The Ontological Status of Artificial Life

back to homepage

 
  

Figure 1. The Founders of Computer-theory and Artificial Life.
Alan Turing, John von Neumann and Christopher Langton


Introduction

This Essay is not ment to be an extensive introduction into the new field of Artificial Life itself, but an evaluation of its ontological status. This means an evaluation about the reality status of its creations, and not in the first place an evaluation of their aliveness.
Good introductions into the field are

EMMECHE, C., 1996, The Garden in he Machine, The Emerging Science of Artificial Life

LEVY, S., 1993, Artificial Life, A Report from the Frontier where Computers meet Biology

Remark : This Essay presupposes many results from all the forgoing Essays.

Why an Essay about Artificial life?
This website is devoted to Ontology, the study of Being, with emphasize on beings. It is meant to be a general theory about some principles of being, especially the generating principles. We want to investigate where those generating principles (which we can denote as the Essences of the beings in question) reside and assess their reality status and their relation with all the other aspects of the beings in question, and in what way and how far these principles determine the holistic- or totality-nature of those beings giving rise to their individuality. Such beings are for example crystals, and especially organisms. The latter are in a way prototypic examples of holistic beings, culminating in man, having selfconsciousness, and so experiencing his individual ONE-ness.
And because the theory should be general and universal, we cannot confine ourselves to the study of life which we encounter just on this planet, because it seems, in contrast with molecules and especially crystals, that life as we know it is not the only possible form of life. While a material life, based on chemistry, is hardly possible with other main materials (other than Carbon compounds), a totally different kind of life maybe possible, for instance life within a computer environment. So in order to be universal we must also study these alternative life-forms. And experience tells us that when we plunge into the problems of Artificial Life, then we are automatically also engaged in deep questions about life-as-we-know-it, especially its ontology.

There are several forms of Artificial Life, like for instance attempts to create life from a soup of chemicals, but in this Essay we will confine ourselves to computer-based Artificial Life (a-life).
On the one hand the relevant computer-based creations can be just simulations of (some aspects of) real life, i.e. the form of material carbon-based life we encounter on Earth. Such simulations are meant to increase our insight in life-as-we-know-it.
On the other hand the relevant computer-based creations are meant to BE a life-form themselves. This is Artificial Life in its strict (" strong ") sense.
Of course at present there still does not exist any computer-based creation that is unanimously ackknowledged as being alife. What is accomplished is just the computer-based creation of some aspects which are believed to belong to one form of life or another. Most of these aspects are organic functions and behavior, not organic form, i.e. morphology (L-systems are an important exception). But of course it could be that computer-based life consists mainly of function and behavior and less of explicit morphological structure. Behavior in the real material world is made possible by morphological structure. In computer-based life, behavior is made possible by imposing formal Rules on simple morphological structures. The creatures, presently crafted or evolved within computers are, as has been said, not yet alife, but can be considered as computer-based proto-organisms, comparable with the real creatures that inhabited the Earth immediately before the advent of life : they subsequently evolved into real living organisms.
In order to assess whether computer creatures are really alife we need, it seems, some criteria for something to be alife, in short we need a definition of life. But such a definition is not yet within reach. It moreover seems that there cannot be such a definition. Life should be gauged on a continuum, and not granted according to a binary decision. Things can be more or less alife. Life is a property showing degrees. A most important general condition for aliveness is the existence of complex systems, as a key component in biological entities. A complex system is one whose component parts interact with sufficient intricacy that they cannot be predicted by standard linear equations. So many variables are at work in the system that its overall behavior can only be understood as an emergent consequence of the holistic sum of all the myriad behaviors embedded within (LEVY, p. 7/8). What, within all the uncertainty about the nature of life, is generally accepted is that life does NOT possess some supernatural component.
To get some flavor of Artificial Life I will cite a part of the opening page of LEVY's book Artificial Life :

" The creatures cruise silently, skimming the surface of their world with the elegance of ice skaters. They move at varying speeds, some with the variegated cadence of vacillation, others with what surely must be firm purpose. Their bodies -- flecks of colors that resemble paper airplanes or pointed confetti -- betray their needs. Green ones are hungry. Blue ones seek mates. Red ones want to fight.
They see. A so-called neural network bestows on them vision, and they can perceive the colors of their neighbors and something of the world around them. They know something about their own internal states and can sense fatigue. They learn. Experience teaches them what might make them feel better or what might relieve a pressing need.
They reproduce. Two of them will mate, their genes will merge, and the combination determines the characteristics of the offspring. Over a period of generations, the mechanics of natural selection assert themselves, and fitter creatures roam the landscape.
They die, and sometimes before their bodies decay, others of their ilk devour the corpses.In certain areas, at certain times, cannibal cults arise in which this behavior is the norm. The carcasses are nourishing, but not as much as the food that can be serendipitously discovered on the terrain.
The name of this ecosystem is PolyWorld, and it is located in the chips and disk drives of a Silicon Graphics Iris Workstation. The sole creator of this realm is a researcher named Larry Yaeger, who works for Apple Computer. It is a world inside a computer, whose inhabitants are, in effect, made of mathematics. The creatures have digital DNA. Some of these creatures are more fit than others, and those are the ones who eventually reproduce, forging a path that eventually leads to several sorts of organisms who successfully exploit the peccadilloes of PolyWorld."

Figure 2. The creatures of PolyWorld have evolved from random genomes : Over generations they become skilled at seeking food (green patches), reproducing, and fighting.
( After LEVY, 1992, Artificial Life )

PolyWorld is a typical example of an Artificial Life effort. The possibility to evolve in an evolutionary sense is considered crucial. And clearly we see the emphasis on behavior, and the absence of almost any individual morphology. It is, it seems, a consequence of computer-based life.

Artificial Life tries to uncover which properties of life-as-we-know-it are typical for just this kind of life, and which properties are universal for living beings, properties, that must be possessed by every kind of life. On the latter properties we will focus our discussion and will guarantee that our ontological interpretation will be general. For us it is important that Artificial Life starts with very simple elements that interact with each other according to explicit rules (dynamical laws), and this will give us a means to assess the general nature of ultimate dynamical laws governing each organism including real ones. As outlined in a number of essays in this website such an ultimate dynamical law, governing the morphology and behavior of an organism can be interpreted as the prime ontological principle of that organism, its Essence.


Investigation into the nature of the Ultimate Dynamical Law of an holistic being, done by an evaluation of the ontological status of Artificial Life

A collection C of (relatively) elementary material entities implicitly and immanently contains the universal Laws of Nature. It also contains several special dynamic laws, say, L1, L2, L3, et cetera, immanent in a particular collection, but often absent in another (this other collection contains other specific dynamical laws). The universal and specific laws determine the possible interactions of the constituents (the elementary material entities) and so their dynamic behavior. Every material entity, and every collection of them must conform to the universal laws of Nature, like the fundamental conservation laws of physics. These laws are then further constrained by the nature of the relevant entities, resulting in the mentioned special dynamical laws. These laws are consequently not universal. But even then there are still several possible special laws that could equally operate within the collection C of material entities. Which one of these special laws finally will operate depends on certain prevailing conditions in which that collection of entities finds itself. This law will determine the interactions between the entities (components) of collection C, and will be responsible for the global change of the collection, for example the emergence of high-level structures and behavior.
The special dynamical law L1 is immanent in the special properties (part of the total set of properties) (L1P1), (L1P2), (L1P3), et cetera, of the constituents (of the collection C).
The special dynamical law L2 is immanent in the special properties (also a subset of the total set of properties) (L2P1), (L2P2), (L2P3), et cetera, of the constituents, where it is possible that some overlap exists, for instance (L2P1) = (L1P3).
If we focus on, say the dynamical law L1, then we in fact focus on the properties (L1P1), (L1P2), (L1P3), etc. If so, then all other properties of the constituents are irrelevant (i.e. per accidens ) with respect to the dynamical law L1. So all these other properties, taken together, form the substrate of the set of properties { (L1P1), (L1P2), (L1P3), ... }, and thus the substrate of the dynamical law L1.
Because of the per accidens nature of that substrate, we can abstract from it and contemplate the dynamical law L1 in itself.
When we consider another set of constituents, that shows the same properties (L1P1), (L1P2), (L1P3), ... , but differing with respect to other properties (all of them or a part thereof) then we have a different substrate for the same dynamical law L1.
So in principle it is possible that a dynamical law is medium-independent. This dynamical law L1 can be considered as FORM, while the substrate can be considered as MATTER (matter and form taken here as ontological principles). It is clear that the notions FORM and MATTER in this context are not absolute, but relative. The only absolute matter is PRIME MATTER, i.e. matter without content, the ultimate substrate of (radical) change. All the matter between PRIME MATTER and the dynamical law L1 (considered as FORM) is not matter in an absolute sense, it has content (because it comprises all the properties that remain when we have abstracted the properties that embody the dynamical law L1).
When we consider the dynamical law L1 as abstracted from the collection C of entities, then this law is not a real entity. To be real it must have a substrate.
The entities of different collections, but harboring the same dynamical law L1, will interact with each other in the same way, but these interactions will result in the generation of different Totalities for those different collections. But the differences will be per accidens when the Totalities exceed a certain size in terms of number of constituents, a complexity-threshold, as is the case for all organisms. So a certain kind of organism can be created from different collections of entities, or said differently, a certain kind of organism can be created with different materials, as long as such a material harbors that same dynamical law (in our instance the law L1.
The dynamical law L1, even in its abstracted form, is not pure FORM. It is FORM, only with respect to its substrate, but it is MATTER as well, because it represents directly the properties (L1P1), L1P2), (L1P3) ..., and these are material properties.

A-life tries to establish dynamical laws like L1, that, when implemented on a substrate, generates an organism.
A-life generally uses the computer as a substrate. It must provide the computer-entities (screen-pixels say) with the relevant properties, but this can only be done by associating them with the dynamical law, which is by necessity transcendent with respect to these entities (in contradistinction with real systems). Now these entities interact according to that transcendent law and in doing so we could perhaps -- from that point onwards -- interpret that law as immanent. The substrate, which is per accidens with respect to the law, IS (identical with) the rest of the relevant computer-hardware. It depends on the very character of the dynamical law whether we will assess (i.e. evaluate) the generated Totality (Totalities) as living. The dynamical law itself implies only a sequence of logical steps, but if this law is implemented in hardware (its substrate) then we will obtain a material process.


Ultimate and Penultimate Law of an Organism

Virtual and Actual History of an Organism

For (more about) the concept of ultimate and penultimate dynamical law, see the Essay on Living Dissipative Systems (Organisms), Part One, fourth section .
A penultimate law for generating an organism is in fact a rule (which is evolutionary changable) which presupposes the presence of already complex building-blocks (for instance aminoacids). So this penultimate law (rule) is a formal set of instructions. This set is constrained by the given structures of the available building blocks (as as penultimate elements of the dynamical system generating the organism) and is as such a descriptional information compression, because the information inherent in those building-blocks does not have to be included in this set of rules. Projecting backwards from those building-blocks to their constituent elements (atoms) yields the Ultimate Dynamical Law, which is selected out of the potential domain of laws inherent in the collection of atoms. So the Ultimate Dynamical Law -- the Essence of the Totality (the generated being) under study -- is revealed by this projecting-back from the penultimate law.
The large potential of possible ultimate dynamical laws, inherent in a collection of atoms, is founded in the enormous number of possible atomic configurations (combinations) that can result from the set of atoms.
An organism is the result of a whole series of constraints (demanded and caused by the environment) which together cause the particular ultimate law to ' crystallize'. We can only contemplate this law from the already formed organism [This, by the way, illustrates the difference between on the one hand the practice and goals of Natural Science, and the way of looking at things in Philosophy (in this case Ontology) on the other]. Only seen from the already formed organism we can say that that particular ultimate dynamical law had actually operated, which can then be interpreted as the Essence of that particular organism. The actual evolutionary history of the organism has of course been very whimsical indeed, with a lot of contingent events. The ultimate dynamical law is seen only with hindsight and represents a kind of conceptual shortcut , from the organism to its constituent atoms (atomic species). So the Essence (= ultimate dynamical law) of an organism (i.e. a particular organism), so conceived, is highly abstract and non-historical. The actual historical (evolutionary) path is per accidens with respect to this Essence. The part of the direct, but conceptual, (path of) generation of the organism, (conceptually) governed by the ultimate dynamical law, leading from the collection of atoms to the establishment of the penultimate law (this law takes care of the ontogenetic development of the individual organism), we can call the organism's ' virtual history ' (in contrast with its actual evolutionary history, that has lead from the origin of life to the formation of that particular penultimate law -- which we must seek in and around the DNA of that particular organism). Probably this Essence (so conceived),of one or another organism, can be equated with the (ultimate) algorithm, for one or another computer organism looked for in Artificial Life research. Such an Essence and also such an algorithm is thereby always integrated within a complex of broader laws that govern ecosystems and evolution.

To explain the concept of the Ultimate Dynamical Law of a particular organism (or whatever Totality) once more, I can state the following :

Actual, Contingent History of the organic Totality

When we trace back the actual ontogenetic and evolutionary (phylogenetic) history of a particular organism, we obtain a succession of material states, leading from the organism to a certain collection of atoms and simple chemicals of the primordial atmosphere. Of course the real succession (the succession that actually took place) goes the other way around, leading from a certain collection of atoms and simple molecules formerly present in the primordial atmosphere to that completely finished organism under investigation. This history still reaches a little more back in time, because we also have to trace back the formation of the mentioned simple molecules from their constituent atoms, which does not have happened in a direct way, straight from the constituent atoms to such a molecule.
This (total) history constitutes the actual history of that particular organism, i.e. the actual historical process of the formation of that organism from a collection of atoms, which can be considered as the ultimate constituents of that organism (See NOTE 1 ).
This actual history consists of a long series of successive atomic configurations (i.e. patterns of atoms, representing molecules and clusters of molecules). And because every non-random series of successive patterns must in principle be describable as a succession of states of a dynamical system, we have to do with a dynamical law, governing the atomic and molecular interactions, leading to the formation of that particular organism.
Is this law then the Essence of that organism?
No, it isn't.

Virtual non-contingent History of the organic Totality

The above described law, governing the actual historical sequence of atomic and molecular patterns, leading to the actual formation of that particular organism, is full of historical contingencies, at least up to the formation of the fertilized egg-cell of that particular organism in question (The history, from that egg-cell to the adult organism does probably not contain any contingent elements).
From the perspective of that organism its historical formation (from atoms) made a detour. Its formation -- as actually happened -- was very indirect, because the whole actual (evolutionary) process consists of trials and errors. Its history is littered with random mutations in the genetic systems, and selection took place on the basis of survival-value with respect to an ever changing environment. Consequently many aspects of this actual history are per accidens with respect to that organism's intrinsic atomic and molecular pattern, i.e. to that organism's actual structure (and its behavior that these structures make possible). And because Essence is something per se (with respect to that organism, i.e. its Essence is all that what necessarily belongs to it), and consequently opposed to anything that is per accidens, this law, governing the actual historical succession of states, leading from atoms to that organism's final structure, CANNOT as such be the Essence of that organism.
What then is the Essence?
The Essence of that particular organism is the dynamical law which remains after subtraction of ALL the contingent elements (i.e. all the per accidens elements) of the actual formational history. So we now consider a non-contingent shortcut of the actual history. This shortcut is a succession of states leading from a relevant collection of atoms, directly to the organism. This shortcut-history must of course be physically and chemically possible, that's why still a long succession of states is necessary (in contrast with, say, a process of the formation of the organism from a collection of atoms in ONE step (or only a few steps)). This shortcut would be actually obtained when we, as experimenters, are able to synthesize the organism directly from a collection of atoms. During this synthesizing in the laboratory there are no historical contingencies, at least not in a final synthesizing method. Of course we, experimenters, design such a synthesis, but that means that we only allow certain conditions to prevail. The atoms and molecules themselves do all the synthesizing. Such a set-up is and must be comparable with the experiments conducted by Stanley Miller, and by successive researchers inspired by him, to synthesize organic building-blocks (nucleotides, aminoacids, etc.) under certain conditions. The synthesizing of that organism could be interpreted as a form of artificial life, but in fact it is only a repetition of the formation of that organism, be it in a direct way. What we normally understand by the term " Artificial Life " is the creation by man of life on a different basis than life as we know it.
So, the dynamical law, inherent in the sequence of states, a sequence leading directly from atoms to that organism, is the Essence of that organism. It concerns a physical and chemical feasable sequence of material states, i.e. it is physically and chemically realistic, but it is stripped from any (historical) contingency, and ' constructed ' (in real experiments, or thought-experiments) NOT (from the) top-down, but (from the) bottom-up, i.e. by self-organizing principles.
Artificial Life in computers concerns a collection of relatively simple elements, as an initial condition (initial state), and a relatively complex Rule that organizes those elements into very complicated dynamic configurations or patterns. Complicated behavior is then emergent and implies functional behavior, like flocking, foraging, reproduction, etc. This Rule, if it generates a certain complete set of behaviors (and, possibly, morphologies), is -- and must be -- comparable with the aforementioned Ultimate Dynamical Law of one or another really existing organism. In both cases the Rule (the Law) must be relatively complex, and consists of much informaion, because the elements (atoms, pixels) themselves contain only a moderate amound of information (in a collection of atoms in fact many many potential dynamical laws inhere, but that means that there is in fact little information present, because information consists in constraints). It all amounds to their way of combining with each other into patterns, and information means a determination toward one or a few such patterns, and this information must reside in the one dynamical law that leads to that pattern. And that in turn means the ' activation ' of certain special properties of the atoms.

The dynamical law must imply a suitable environment for sustaining the ongoing dynamics, i.e. that law is not demanding such an environment but brings it with it. Such an environment consists of (physical) matter and energy or their (computer-)equivalents. Besides this the dynamical law demands (and now : not implies) some external conditions or a succession of external conditions like temperature and pressure (or their computer-equivalents) to prevail in order to let the law operate.
In real organisms their Essence is inherent in the relevant (physical) matter, i.e. in some properties of the relevant elements. In order to conceptually isolate these properties (from other not relevant properties) we describe their collection as a dynamical law. This dynamical law is accordingly equivalent to a set of properties and can be interpreted as FORM. The remaining properties, that do not enter the domain of the dynamical law (i.e. that do not go into the constitution of that law), can be interpreted as the substrate or MATTER relative to that FORM. As has been said, FORM and MATTER are in this context only relative, not absolute. This means that the dynamical law is not ' just formal ' and the substrate ' just material'.


Boundary Conditions

In the case of real entities, and we thereby focus on organisms, we have to do with physical laws. These laws are global, they are the general physical laws of Nature. They reign everywhere in the Cosmos. They leave open a vast range of possible patterns that can be generated. But locally we find constraints, which we can call boundary conditions. These boundary conditions are external, but local, and also internal but, again, local. Internal boundary conditions are related to certain structures of the (proto-)Totality (i.e. the being to be generated). External boundary conditions are related to local physical conditions like temperature, pressure, concentrations, etc.
The boundary conditions limit the (number of) possible patterns that could be generated, or, in other words, they harness the general physical laws : the boundary conditions are imposed on those laws and the result is the emergence of SPECIFIC RULES. Such a specific Rule is the dynamical law finally generating and maintaining the (organic) Totality.
With such and such boundary conditions, actually present, such an such dynamical law will (start to) operate. This law will be the actual law, which generates the organism very indirectly, and which is represented by the actual evolutionary (and ontogenetic) succession of material states. When we conceive of the virtual boundary conditions, in the context of the non-contingent generation of the organism, then those boundary conditions let the non-contingent Ultimate Dynamical Law (start) to operate. Such a dynamical law (the historical one and the conceptual one) is accordingly the result of constraining the general physical laws, limiting them to the emergence of ONE special process generating and maintaining the Totality. Those (local) boundary conditions, and by consequence the dynamical law, is NOT determined by the (global) physical laws, but those laws cannot be violated by the process resulting from harnessing those general physical laws (The general Laws of Nature).

Artificial Life

In computer-generated artificial life a Rule is set up that is interpreted as the implication of, or, if one wishes to say, the equivalent to, certain boundary conditions. The physical set-up of the computer cannot be violated by such a Rule. In artificial life experiments such a Rule, together with an initial condition forms the a-life system (In real systems the Rule and the initial condition are not separated). Taken in itself such a system is just a set of symbols 0 and 1 which are first of all interpeted as numbers. These numbers are replaced by other numbers according to the Rule. As such this system does not embody movement and space, and consequently no dynamic spatial patterns and processes. But the system is geared to a further interpretation. This further interpretation is realized with the display devices of the computer. And now we can see the movements and the spatial relationships, and consequently the dynamic patterns. With respect to life on Earth these patterns can only be a simulation(of those earthly patterns), not that Earthly life itself. But these computer-generated dynamic patterns can themselves be another kind of life, in this case a computational kind of life, different from life as we know it. It is important to realize that there is probably not a sharp transition between inorganic structures and living structures (living structures not necessarily of the Earthly type) -- the latter only demand a certain degree of complexity, that allows for self-maintenance and reproduction. So there is no sharp boundary between certain complex inorganic structures and certain relatively simple organic structures. The intermediate forms (between inorganic and organic) have existed on Earth long ago, but after a certain brand of life had finally established itself, those intermediate forms (existing as relicts) could not maintain themselves much longer anymore, they were eaten by the existing living forms (maybe we could still find them at the volcanic vents at the bottom of the oceans where new crust material is being formed). The same phenomenon we observe with respect to intermediate forms between organic species.
So for computer-generated ' life ' to be evaluated (as either living or non-living), we will not have at our disposal a complete list of criteria, only some general features, which must in some way be present in something which is living. So the debate on the status of the products of Artificial Life will until now be fruitless. We have to await further developments in Artificial Life research.
I spoke about tracing -- conceptually -- the direct (virtual) history of an organism (whose Essence we want to assess in general terms), a history without all the contingencies, contingencies that actually occurred during its real history. How must we -- conceptually -- trace back this direct history? It seems that we must first trace back, starting from the complete organism back to the fertilized egg-cell from which it had ontogenetically developed, and then starting our virtual history from this fertilized egg-cell to the set of yet unorganized atoms. This implies a sequence of feasable chemical reactions, leading from this unorganized set of atoms to the fertilized egg-cell. Of course in the actual history the final synthesis of that egg-cell was accomplished via a huge number of (genetic) generations of organisms in the context of the evolutionary development via Self-organization, Mutation and Natural Selection, but this history is, as I noted already several times, is, from the viewpoint of the final structure of that particular organism, full of contingencies, and trial and error processes. So the virtual history, leading from the unorganized set of atoms to the egg-cell (of that particular organism), and from there to the adult organism (in fact all the way to its death) is composed of a non-random sequence of states, and such a sequence can in principle be described by a dynamical law. This law then is the Ultimate Dynamical Law, or Essence, of that organism. It is a law, determining a sequence of states. How, i.e. by what structures, are those states represented? Are they the (virtual) phenotypes or the (virtual) genotypes? It seems to me that they are the latter. The virtual history consists of the direct (as direct as chemically possible) synthesis of the particular DNA molecules present in the fertilized eggcell, the eggcell that now gives rise to the particular organism of which we want to assess its Essence. The accompanying phenotypes of the (virtual) generations that lead to that organism are in fact necessary conditions for that particular DNA to develop, but they are only conditions, and thus per accidens with respect to the structure of the final product, the DNA of the organism under investigation. That particular DNA is chemmically formed out of other DNA molecules, and these ultimately out of (collections of) atoms.
That part of the history, leading from the fertilized egg-cell onwards is largely already determined by the structure of that cell, especially its DNA. So whether the organism dies prematurely or not, will not affect the very structure of the dynamical law (which is already determined by a (significant) part of the (total) succession-series of states) and is per accidens. The penultimate law moreover is already written in that cell. All kinds of perturbations during the organism's life are thus per accidens with respect to its Essence.

Aspects of Information, Life Reality, and Physics

Whithin the Artificial Life Community one occasionally ponders about an ontological interpretation of artificial life creations within computers.
At the Second A-life Conference, Rasmussen, a Danish physicist, involved in artificial life, distributed an intentionally provocative crib sheet titled " Aspects of Information, Life Reality, and Physics." It contained seven statements, 1, 2, 3, 4, 5, 6, and 7, logically connected. I will reiterate these statements accompanied with some comments between [ ]. (See LEVY, p. 145) :


Information and Life :

Accepting 1, 2, and 3 implies the possibility of life in a computer [ Because IF we accept that life is a process (2), and IF we accept that a universal computer -- every ordinary computer is a universal computer -- can emulate (i.e. reproduce) any process (1), and IF we accept that we indeed have some criteria (for distinguishing life from non-life) at our disposal (3), THEN we can have, and recognize, life in a computer.].

Life and Reality :

Reality and Physics :


As LEVY remarks, the final proposition provided a wondrous justification for creating artificial universes.
When we claim to have life in a computer (which we can then duplicate to several other computers), then someone could say that this " life " is confined to a computer environment, and cannot exist outside it. But life as we know it is also confined to earth-like planets and cannot exist outside such environments.


Images of Artificial Life

Self-reproduction is a very characteristic of Life (In any case the SELF is strongly developed in organic Totalities, culminating on Earth in Self-consciousness).
LANGTON devised a relatively simple system, Langton's Loop, based on a Cellular Automaton, a loop that can replicate itself. In this case the organism is a loop, and while it replicates it forms a colony of individuals.

Figure 3. Chris Langton's cellular automaton ' loops ' reproduce in the spirit of life. Beginning from a single organism, the loops form a colony. As the loops on the outer fringes reproduce, the inner loops -- blocked by their daughters -- can no longer produce offspring. These dead progenitors provide a base for future generations' expansion, much like the formation of a coral reef. This self-organizing behavior emerges spontaneously, from the bottom up -- a key characteristic of artificial life.
( From LEVY, 1992, Artificial Life )

It is asserted that the ability and actual occurrence of open-ended evolution is a necessary characteristic of life. Indeed, this is true for organisms in order to be able to maintain themselves in the long run (i.e. to prolong themselves in future generations) in an ever changing environment. This is what has happened (and still happens) on the planet Earth.
It is however not a necessary characteristic of a living organism as far as its actual constitution is concerned. Most creatures of Artificial Life are indeed just some sort of Totalities that execute some biological functions, which can be summarized by (the function of) LIVING, at least an analogue of it. These creatures are constituted in a certain way, exhibiting a stronger or lesser degree of a SELF. They live in a stable environment. That does not make them less real. An on the long run stable or changing environment is ontologically per accidens with respect to their individual SELF, their Essence (i.e. what they are in themselves). Differently said, an on the long run stable or changing environment is ontologically , i.e. metaphysically -- remember that Metaphysics is about the (beingness of) the individual being -- per accidens with respect to its Essence, it is however physically (biologically) crucial, and consequently per se with repect to the maintenance of the organismic species (and beyond). A living individual organism can cope with the prevailing environment because of its very constitution. But it cannot normally survive when the environment changes drastically within its life-time. This it must accomplish by evolution , thus at the level of the species, or beyond. So in order to maintain itself in the long run, life creates the phenomenon of succeeding generations that slightly change and adapt to the changing environment. So only for a supra-individual maintenance of life, is its ability for open-ended evolution (in which fitness-criteria are determined by the changing environment) a necessary characteristic. An individual organism, insofar as individual, does not have to anticipate changes in the environment way out in the future. But for the maintanance of life as such, and with it the supra-individual maintenance of that organism, it must do so (which means : be so).
When we want to create something living (in a general sense), then the creation of a Totality with a SELF, and the ability to be born, grow, reproduce and die, could already be sufficient.
But to simulate life as we know it, is a different matter. In this case it must have an ability for open-ended evolution, in order to maintain itself super-individually under changing environments.
Such a simulation is accomplished by some a-life researchers, especially by Thomas RAY. His simulator is called Tierra (which means Earth). The creatures in Tierra do indeed have the ability for open-ended evolution. Open-ended evolution means that the system is not task-oriented but environment-oriented. In earthly life this environment is mainly its biotic part, the pure physical part is relatively insignificant. The same applies to artificial systems : The environment of a digital organism to cope with, is the collection of the other organic individuals, and this environment constantly changes because those organisms change (for example by mutations of their code). The artificial evolution is thereby subjected to the laws of Logic rather than those of physics and chemistry.
TIERRA itself is a parallel virtual computer (wihin a ' real ' computer, this is done for reasons of security : the digital organisms cannot escape and multiply over computer networks). Each digital organism has its own Central Processing Unit (CPU). Such an organism consists of a line of code (i.e. series of instructions), which is its genotype. The execution of that code can be interpreted as its phenotype. Just as earthly life consumes energy (that ultimately comes from the sun) in order to organize its material, digital life can be seen as a consumer of the main computational device in the computer, the CPU, and this CPU is used for a certain period of time to organize the memory of the computer. In TIERRA CPU-time (concerning the CPU of the virtual computer) is the analogue of energy resources, and the computer memory corresponds to the spatial resource. The computer memory, the CPU and the computer's operating system correspond to the abiotic environment, the self-replicating programs (in assembler code) correspond to the organisms. So the Tierran organims drew their energy from the virtual computer's Central Processing Unit and used that energy to power the equivalent of their own energy centers, virtual CPU's assigned to each organism.
TIERRA starts, not from the inoculation of one or another random structure -- i.e. it is not going to simulate the origin of some kind of life -- but its environment was inoculated by a full-fledged digital organism, capable of mutation and replication. The system is designed in such a way that a competitive situation arises with respect to CPU-time and memory resources. When run, the system will display open-ended evolution resulting in more and more efficient codes (representing the organisms). Organisms with smaller instruction-sets normally use less resources and consequently are fitter. But their instruction-sets must enable the digital organism to replicate, so they cannot be too short. The first organism that was inoculated into the Tierran environment, the Ancestor, had 80 instructions. Soon mutants appeared with shorter instruction-sets using less resources. The system had discovered more successful strategies to cope with their environment. The proceedings were tracked by means of a dynamic bar chart, which identified the types of organisms and the degree to which they proliferated in the ' soup'.
Then something very strange happened.
Let me quote LEVY, p. 222 :

"In the lower regions of the screen a bar began pulsing [ the length of the bar represents the amound of organisms of the relevant type ]. It represented a creature of only forty-five instructions! With so sparse a genome [ its instruction-set ], a creature could not self-replicate on its own in TIERRA -- the process required a minimal number of instructions, probably, Ray thought, in the low sixties. Yet the bar representing the population of forty-five instructions soon matched the size of the bar representing the previous most populous creature. In fact, the two seemed to be engaged in a tug-of-war. As one pulsed outward, the other would shrink, and vice versa.
It was obvious what had occurred. A providential mutation had formed a successful parasite. Although the forty-five instruction organism did not contain all the instructions necessary for replication, it sought out a larger, complete organism as a host and borrowed the host's replication code. Because the parasite had fewer instructions to execute and occupied less CPU time, it had an advantage over complete creatures and proliferated quickly. But the population of parasites had an upper limit. If too successful, the parasites would decimate their hosts, on whom they depended for reproduction. The parasites would suffer periodic catastrophes as they drove out their hosts.
Meanwhile, any host mutations that made it more difficult for parasites to usurp the replication abilities were quickly rewarded. One mutation in particular proved cunningly effective in " immunizing " potential hosts -- extra instructions that, in effect, caused the organism to " hide " from the attacking parasite. Instead of the normal procedure of periodically posting its location in the computer memory, an immunized host would forgo this step. Parasites depended on seeing this information in the CPU registers, and, when they failed to find it, they would forget their own size and location. Unable to find their host, they could not reproduce again, and the host would be liberated. However, to compensate for its failure to note its size and location in memory, the host had to undergo a self-examination process after every step in order to restore its own " self-concept ". That particular function had a high energy cost -- it increased the organism size and required more CPU time -- but the gain in fitness more than compensated. So strains of immunized hosts emerged and virtually wiped out the forty-five-instruction parasites.
This by no means meant the end of parasitism. Although those first invaders were gone, their progeny had mutated into organisms adapted to this new twist in the environment. This new species of parasite had the ability to examine itself, so it could " remember " the information that the host caused it to forget. Once the parasite recalled that information it could feast on the host's replication code with impunity. Adding this function increased the length of the parasite and cost it vital CPU time, but, again, the trade-off was beneficial."

 1                               2 

 3                               4 

Figure 4. Evolutionary arms race between hosts and parasites in the TIERRA Synthetic Life program. Images were made using the Artificial Life Monitor (ALmond) program developed by Marc Cygnus. Each creature is represented by a bar. The color corresponds to genome size (e.g., red = 80, yellow = 45, blue = 79).
(1) Hosts, red, are very common. Parasites, yellow, have appeared but are still rare.
(2) Parasites have become very common. Hosts immune to parasites, blue, have appeared.
(3) Immune hosts now dominate memory, while parasites and susceptible hosts decline.
(4) The parasites will soon be driven to extinction.

( From COVENEY & HIGHFIELD, 1995, Frontiers of Complexity )


TIERRA also developed hyper-parasites, creatures that force other parasites to help them multiply, although they can reproduce in their own right. Also symbiotic -- cooperative -- behavior developed when each creature relied on at least one other to reproduce.
TIERRA shows the logical nature of Darwinian evolution. In itself it can be considered as another life-form, different from the way we know it. Just like life-as-we-know-it is confined to planets like Earth to support it, so Tierran life is confined to certain computer environments that support it.
For more details on TIERRA, See
LEVY, 1992, Artificial Life .
COVENEY & HIGHFIELD, 1995, Frontiers of Complexity .
EMMECHE, 1996, The Garden in the Machine .


In most Artificial Life systems the phenotype is behavior, we don't normally see life-like material structures. These latter we see in L-systems. These systems I treated of in a separate Essay, L-systems , because L-systems do not show behavior in the strict sense, but simulate one important biological function, namely morphological construction. The reader is referred to the Essay mentioned.
Below I show some more images of L-system outputs :

Plant05
Number of iterations 8
Axiom X
Rules :

X = F[+X][-X]FX
F = FF

Angle = 14

Output (Figure 5.) :

Figure 5.


Plant05
Number of iterations 10
Axiom X
Rules :

X = F[+X][-X]FX
F = FF

Angle = 14

Output (the effect of more iterations is clearly observable)(Figure 6.) :

Figure 6.


Plant08
Number of iterations 8
Axiom SLFFF
Rules :

S = [+++Z][---Z]TS
Z = +H[-Z]L
H = -Z[+H]L
T = TL
L = [-FFF][+FFF]F Angle = 20

Output (Figure 7.):

Figure 7.


Plant08
Number of iterations 10
Axiom SLFFF
Rules :

S = [+++Z][---Z]TS
Z = +H[-Z]L
H = -Z[+H]L
T = TL
L = [-FFF][+FFF]F Angle = 20

Output (Figure 8.):

Figure 8.


Beautiful images can be created with Laurens Lparser, a sophisticated L-system program and associated viewer (click on the small images):

Figure 9.
Here is a fractal fern type leaf on a nice reflecting background. Using the -i option with the Lparser generates a series of linked cylinder shapes.

Figure 10.
Here is a series of mutations based on a flower shape. Using polygons together with the basic F elements.

Figure 11.
These 4 images demonstrate the principle of mutation. The mother shape is at top left and the rest are her offspring. The children are clearly descendant from the basic shape but their differences increase with the number of mutations performed on the original Lsystem.


Another interesting a-life creation are Boids.
Here it concerns flock-behavior, as we see with some species of birds and fish. Such a flock consists of a hundred or more individuals and looks like a cloud that changes its form constantly. When it moves the flock seems to be governed by a central authority, a master-bird. But probably this is not so. The phenomenon can be simulated on a computer. In such a simulation there is no master bird. Each digital bird (which are called " boids ") follows some simple rules on the basis of local circumstances. The flock-behavior (which is a global phenomenon) then emerges from these simple rules. These rules are based on the following individual behaviors :
  1. Each individual tries to keep itself at a minimum distance from its neighbors. This results in a clumping force that keeps the flock together.
  2. Each individual tries to match its velocity with its neighbors, so that the boids in the flock would move at the same speed, and the flock is moving as one unified whole.
  3. Each individual moves away from those neighbors which are too close to it. This results in a separation force that prevents collisions.
When the rules are properly implemented in the computer, the boids -- represented by, say, little arrows -- form a flock that moves fluidly. Such an artificial flock even avoids obstacles, by splitting itself temporarily, as can be seen in the figure. We thus have a simulation of flocking behavior. But although the simulation is not the same as the thing simulated, what we see on the computerscreen is not only a simulation of flocking it IS flocking, be it of a digital kind.

Figure 12. A flock of Craig Reynolds's " boids " confront a cluster of pillars. Since their behavior is emergent, it was not clear even to Reynolds how the columns would affect the flight of the boids. As it happened, the flock was undaunted -- it temporarily split into two flocks and then reunited.
(From LEVY, 1992, Artificial life)


Are the creatures of Artificial Life really alife ?

In fact this question is -- in some sense at least -- irrelevant, because in the philosophy presented on this website aliveness is not considered a special ontological category. This is because probably there is no sharp boundary between complex inorganic beings and relatively simple organic beings. This in turn is because organic beings evolved from inorganic beings. The capacity for open-ended evolution -- which is often cited as an important criterium for something to be a living being -- is only relevant in complex ever-changing environments, and only serves the prolonged existence of life, it does not concern the livingness (of a being) itself.
One of the best digital candidates to date for claiming themselves as being alife (besides the digital creatures we have in systems like TIERRA) are computer viruses. A computer virus is a stretch of computer code that can copy itself into one or more larger ' host ' computer programs when it is activated. It will use those programs for its replication. Although until now no computer virus is able to undergo open-ended evolution, future versions will certainly accomplish this. They then should be considered as really alive, albeit a different form of life, a more or less disembodied form of life. Some researchers keep saying that they never can be really alife, and cite the well-known simulation problem in favor of their stance. Concerning this, LEVY, p. 337/8 ( Artificial Life) says the following :
" This is the telescoping of an obvious truism -- any simulation of something cannot be the same as the object it simulates -- to a general criticism of the methodology of simulation. Even those enthusiastic about the Weak Claim to a-life [ 'Weak Artificial Life ' only claims that its creations are simulations of living things or aspects of living things, they themselves are not claimed alive by that fact ], like Howard H. Pattee, have warned that its practitioners should resist the temptation to assume that fascinating results of computer experiments had relevance to the physical world. But in practice a-life frees itself from that dilemma by insisting that, although, indeed, a computer experiment is by no means equivalent to something it may be modeled on, it certainly is something. Maps are not the territory, but maps are indeed territories.
The methodology of a-life also shatters a related objection, that computer experiments, by their deterministic nature, can never attain the characteristics of true living systems. According to this argument, life cannot emerge from a mere execution of algorithms : in the natural world any number of chance occurrences contributed to the present biological complexity. But random events are indeed well integrated into artificial life. Von Neumann himself proposed, although he did not have the chance to design, a probabilistic version of his self-reproducing cellular automaton, which obviated the deterministic nature of the previous version. Many experiments in a-life, especially those that simulate evolution, include a step where random events make each iteration unpredictable. In any case, even in deterministic systems such as the CA game, Life [ ' Life is a famous two-dimensional Cellular Automaton that displays very complex behavior ], the cacophony of variables is sufficiently complex that the system can yield unbidden, or emergent, behavior."
This brings us back to the evaluation of the nature of computer viruses.
Biological viruses are just naked strands of nucleic acid, either RNA or DNA, surrounded by a sheath of protein. They need biological cells for their reproduction. Because of this they are somtimes regarded as non-living. But also among " real " organisms we encounter dependence (often very specific) of one organism on another, so for instance all the non-viral internal parasites which need other organisms' interior in order to complete their life-cycle. Maybe, as Andrew SCOTT (cited in LEVY) suggested LIFE is not just a collection of separate organisms, but a complete and integrated biosystem. In this way biological viruses are a part of life, just like we ourselves are just a part of life.
Do computer viruses satisfy the same, or equivalent, conditions that biological viruses satisfy? Are they " parts of life "? According to FARMER & BELIN a computer virus satisfies most, and potentially all, of the criteria for life. LEVY, p. 327/8 summarizes those criteria : From this it should be clear that the similarities between natural and artificial viruses are considerable.


What is Life any way?

In an article, published in the Artificial Life Bibliography of On-line Publications, Is Life as a Multiverse Phenomenon? by Claus EMMECHE, the discussion is about whether Artificial Life is possible. It concerns the question whether LIVING beings can in principle be created IN OTHER MEDIA, especially media like silicon-chips. To answer such a question one of course needs a valid definition of life. Such a definition could be presented in the form of a list of criteria for simething to be alive. EMMECHE cites such a list proposed by FARMER & BELIN (See for the exact reference the article mentioned):

Life (as defined by Farmer & Belin)

  1. A pattern in space-time (rather than a specific material object)
  2. Self-reproduction, in itself or in a related organism.
  3. Information-storage of a self-representation
  4. Metabolism that converts matter/energy
  5. Functional interactions with the environment
  6. Interdependence of parts within the organism (Preserve the identity of the organism. Ability to die)
  7. Stability under perturbations of the environment
  8. The ability to evolve
  9. Autonomy [added feature, C.E.]
This is a pretty broad definition of Life, but this -- which I have already emphasized above -- should be so, in order not to confine such a definition to the form of Life that we happen to encounter here on Earth. Furthermore the list must not be seen as something definitive, that rigourously and exclusively characterizes Life.

It could be interesting to accompany this list with the intentions and ideas of the new research program Artificial Life in its strongest and most ambitious form. Seven such ideas are cited by EMMECHE in his book The Garden in the Machine, 1996, p. 17-20 :

  1. The biology of the possible. Artificial Life does not concern the special wet and carbon-based life as we know it here on earth, which is the subject of experimental biology. Artificial Life deals with life as it could be. Since biology is only based on one example, life on earth, it is too empirically limited to help create truly general theories. Here Artificial Life is a clear and quite decisive supplement. It is not certain that we appear as we do simply because of previously existing earthly materials and the accidental evolutionary sequence. Evolution could rest on much more general organizational laws, but these are laws that we simply do not know yet. Biology today is only the biology of actual life. It must become a biology of any possible life-forms.
  2. Synthetic method. Where traditional biological research has placed emphasis on analyzing living beings and explaining them in terms of their smallest parts, the artificial-life perspective attempts to synthesize life-resembling processes or behavior in computers or other media.
  3. Real (artificial) life. Artificial Life is the study of humanly created systems that exhibit behavior characteristic of natural, living systems. However, what is it in artificial life that is artificial, in the sense of false, unnatural, or humanly created? That which is " artificial " about life in silico -- all gadgets and information structures in the form of machines, models, and constructed " organisms" -- is not the behavior as such. The behavior, the generalized process, is just as genuine as the behavior exhibited by real-life organisms. No, the " artificial " of artificial life rests solely in the components (like the silicon chips, formulas, computational rules, and the like) of which it consists. These are designed by us [ This is not quite correct anymore. In many a-life systems, like TIERRA, open-ended evolution occurs which is by definition beyond our control. New, no man-made formulas can appear. They evolve from within the system. Only the system as such -- we could say its initial state -- is designed by humans. ]. The behavior, however, is produced by the artificial life itself.
  4. All life is form. Neither actual nor possible life is determined by the matter of which it is constructed. Life is a process, and it is the form of this process, not the matter, that is the essence of life. One can therefore ignore the material and instead abstract from it the logic that governs the process, taking it out of the concrete material form of the life we know. Hence, one can thus achieve the same logic in another material " clothing " or substratum. Life is fundamentally independent of the medium.
  5. Bottom-up construction. The synthesis of artificial life takes place best via a principle of computer-based information processing called " bottom-up programming ": at the bottom many small units and a few simple rules for their internal, purely local interaction are defined (This is the real programming). From this interaction arises the coherent " global " behavior at the general level, behavior not previously programmed according to specific rules. Bottom-up programming corresponds to the fact that our proteins are " programmed " relatively explicitly by DNA, but there is no gene that directly specifies the form of the face or the number of fingers. This kind of programming contrasts with the dominant programming principle within [ classical ] artificial intelligence (AI). Here one attempts to construct intelligent machines by means of programs made from the top down : the total behavior is programmed a priori by dividing it into strictly defined subsequences of behavior, which are in turn divided into precise subroutines, smaller subsubroutines, etc., all the way down to the program's own machine code.
    The bottom-up method in artificial life imitates or simulates processes in nature that organize themselves. We might also call these processes " simulated self-organization".
  6. Parallel processing. While information processing in a classical comoputer takes place sequentially -- similar " one-logical-step-at-time " thinking is also found in classical AI -- the principle for information procesing in artificial life is based on a massive parallelism that occurs in real life. In real life the brain's nerve cells work alongside each other without waiting for their neighbor to " finish his work ". In a flock of birds it is the simultaneity of the many birds' individual small changes in the direction of flight that gives the flock its dynamic character. Artificial neural networks are a typical example of parallel information processing and, hence, a kind of artificial life (That the parallelism of neural network models can be simulated on sequential computers is simply a stroke of luck and says nothing about the computational principle itself).
  7. Allowance for emergence. The essential feature of artificial life is that it is not predesigned in the same trivial sense as one designs a car or a robot [ Today in modern robotics one is also programming them in the bottom-up way. ]. The most interesting examples of artificial life exhibit " emergent behavior". The word " emergence " is used to designate the fascinating whole that is created when many semisimple units interact with each other in a complex (nonlinear) fashion. In computational terms, it is the bottom-up method that allows for the emergence of new, unforseen phenomena on the superordinate level, a phenomenon that is crucial for living systems.

All this kind of research is directed to an understanding of living systems in general. Its procedures and results directly evokes fundamental questions, involving ontology, origin of life, consciousness, the nature of computation, security, and even morality. As any other science it should be done with caution, theoretically as well as practically. The existence of computer viruses already teaches us about those two aspects. Although the new science is now well under way it is just the beginning of a series of staggering creations that cannot be denied the status of aliveness anymore.

back to homepage

************