, and/or they may change as time passes. Often, too, the rules allow or disallow certain changes depending only on the current state of the cell's immediate neighbors. Nevertheless, the state of far-distant cells can also have an influence. Finally, the rules are normally deterministic, but can be probabilistic.

A CA can be seen as a space for exploratory creativity. The primary interest, here, is in exploring the patterns that can be generated by the CA in question. But, as the previous paragraph implied, the overall nature of the CA (i.e., its defining rules) can be varied too. A CA model of biological development, for example, could include successive rule changes modeled on the switching on and switching off of genes.

Very simple CA rules, within a very simple matrix (a single line, generating another line below it), can generate surprisingly complex patterns. Some of these are highly naturalistic—like the markings seen on many sea shells [56]. This creative journey, from apparent simplicity to evident complexity, is of interest to biologists and artists alike.

Paul Brown is a computer artist whose CA work has been commissioned internationally. His exploration of CA space has recently raised an intriguing puzzle, namely, whether it is possible—and if so, how—to use CAs not to express his personal signature, but to delete it [5]. Committed as a young man to the modernist aesthetic, and drawn to computer art precisely because of its impersonal nature, Brown has tried for years to lose his authorial recognizability. But without success: A new “Brown” is still instantly recognized by the afficionado as a work by Brown.

In other words, Brown's explorations of CA space (and his many novel combinations of CA rules) have not come up with a fundamentally different style. The creative explorer, howsoever intrepid, has remained within the one country.

Given the threefold categorization of creativity in Section 3, one might say that this is only to be expected. Transformation, not merely exploration, is required for stylistic change. And Brown himself has recently said the same thing. Specifically, he has suggested that stylistic transformation could result from evolutionary ALife techniques [10]. The next section outlines Brown's attempts to achieve this, and asks whether it is even in principle possible for GAs to produce truly radical change.

6 Transformational Creativity in ALife

The most obvious—although, as we'll see in Section 7, not the only—candidate for ALife work on transformational creativity is evolutionary programming. Whether this technique can ever produce truly radical novelty is controversial. But evolutionary ALife can certainly give the appearance of transformational creativity.

This is evident in the Tierra parasites mentioned above, and in some computer art that uses evolutionary methodology to produce H-novel forms. For instance, the forms generated by Sims' system often share no visible family resemblance, appearing to come from utterly different styles.

However, some critics insist that this is a superficial illusion. Truly radical transformation in evolutionary computation systems, they argue, is in principle impossible. The GA program defines a particular set of possibilities, and no structure lying outside that range can occur. (In principle, to be sure, GAs and CAs are universal machines; but any actual instance is not.)

Such criticism is especially likely to come from people interested in biology. For in biological evolution, fundamental change has sometimes happened. Seven large-scale “major transitions” have been identified, including the passage from prokaryotes to eukaryotes and from asexual to sexual reproduction [33]. Many more specific instances have been noted too. As well as improving on a primitive eye, for instance (which these critics admit that an ALife GA can do), biology has come up with a light sensor where no such sensor existed before [11,13]. And this, they argue, is impossible within ALife.

I discuss the problem of radical novelty at length elsewhere [7]. Briefly, the answer is that radical novelties are possible in GA systems provided that there is some input to the system from outside itself.

If robots are involved, this input may be physical. At least one case has already occurred where a radically new sensor has evolved unexpectedly because of a physical contingency in the environment [1]. The research team was evolving electronic circuits, hoping to generate an oscillator. On one occasion, however, they evolved a primitive radio receiver. It turned out that the radio wave sensor depended on unforeseen parameters such as proximity to a PC monitor, the aerial-like properties of printed circuit boards, and the fact that the soldering iron left on a nearby workbench happened to be plugged in at the mains.

One of the team members later worked with the computer artist Brown in his attempt (noted in Section 5) to evolve a signature-free style of drawing [10]. Their strategy was to evolve robots equipped with pens, which could be raised/lowered as the robots moved. The resulting lines would then be subject to a fitness function (primarily determined by Brown), which—so it was hoped—would eventually evolve a non-Brown, but aesthetically acceptable, drawing style.

The experiment failed, in the sense that no aesthetically acceptable style emerged—still less, one lacking Brown's signature. For present purposes, however, the important point is that the team's decision to use robots was not a mere gimmick. They could have relied on a GA-driven computer graphics system, to evolve virtual robots brandishing virtual pens in cyberspace. But that would have been subject to the principled objection against radical GA transformation. Instead, they used real robots, which would encounter a variety of unexpected contingencies in the real world (a flex lying on the floor, for instance)—which contingencies, like the soldering iron mentioned above, might then feed constructively into the evolutionary process.

In principle, “input from outside the GA system itself” could be semantic rather than physical. Access to the Internet, for instance, has already been mentioned in reference to The Living Room (see Section 4). However, that possibility is more germane to evolutionary models of psychological creativity than of biological creativity. Where biological evolution is concerned, the system's physical environment—including the predator/prey and mating behavior of other creatures—is what matters.

7 Morphogenesis as Transformational Creativity

P-creativity (of all three types) is often unappreciated. In the worlds of art and science, the interest, respect, and international accolades (Nobel Prizes, exhibitions at the Tate or Venice Biennale) aren't offered to mere P-creativity. Individual artists and scientists want to be seen as H-creative. Quite apart from anything else, getting into the history books eases the burden of mortality. So, insofar as ALife work is focused on art, it seeks to foster H-creativity.

For science, by contrast, P-creativity is more interesting. So psychologists want to understand how an individual person can generate ideas that are novel to them (some of which may also be novel to all human thought). Moreover, people seeking to facilitate creativity focus on the more mundane examples, since H-creativity can't be conjured up on demand. Richard Feynman, for example, took pride in his H-creative pedagogical skills as well as in his H-creative physics [15]. But even Feynman, when acting as a teacher, focused primarily on encouraging P-creativity in his students. With any luck, H-creativity might sometimes arise as a special case.

Biology, too, offers a huge range of examples of non-historical creativity. As remarked in Section 2, these are better termed I-creative than P-creative, because they are individual but not psychological. What's more, many of these are instances of transformational I-creativity.

They concern self-organization in general, and morphogenesis in particular. Self-organization is the spontaneous generation of order (or “form”) from a base that is ordered to a lesser degree—where “spontaneous” means not magical, but following from the intrinsic nature of the system. As such, it counts as a form of transformational creativity.

Intriguing computational models of self-organization include ALife work on cellular autopoiesis [57,59], and connectionist research on the formation/development of neural networks as a result of random input [27,29, 52, 55].

But perhaps the most intriguing of all concern morphogenesis. The basic genome, of course, remains the same throughout embryogenesis. (No creativity there.) However, at the level of organs and overall bodily form there are huge transformational changes.

ALife research on morphogenesis has grown from the insights of the biologist D'Arcy Thompson [47] and, particularly, from Turing's article [50] on chemical morphogenesis. Both these men argued that the origin of new form in a developing organism is largely governed by basic laws of physics and chemistry. For Turing, the universal laws of interest were those governing reaction diffusion. Different chemicals, present in different concentrations and diffusing at different rates, will—in certain (quantitatively identifiable) circumstances—inevitably interact to produce, for example, spots, stripes, whorls, and 3D invaginations (e.g., gastrulation).

Using innovative mathematics, and aided by primitive computer calculations, Turing proved that this is in principle possible. However, he couldn't specify the chemicals (“morphogens”) concerned. Knowing nothing of what today's biologists term Hox genes, for example, he had to hypothesize an unknown “leg-evocator” [50, p. 39]. Similarly, his Ph.D. student Bernard Richards [37, 38] proved Turing's suggestion that distinct bodily symmetries of Radiolaria would arise if the morphogens were distributed on different axes within the cell—but the actual chemistry was a mystery.

Even Turing himself, had he lived, could not have taken these ideas much further. For that would have required advanced computer graphics and computer power—not to mention sophisticated experimental techniques for identifying the chemical reactions concerned. He had focused on the origin of order from a homogeneous base. But, as he explicitly acknowledged, this is only the first step. Most morphogenesis concerns the transformation of one already existing pattern into another, and then another, and yet another, and so on. (Think of the embryology of the neural tube, for instance.) The computer modeling of pattern from pattern by means of diffusion reactions couldn't be done until many years later. It was achieved, using Turing's own equations, by the computer graphics expert Greg Turk [51], and, using other equations, by the developmental biologist Hans Meinhardt [16].

Today, many developmental biologists are studying the I-creativity of morphogenesis. And some are constructing computer models of this type of self-organization. An interesting example is due to Brian Goodwin and colleagues [18, pp. 88–103; 20], who have provided a computational model of the development of an entire organism: the unicellular alga Acetabularia.

The cell in question, which can grow to a length of 2 centimeters, changes its shape significantly throughout its life cycle: at one stage, a shapeless blob; then, an elongated stalk; and next, a flattened top. The flat top develops a ring of knobs around the edge, sprouting into a whorl of laterals; finally, the laterals consolidate to form an umbrella-shaped cap. In other words, we are not here considering the autopoietic formation of a simple cell or cell wall from an unstructured base. Rather, we are dealing with a sequence of morphogenetic changes (pattern from pattern), during which the bodily form is successively transformed.

Based on extensive experimental work, the model features over thirty parameters representing the chemical changes concerned. They reflect factors such as the diffusion constant for calcium, the affinity between calcium and certain proteins, and the mechanical resistance of the elements of the cytoskeleton. The model simulates complex, iterative feedback loops wherein these parameters can change from moment to moment. The cell's control of the concentration of calcium ions affects, and is affected by, other conditions, such as the mechanical properties of the cytoplasm. So calcium metabolism underlies, and generates, various types of bodily morphology: e.g., stalks, flattened tips, and whorls.

The quantitative values of the parameters were restricted to the ranges observed in the actual organisms, but within those constraints they were randomized. Much as Turing (and Turk and Meinhardt) had played around with varying numerical values, to see which ones actually (and unpredictably) would generate new forms, so Goodwin's group did the same.

This required huge computer power. Indeed, it might have seemed impracticable, for varying thirty-plus numerical parameters is a tall order even for the largest computers. (And, one might add, even for biological evolution.) But it turned out that certain patterns—for example, alternating high and low concentrations of calcium at the tip of a stalk, seen by Goodwin as the emerging symmetry of a whorl—arose over and over again. They did not depend on a highly specific choice of parameter values. To the contrary, they emerged if the parameters were set anywhere within a large range of values.

Moreover, once the whorls had originated, they persisted. Accordingly, they could in principle become the ground for transformations leading on to other features. Indeed, this could take place at the phylogenetic level as well as in ontogenesis: H-creativity as well as I-creativity. So the evolution of organs such as the tetrapod limb [40], for instance, as well as the embryology, might be illuminated by such an approach.

The Acetabularia umbrella cap, by contrast, was never generated by Goodwin's model. Possibly, that would require one or more extra parameters, representing as yet unknown chemical interactions occurring in the real organism. Or perhaps such caps do lie within the possibility space defined by Goodwin's model, and so could in principle arise from it, but only if the numerical values are so highly limited that they are unlikely to be found by random search. (For the record, whorls of laterals weren't generated either, but that was purely because of a limitation in computational power. The whole program would have had to be executed on a lower level, and repeated for every lateral in the whorl.)

Goodwin was especially interested in the fact that whorls were generated many times, because they are generic forms. That is, they occur in many living things, both animals and plants. This implies, he claimed, that their explanation is due not primarily to specific biochemical mechanisms directed by contingently evolved genes, but rather to general mechanisms (such as reaction diffusion) found in most or even all living things.

Goodwin drew this implication because he was already committed to a structuralist view of biology, explicitly based on the ideas of Johann von Goethe and D'Arcy Thompson [53, 6]. Structuralism holds that a general science of morphology is possible. If that's so, it's highly significant for ALife.

8 Spaces of Impossibility?

Section 3 referred to “impossibilist” surprise, caused by the generation of forms that could not have arisen before some transformation of the preexisting possibility space had happened. The transformation had made structures possible that were previously impossible. But some things, including some things relevant to biology, really are impossible. No amount of transformational creativity could bring them about.

Even psychological creativity may be limited in this way. There may be aspects of the universe that the human mind is simply not capable of understanding—much as a dog is incapable of understanding algebra. (Some philosophers argue that the relation between brain and consciousness is one such feature [31].)

Biological creativity is apparently even more limited. Think of the perfectly cuboidal, shiny, black slab found by the cavemen in 2001: A Space Odyssey. Obviously, it's not a biological form. But why not?

Or, to pose a similar question: Why so few umbrellas? Biology has given us umbrella-shaped forms: jellyfish, for instance, and Acetabularia caps too. But there are relatively few of them. There are even fewer biological gears, or rotary organs mounted on axles (seen on the insect Issus coleoptratus and bacterial flagella respectively). Why is biological creativity constrained in this way? Is it just a historical accident, due to some contingency lying way back in the phylogenetic record? Or is there some deeper reason?

D'Arcy Thompson, Turing, and structuralist biologists in general focused on how the laws of physics and chemistry implicitly define realms of biological (morphogenetic) possibility, such as stripes, spots, rings of petals or tentacles, and invagination. But the same physical laws imply some impossibilities, too—for instance, 100-ton elephants, or human beings one mile high. D'Arcy Thompson remarked the impossibility of various apparently conceivable organisms.

Most biologists have more than enough to do than to venture into such exotic realms. Sometimes, to be sure, they have no choice. For living things surviving in seemingly lethal environments are occasionally found. Prior to the discovery of organisms in the extreme conditions found in deep-sea vents, for instance, such creatures would have been thought impossible by most biologists. Similarly, most biologists would have scorned the idea that a female frog might swallow her own fertilized eggs and allow them to develop into tadpoles and then young frogs inside her stomach. But such a creature (now extinct, though being cloned) existed in Queensland, Australia, until about 1980.

These examples, of sea-vent organisms and gastric-brooding frogs, show how risky it is to speculate that something is biologically impossible. One biologist who did this was the reproductive specialist Jack Cohen. He worked with various science fiction writers and filmmakers, advising them on what imaginary creatures are or are not possible on planets having physical conditions different from ours [14].

The ALife researcher Sims threw light on these matters both intentionally and accidentally. He ran an experiment wherein he evolved both the bodily form and the behavior of creatures competing in a virtual world [42, 43]. In most cases, the simulated physics of their world was the same as ours. And in those cases, lifelike (although sometimes highly unfamiliar) anatomies and behavior emerged. But on one occasion, when defining the physics of the virtual world, he forgot to include a principle concerning inertia. As a result, ludicrously un-lifelike—alias impossible—creatures appeared. They were constrained by, and adapting to, a different sort of physical world. They simply could not have functioned on good old terra firma.

Other ALife work that is potentially relevant here includes simulations of behavior detailing the effects of specific muscles. For instance, consider virtual fish that exist and behave in a virtual world whose physics—in particular, the physics of water—is fundamentally like ours [46]. The major bodily movements, and associated changes in body shape, result from twelve internal muscles (conceptualized as springs). Minor movements result from the definitions of seventy-nine other springs and twenty-three nodal point masses, whose (virtual) physics generates subtly lifelike locomotion. A host of questions about muscular possibilities—and impossibilities—could be explored by altering and/or deleting one or more of these simulated organs.

An example of ALife work explicitly focused on questions about unrealized possibilities is David Willshaw's research on the control of swimming in lampreys [19]. Lampreys swim not by using paired fins, as fish do, but by rhythmic undulations of their eel-like bodies. By the early 1990s, many of the biological details were known: the neural circuitry, the transmitters, and the membrane properties involved. Willshaw's team wanted to discover whether any (nonexistent but) viable lampreys might have evolved, had the genetic mutations been different.

Using GAs, they showed that there are many network architectures capable of controlling this type of swimming. Moreover, some of their artificial networks, despite being composed of “neurons” closely based on the biological data, were much more efficient than those found in real lampreys. Some had a frequency range five times larger. Even when the connections (and their type: excitatory or inhibitory) were fixed to be identical with those in real lampreys, some networks evolved—within only 100 generations—with frequency ranges three times larger. (This is one of many examples showing that biological evolution cannot be trusted to find the optimal solution for a given computational task; cf. [32, pp. 53–73; 17].)

To be sure, the emphasis in these ALife studies of fish and lampreys was on how certain types of movement are possible, not impossible. But a shift of vision would enable them to throw light also on what behaviors certain (perhaps imaginary) muscular anatomies could not support.

In sum, and despite the astonishing ingenuity of the living world we see around us, the limits on biological creativity are interesting too. ALifers might want to consider that their task is not only to describe “life as it is” and “life as it could be” [26], but also life as it couldn't be—and to explain why.

References

1
Bird
,
J.
, &
Layzell
,
P.
(
2002
).
The evolved radio and its implications for modelling the evolution of novel sensors
. In
Proceedings of Congress on Evolutionary Computation, CEC-2002
,
1836
1841
.
2
Boden
,
M. A.
(
1999
).
Is metabolism necessary?
British Journal for the Philosophy of Science
,
50
,
231
248
.
Reprinted in M. A. Boden (2010). Creativity and art: Three roads to surprise (pp. 235–254). Oxford, UK: Oxford University Press
.
3
Boden
,
M. A.
(
2004
).
The creative mind: Myths and mechanisms
, 2nd ed., revised, expanded.
London
:
Routledge
. (
1st ed., 1990
.)
4
Boden
,
M. A.
(
2006
).
Mind as machine: A history of cognitive science
.
Oxford, UK
:
Clarendon Press
.
5
Boden
,
M. A.
(
2010
).
Personal signatures
. In
M. A.
Boden
,
Creativity and art: Three roads to surprise
(pp.
92
124
).
Oxford, UK
:
Oxford University Press
.
6
Boden
,
M. A.
(
2013
).
From Goethe to Goodwin, via von Foerster
. In
D.
Lambert
,
C.
Chetland
, &
C.
Millar
(Eds.),
The intuitive way of knowing: A tribute to Brian Goodwin
(pp.
108
119
).
Edinburgh
:
Floris Books
.
7
Boden
,
M. A.
(
forthcoming
).
Can evolutionary art produce radical novelty?
In
M. A.
Boden
&
E. A.
Edmonds
(Eds.),
From fingers to digits: Towards a new aesthetic
(
Chap. 11
).
Cambridge, MA
:
Leonardo/MIT Press
.
8
Boden
,
M. A.
, &
Edmonds
,
E. A.
(
2010
).
What is generative art?
In
M. A.
Boden
(Ed.),
Creativity and art: Three roads to surprise
(pp.
125
163
).
Oxford, UK
:
Oxford University Press
.
9
Bown
,
O.
(
2012
).
Generative and adaptive creativity: A unified approach to creativity in nature, humans, and machines
. In
J.
McCormack
&
M.
d'Inverno
(Eds.),
Computers and Creativity
(pp.
361
381
).
Berlin
:
Springer
.
10
Brown
,
P.
,
Bigge
,
B.
,
Bird
,
J.
,
Husbands
,
P.
,
Perris
,
M.
, &
Stokes
,
D.
(
2007
).
The drawbots
. In
Proceedings of MutaMorphosis: Challenging arts and sciences
.
11
Cariani
,
P.
(
1992
).
Emergence and artificial life
. In
C. G.
Langton
,
C.
Taylor
,
J. D.
Farmer
, &
S.
Rasmussen
(Eds.),
Artificial life II
(pp.
775
797
).
Redwood City, CA
:
Addison-Wesley
.
12
Cariani
,
P.
(
1993
).
To evolve an ear: Epistemological implications of Gordon Pask's electrochemical devices
.
Systems Research
,
10
(
3
),
19
33
.
13
Cariani
,
P.
(
1997
).
Emergence of new signal-primitives in neural systems
.
Intellectica
,
2
,
95
143
.
14
Cohen
,
J.
, &
Stewart
,
I.
(
2004
).
What does a Martian look like? The science of extra-terrestrial life
.
London
:
Ebury Press
.
15
Feynman
,
R. P.
,
Leighton
,
R. B.
, &
Sands
,
M.
(
1995
).
The Feynman lectures on physics including Feynman's tips on physics: The definitive and extended edition
, 2nd ed.
New York
:
Addison-Wesley
. (
1st ed., 1964
.)
16
Gierer
,
A.
, &
Meinhardt
,
H.
(
1972
).
A theory of biological pattern formation
.
Kybernetik
,
12
,
30
39
.
17
Gigerenzer
,
G.
, &
Goldstein
,
D. G.
(
1996
).
Reasoning the fast and frugal way: Models of bounded rationality
.
Psychological Review
,
103
,
650
669
.
18
Goodwin
,
B. C.
(
1994
).
How the leopard changed its spots: The evolution of complexity
.
London
:
Weidenfeld & Nicolson
.
A new preface by the author added in 2nd ed., Princeton, NJ: Princeton University Press (2001)
.
19
Ijspeert
,
A. J.
,
Hallam
,
J.
, &
Willshaw
,
D.
(
1997
).
Artificial lampreys: Comparing naturally and artificially evolved swimming controllers
. In
P.
Husbands
&
I.
Harvey
(Eds.),
Fourth European Conference on Artificial Life
(pp.
256
265
).
Cambridge, MA
:
MIT Press
.
20
Jaeger
,
J.
, &
Monk
,
N.
(
2013
).
Keeping the gene in its place
. In
D.
Lambert
,
C.
Chetland
, &
C.
Millar
(Eds.),
The intuitive way of knowing: A tribute to Brian Goodwin
(pp.
153
186
).
Edinburgh
:
Floris Books
.
21
Jones
,
D. E. H.
(
as Daedalus
) (
1966
).
Note in Ariadne column
.
New Scientist
,
32
(
3 Nov.
),
118
119
.
22
Jones
,
D. E. H.
(
1982
).
The inventions of Daedalus
.
Oxford, UK
:
W. H. Freeman
.
23
Koestler
,
A.
(
1975
).
The act of creation
.
London
:
Picador
. (
First published 1964
.)
24
Land
,
M.
(
2001, 2012
).
Animal eyes
, 1st and 2nd eds.
Oxford, UK
:
Oxford University Press
.
25
Langton
,
C. G.
, Ed. (
1989
).
Artificial life, proceedings of an interdisciplinary workshop on the synthesis and simulation of living systems, Sept. 1987
.
Redwood City, CA
:
Addison-Wesley
.
26
Langton
,
C. G.
(
1989
).
Artificial life
. In
[25]
(pp.
1
47
). Revised version in
M. A.
Boden
(Ed.)
(1996). The philosophy of artificial life (pp. 39–84)
.
Oxford, UK
:
Oxford University Press
.
27
Linsker
,
R.
(
1986
).
From basic network principles to neural architecture (series)
.
Proceedings of the National Academy of Sciences of the U.S.A.
,
83
,
7508
7512
,
8390–8394, 8779–8783
.
28
Linsker
,
R.
(
1988
).
Self-organization in a perceptual network
.
Computer
,
21
,
105
117
.
29
Linsker
,
R.
(
1990
).
Perceptual neural organization: Some approaches based on network models and information theory
.
Annual Review of Neuroscience
,
143
,
257
281
.
30
McCormack
,
J.
(
2004
).
Impossible nature: The art of Jon McCormack
.
Melbourne
:
Australian Centre for the Moving Image
.
31
McGinn
,
C.
(
1989
).
Can we solve the mind-body problem?
Mind
,
98
,
349
366
.
32
Marr
,
D. C.
(
1982
).
Vision: A computational investigation into the human representation and processing of visual information
.
San Francisco
:
W. H. Freeman
.
33
Maynard-Smith
,
J.
, &
Szathmary
,
E.
(
1995
).
The major transitions in evolution
.
Oxford, UK
:
Oxford University Press
.
34
Paul
,
E. S.
, &
Kaufman
,
S. B.
(
2014
).
The philosophy of creativity: New essays
.
Oxford, UK
:
Oxford University Press
.
35
Ray
,
T. S.
(
1992
).
An approach to the synthesis of life
. In
C. J.
Langton
,
C.
Taylor
,
J. D.
Farmer
, &
S.
Rasmussen
(Eds.),
Artificial Life II
(pp.
371
408
).
Redwood City, CA
:
Addison-Wesley
.
36
Ray
,
T. S.
(
1994
).
An evolutionary approach to synthetic biology: Zen and the art of creating life
.
Artificial Life
,
1
,
179
210
.
37
Richards
,
B.
(
2005
).
Turing, Richards and Morphogenesis
.
Rutherford Journal
,
1
().
38
Richards
,
B.
(
2013
).
A solution for the morphogenical equations for the case of spherical symmetry
. In
S. B.
Cooper
&
J.
van Leeuven
(Eds.),
Alan Turing: His work and impact
(pp.
818
826
).
Amsterdam
:
Elsevier
.
39
Runco
,
M. A.
(
2010
).
Parsimonious creativity and its measurement
. In
E.
Villalba
(Ed.),
Measuring creativity: Proceedings of European Council Meeting on Creativity and Innovation
(pp.
393
405
).
Luxembourg
:
Publications Office of the European Union
.
40
Shubin
,
N. H.
, &
Alberch
,
P.
(
1986
).
A morphogenetic approach to the origin and basic organization of the tetrapod limb
.
Evolutionary Biology
,
20
,
319
387
.
41
Sims
,
K.
(
1991
).
Artificial evolution for computer graphics
.
Computer Graphics
,
25
(
4
),
319
328
.
42
Sims
,
K.
(
1994
).
Evolving 3D-morphology and behavior by competition
.
Artificial Life
,
1
,
353
372
.
43
Sims
,
K.
(
1994
).
Evolving virtual creatures
. In
Computer graphics (SIGGRAPH '94 Proceedings)
(pp.
15
22
).
44
Smalley
,
R. E.
(
1996
).
Discovering the fullerenes
.
Reviews of Modern Physics
,
69
(
3
),
722
730
.
45
Stokes
,
D.
(
2011
).
Minimally creative thought
.
Metaphilosophy
,
42
(
5
),
658
681
.
46
Terzopoulos
,
D.
,
Tu
,
X.
, &
Gzeszczuk
,
R.
(
1994
).
Artificial fishes with autonomous locomotion, perception, behavior, and learning in a simulated physical world
. In
R. A.
Brooks
&
P.
Maaes
(Eds.),
Artificial Life IV: 4th International Workshop on the Synthesis and Simulation of Living Systems: Selected papers
(pp.
17
27
).
Cambridge, MA
:
MIT Press
.
47
Thompson
,
D. W.
(
1917, 1942
).
On growth and form
.
Cambridge, UK
:
Cambridge University Press
.
48
Todd
,
S. C.
, &
Latham
,
W.
(
1992
).
Evolutionary art and computers
.
London
:
Academic Press
.
49
Tofts
,
D.
(
2005
).
Artificial nature
. In
D.
Tofts
(Ed.),
Interzone: Media art in Australia
(pp.
80
103
).
Fisherman's Bend, Victoria
:
Craftsman House
.
50
Turing
,
A. M.
(
1952
).
The chemical basis of morphogenesis
.
Philosophical Transactions of the Royal Society, Series B
,
237
,
37
72
.
51
Turk
,
G.
(
1991
).
Generating textures on arbitrary surfaces using reaction-diffusion
.
Computer Graphics
,
25
,
289
298
.
52
Von der Malsburg
,
C.
(
1973
).
Self-organization of orientation sensitive cells in the striate cortex
.
Kybernetik
,
14
,
85
100
.
53
Webster
,
G.
, &
Goodwin
,
B. C.
(
1996
).
Form and transformation: Generative and relational principles in biology
.
Cambridge, UK
:
Cambridge University Press
.
54
Whitelaw
,
M.
(
2004
).
Metacreation: Art and artificial life
.
Cambridge, MA
:
MIT Press
.
55
Willshaw
,
D. J.
, &
von der Malsburg
,
C.
(
1976
).
How patterned neural connections can be set up by self-organization
.
Proceedings of the Royal Society, Series B
,
194
,
431
445
.
56
Wolfram
,
S.
(
1984
).
Cellular automata as models of complexity
.
Nature
,
311
,
419
424
.
57
Zeleny
,
M.
(
1977
).
Self-organization of living systems: A formal model of autopoiesis
.
International Journal of General Systems
,
4
,
13
28
.
58
Zeleny
,
M.
(Ed.) (
1981
).
Autopoiesis: A theory of living organization
.
New York
:
Elsevier
.
59
Zeleny
,
M.
,
Klir
,
G. J.
, &
Hufford
,
K. D.
(
1989
).
Precipitation membranes, osmotic growths, and synthetic biology
. In
[25]
(pp.
125
139
).

Author notes

*

Department of Informatics, University of Sussex, Brighton, BN1 9QJ, United Kingdom. E-mail: M.A.Boden@sussex.ac.uk