Artificial life is a cross-disciplinary field of research devoted to the study and creation of lifelike structures in various media (computational, biochemical, mechanical, or combinations of these). A central aim is to model and even realize emergent properties of life, such as self-reproduction, growth, development, evolution, learning, and adaptive behavior. Researchers of artificial life also hope to gain general insights about self-organizing systems, and to use the approaches and principles in technology development.

Evolution of research

The historical and theoretical roots of the field are manifold. These roots include:

• early attempts to imitate the behavior of humans and animals by the invention of mechanical automata in the sixteenth century;
• cybernetics as the study of general principles of informational control in machines and animals;
• computer science as theory and the idea of abstract equivalence between various ways to express the notion of computation, including physical instantiations of systems performing computations;
• John von Neumann's so-called self-reproducing Cellular Automata;
• computer science as a set of technical practices and computational architectures;
• artificial intelligence (AI)
• robotics;
• philosophy and system science notions of levels of organization, hierarchies, and emergence of new properties;
• non-linear science, such as the physics of complex systems and chaos theory; theoretical biology, including abstract theories of life processes; and
• evolutionary biology.

Despite the field's long history, the first international conference for artificial life was not held until 1987. The conference was organized by the computer scientist C. G. Langton, who sketched a future synthesis of the field's various roots and formulated important elements of a research program.

In the first five years after 1987, the research went through an exploratory phase in which it was not always clear by what criteria one could evaluate individual contributions, and some biologists were puzzled about what could falsify a specific piece of research. Later the field stabilized into clusters of research areas, each with it own models, questions, and works in progress. As in artificial intelligence research, some areas of artificial life research are mainly motivated by the attempt to develop more efficient technological applications by using biologic inspired principles. Examples of such applications include modeling architectures to simulate complex adaptive systems, as in traffic planning, and biologically inspired immune systems for computers. Other areas of research are driven by theoretical questions about the nature of emergence, the origin of life, and forms of self-organization, growth, and complexity.

The media of artificial life

Artificial life may be labeled software, hardware, or wetware, depending on the type of media researchers work with.

Software. Software artificial life is rooted in computer science and represents the idea that life is characterized by form, or forms of organization, rather than by its constituent material. Thus, "life" may be realized in some form (or media) other than carbon chemistry, such as in a computer's central processing unit, or in a network of computers, or as computer viruses spreading through the Internet. One can build a virtual ecosystem and let small component programs represent species of prey and predator organisms competing or cooperating for resources like food.

The difference between this type of artificial life and ordinary scientific use of computer simulations is that, with the latter, the researcher attempts to create a model of a real biological system (e.g., fish populations of the Atlantic Ocean) and to base the description upon real data and established biologic principles. The researcher tries to validate the model to make sure that it represents aspects of the real world. Conversely, an artificial life model represents biology in a more abstract sense; it is not a real system, but a virtual one, constructed for a specific purpose, such as investigating the efficiency of an evolutionary process of a Lamarckian type (based upon the inheritance of acquired characters) as opposed to Darwinian evolution (based upon natural selection among randomly produced variants). Such a biologic system may not exist anywhere in the real universe. As Langton emphasized, artificial life investigates "the biology of the possible" to remedy one of the inadequacies of traditional biology, which is bound to investigate how life actually evolved on Earth, but cannot describe the borders between possible and impossible forms of biologic processes. For example, an artificial life system might be used to determine whether it is only by historical accident that organisms on Earth have the universal genetic code that they have, or whether the code could have been different.

It has been much debated whether virtual life in computers is nothing but a model on a higher level of abstraction, or whether it is a form of genuine life, as some artificial life researchers maintain. In its computational version, this claim implies a form of Platonism whereby life is regarded as a radically medium-independent form of existence similar to futuristic scenarios of disembodied forms of cognition and AI that may be downloaded to robots. In this debate, classical philosophical issues about dualism, monism, materialism, and the nature of information are at stake, and there is no clear-cut demarcation between science, metaphysics, and issues of religion and ethics. If it really is possible to create genuine life "from scratch" in other media, the ethical concerns related to this research are intensified: In what sense can the human community be said to be in charge of creating life de novo by non-natural means?

Hardware. Hardware artificial life refers to small animal-like robots, usually called animats, that researchers build and use to study the design principles of autonomous systems or agents. The functionality of an agent (a collection of modules, each with its own domain of interaction or competence) is an emergent property of the intensive interaction of the system with its dynamic environment. The modules operate quasi-autonomously and are solely responsible for the sensing, modeling, computing or reasoning, and motor control that is necessary to achieve their specific competence. Direct coupling of perception to action is facilitated by the use of reasoning methods, which operate on representations that are close to the information of the sensors.

This approach states that to build a system that is intelligent it is necessary to have its representations grounded in the physical world. Representations do not need to be explicit and stable, but must be situated and "embodied." The robots are thus situated in a world; they do not deal with abstract descriptions, but with the environment that directly influences the behavior of the system. In addition, the robots have "bodies" and experience the world directly, so that their actions have an immediate feedback upon the robot's own sensations. Computer-simulated robots, on the other hand, may be "situated" in a virtual environment, but they are not embodied. Hardware artificial life has many industrial and military technological applications.

Wetware. Wetware artificial life comes closest to real biology. The scientific approach involves conducting experiments with populations of real organic macromolecules (combined in a liquid medium) in order to study their emergent self-organizing properties. An example is the artificial evolution of ribonucleic acid molecules (RNA) with specific catalytic properties. (This research may be useful in a medical context or may help shed light on the origin of life on Earth.) Research into RNA and similar scientific programs, however, often take place in the areas of molecular biology, biochemistry and combinatorial chemistry, and other carbon-based chemistries. Such wetware research does not necessarily have a commitment to the idea, often assumed by researchers in software artificial life, that life is a composed of medium-in-dependent forms of existence.

Thus wetware artificial life is concerned with the study of self-organizing principles in "real chemistries." In theoretical biology, autopoiesis is a term for the specific kind of self-maintenance produced by networks of components producing their own components and the boundaries of the network in processes that resemble organizationally closed loops. Such systems have been created artificially by chemical components not known in living organisms.

Conclusion

Questions of theology are rarely discussed in artificial life research, but the very idea of a human researcher "playing God" by creating a virtual universe for doing experiments (in the computer or the test tube) with the laws of growth, development, and evolution shows that some motivation for scientific research may still be implicitly connected to religious metaphors and modes of thought.

See also ARTIFICIAL INTELLIGENCE; CYBERNETICS; CYBORG; INFORMATION TECHNOLOGY; PLAYING GOD; ROBOTICS; TECHNOLOGY

CLAUS EMMECHE