Artificial Life

One could answer like Langton: that it is about systems, built by man, which present characteristic behaviors of natural living. But then, what are these characteristics? How much of it is necessary to speak about artificial life? This definition is, as be seen, much too vague. It has however the merit to give us a base of reflection.

When Langton speaks about characteristics of natural livings, it is doubtless that he defines life as a group of properties. Now, it is impossible to make a list, in a strict way, of all the properties of living. To remedy this, Maturana and Varéla proposed as from 1979 another much more relevant definition, nevertheless still too much underestimated by scientists. It is a question of conceiving a being as an autopoeitic organization. It characterizes the fact that a being is a network of dynamic transformations manufacturing its own components (metabolism), and that builds a topological barrier (membrane), which in turn, is the necessary condition for the functioning in unity of the network of transformations that engendered it. Which means that the livings are constantly self-producing.

The living beings are thus characterized by their autopoeitic organization. They differ from each other by their structure, but they are identical in their organization. Maturana and Varéla wanted their definition of the living to be a scientific one: " If we can supply the list of characteristics of a being, why not propose a system which generates all the phenomena characteristic of the living? ". The fact that an autopoeitic entity has all these characteristics appears evident in the light of what one knows on interdependence between the metabolism and cellular structure.

Evidently, not only the livings have organization. It can be found in all that can be considered as a system. What characterize however beings, is that their organization is such as their only product is themselves, and the absence of separation between the producer and the product. The being and making of an autopoeitic unit are inseparable, and it is there their particular mode of organization.

So, if we can define life otherwise that by a list of characteristics (list that we can not formally establish); we cannot define artificial life as Langton did. We cannot say that artificial life constitutes of systems presenting characteristic behaviors of natural living systems, but built by the man.

What alternatives do we have? It would be very interesting and logical to map the autopoeitic definition of natural life on a definition of artificial life. Effectively artificial life seems to appear under a great number of different forms. But before going any further, it is important to visualize well the various domains that could be englobed by artificial life. It is still impossible today to define the limits between what is of artificial life and what is not for the simple reason that there is no definition relevant and enough precise.

During the third Santa Fe conference, Claus Emmeche evoked the various forms of artificial life. He classified them in two main groups: trivial versions and non-trivial versions. For each of these groups, we can find a weak or a strong artificial life. Strong artificial life corresponds to a search for the creation of life, whereas the weak approach corresponds to an imitation of the natural life.

The trivial group includes:

Simulations and models. It corresponds to computer simulations based on mathematical models. Conceptual or physics of biological systems. It cannot end in real forms of lives. Modified organisms. It is about real beings, modified by man by means of genetic manipulations for example.

Whereas the non-trivial aspect corresponds:

Computational systems. These realizations are by definition strictly in computers. Their aim is the creation of virtual organisms considered as alive (at least by their author). It is the most controversial strong approach of artificial life.
The evolvable robotics. It regroups autonomous and evolvable robots.
Experiments relating to biochemistry. This last category encloses the synthesis of pre-biotic processes and primitive organisms thanks to physico - chemical in vitro experiments.

It seems so that there are several forms of artificial life; and we see that they differ enormously. It would be maybe more convenient not to group them under the same term, but to create new names for each of these categories.

One could again wonder on borderline case, of prosthesis for instance. Do we have an artificial life, if we have an artificial heart or even for the case of a transplanted heart patient? This brings us to the question about what is a prosthesis? Are eyeglasses prosthesis? If yes, then we can say that when we shoe on skis, it is also about prostheses; everything as could be finally, in the same way, a car.

So, the definition of the artificial life is very problematic: difficult to decide on the border between life and artificial life (problem of prosthesis); and difficult also to decide on the limit between artificial life and a computer program with artificial intelligence for instance.

One may choose to consider here as artificial life what Claus Emmeche called "strong artificial life" which corresponded to creation of life, that is: computational systems, the evolvable robotics and experiments relating to biochemistry.

Artificial Intelligence (AI) and Artificial Life (AL)

Cybernetics' paradigm saw its technological realization through two fields whose stakes and defining concepts differ: Artificial Intelligence and Artificial Life. AI, the former of the two, dated from the first developments in cybernetic. It concerns the conception of machines able to feign human cognition. As we shall see, AL, more recent, diverges from AI on some points and joins it on others. The field of study covered by AL is vaster. It investigates characteristics of the living in general.

So as to have some landmarks, we shall explain here the main stages of development of these concepts. We will also consider their technological applications.

Artificial Intelligence: human cognition as a reference

Artificial Intelligence arises directly from concepts on cognition. That is, it envisages intellectual functioning under a logical-deductive angle, and thus considers that the cognitive action is made through a manipulation of elementary symbols. Through cognition, AI establishes a fundamental analogy between the intellectual functioning and that of the computer. That is why some authors refer the term computational to indicate this theoretical orientation of the cognitive sciences of which AI is the most evident technological development.

IA has met numerous successes in extremely varied applications and particularly in designing expert systems. However, in the course of time while it acquired a scientific status in its own right, it departed from its first objective: create artificial systems capable of reproducing the intelligent behavior of biological systems.

Initially, the ambitions of AI were particularly high. One hoped to be able to create machines capable of dealing with very general problems such as the recognition of natural language, the translation of natural languages or the universal resolution of problems. It was in question to model the know-how of experts. But, as underline by Fransisco J. Varela:

" as these attempts became more modest, it became clearer that the deepest and most fundamental intelligence is that of the baby who acquires language... "
[VARELA (Fransisco). - Invitation aux sciences cognitives (Seuil, Paris, 1996), p. 56.]

The acquisition of natural language by a machine is one of the essential stakes in AI and indicates clearly its theoretical orientations.
However, the possibility that a computer can "learn to speak " stays completely hypothetical and the idea sustaining this possibility is that the machine can and must imitate the living. Another point of view states that while the computer has remarkable capacities to resolve some classes of problems, it is completely quite impossible to endow it with the perceptive and cognitive characteristics of the living. Given the technological and financial stakes, these assertions are object of numerous debates. We shall consider these in the following section.
Let us just note here that in spite of the progress in AI, the dream of a universal machine is far from being realized.

We observe so a redirection from general to more particular considerations. The ambition to create a universal machine fades and is replaced by the implementation of AI to create machines resolving a class of problems. It should be noted that it is a redirection that can be observed presently in numerous fields.

Artificial Life: approaching biological functioning

Contrarily to what one could believe, AL had not so general ambitions as AI in its beginnings and thus it avoided the great pitfall of universality. It is true that AL is a more recent field and that it benefits, consequently, of the experience of AI. The current AL program seems well enough defined and has more diversified theoretical and conceptual bases. Artificial Life exploits concepts stemming from cognition and connectionist that make up cognitive sciences. It proposes solutions that sometime stress on the symbolic approach, sometime on the auto-organizational approach, or builds the models from one combination of the two methodologies.

The program of this new science is defined by Christopher Langton as:

"the study of manmade systems displaying behaviors characteristic of living systems".

[Quoted by FERNÁNDEZ (Julio) and MORENO (Álvaro). - Op. cit., p. 23.]

Actually, AL often attempts to model behaviors or primordial mechanisms of living systems such as auto-regulation, breathing, predating or reproduction.

AL dedicates a large part of its efforts in realizing of virtual creatures, that is, to simulate living beings with the help of formal computer languages. Another branch of this field is concerned with conception of material realizations in the form of autonomous robots (using partially the concepts of AI), or from the materials of living beings.

We can hence distinguish two types of work stemming from AL. On one hand, simulations that use computers exclusively. Admittedly, in this case, any system can be modeled on computer. That is, a formal system is susceptible to represent satisfactorily a physical system. On the other hand, realizations where the concrete and material character of the system is primordial. It is considered here that the physical aspect of the system cannot be reduced to a symbolic representation.

As to knowing whether AL is susceptible to create life, let us just refer to Christopher Langton:

" [... ] It was 3 o'clock in the morning. Suddenly, I felt a presence in the room. But there was no one. I realized that there had been something going on the screen and that this thing had struck my subconscious. I had reacted to something living. Of course, the computer was not physically living, but it was able of some kind of behavior to which had reacted some primitive thing in me, as if there was life. " [LANGTON (Christopher). - " Et si les informaticiens découvraient la vie ? ", Courrier International (n° 360, 25 sept. - 1° oct. 1997) (initially published by Facts, Zurich)]

Christopher Langton recalls this experience as one of the most striking that he had lived. Maybe it is for this reason that he is trying to define and to understand what life is while preserving himself from any prejudice.

If we understand well how AI joins in our view, as a technological outcome of the logical - mathematical spirit introduced by Aristotle, the evocation of AL can seem misplaced. It is true that AL is more concerned by "reflex" behavior than logical reasoning or cognitive acts. AL deserves however all our attention, and this, for two reasons. By the concepts it carries, AL represents an interesting alternative to the strictly computational approach. Then, it is much evident that computational treatment of information implies a manipulation of symbols whose meaning is unconsidered by the system. However, on wondering about emergence, by trying to approach the most possible the essential mechanisms of living, AL opens a major debate: How the living can bear sense and from what mechanisms? How can an artificial system have equivalent characteristics? From then on, what relationships are established with the observer of the system?

A Definition of Artificial Life

Artificial life was defined by Langton in 1989 as the study of systems built by man having characteristic behaviours like living systems. However, we will try to go further in this definition by giving details. The definition of artificial life is tightly linked to the one of life, which definition is a problem in itself. It is thus particularly difficult to define artificial life and especially its limits. Where is/stands the frontier between life and artificial life? What are the differences between a machine or a program, and an artificial form of life? This problem is developed in "Artificial Life: a complex problem".

The definition we give here comes from J. Doyne Farmer and Alleta d'A. Belin's definition of life. Here are the basic properties that characterise any artificial form of life:

1. Human beings contributed to the conception of any artificial way of life;
2. An artificial form of life is autonomous;
3. An artificial form of life is in interaction with its environment;
4. There is emergence of behaviours in an artificial form of life;

The three next properties are not essential but remains, nevertheless, very present:

5. An artificial form of life can reproduce;
6. An artificial form of life has the ability of adaptation;
7. An artificial form of life is not a unit.

Contrarily to a biologically living system, an artificial life system can be divided into several parts that might not be physically linked: for instance, a computer can pilot a robot thanks to waves. Even inside a computer, nothing guarantees that all the bytes of this system are gathered.