Already for about 40 years from today, natural scientists more and more often reach the limits of a reductionist world view. The top-down causality of principles, which dominated scientific methods for centuries, lacks explanation for complex phenomenons, for example, the spontaneous self-organisation of animal herds. Even since the 1940ies, mathematicians, physicists, biologists and computer scientists have developed new models to explain the complexity of our world, however most of them were lacking cross-disciplinary exchange to be able to work on a bigger picture.

In 1987, a first conference in New Mexico joined avantgarde scientists from various fields and coined the name of a new field of research: Artificial Life.

Around the same time, the research on Artificial Intelligence had reached a zero point. After ambitious prognoses for the develoment of human-like machine intelligence in the early years of computer science (1940ies to 60ies), years of relatively fruitless research followed.

Useful AI

In 1986, computer scientist Rodney Brooks noted that the question how AI could be “useful” should be as important a research goal as the big philosophical questions about Artificial Intelligence (cf. Rodney Brooks, Elephants Don´t Play Chess).

He developed a new method for a “robust” and applicable AI and robotics: particular tasks of an artificial intelligence system are solved as independent modules, e.g. sensoric or motoric tasks, and then linked up in a control unit – a method called Subsumption Architecture. Brooks was certain that a more complex intelligence might emerge from a system that consist of more simple and limited intelligence units. Until today, Subsumption Architecture is the basic structure for Robots and applied AI, not only in those helpful little creatures whose intelligence is limited to cleaning the floor of our swimming pools, but in more powerful systems like computer-aided medical diagnosis based on image analysis.

Complex adaptive systems that we all know are for example the evolution, the weather, the stock exchange, or the immune system. From a mathematic point of view, an Alife simulation is a dynamic system of various differential- and difference equations. In computer science, some complex adaptive system are called Genetic or Evolutive Algorithms.

The phenomenon of complexity in a system based on simple rules is called Emergence. Examples would be the forming of chaotic, periodic or homogenous patterns (like the patterns of butterflies, a leopard, or a snail house), or the self-organisation of animals in herds or swarms. The emerging complex behaviour in a system of individuals is known as Collective Intelligence.

The elements in a software simulation are often called Agents or Particles, the simulation itself Agent-based System or Particle System. AI research works with agent-based systems or Multi Agent Systems, too. While in AL, the elements are often passive, agents in Multi Agent System usually have a limited autonomy for decision-making.

Example: Swarming Behaviour

Many bird and fish species move in flocks. Likely do mammals in a herd always seem to know their place. This self-organisation is spontaneous and instinctive; there is no central navigation or a leader supervising the swarm. The behaviour appears chaotic and complex, yet harmonic and incomprehensibly organised.
At the same time, swarming behaviour is one of the most disctinct examples of how complex behaviour can be simulated with a system of simple rules. Computer graphics engineer Craig Reynolds has first imitated swarming behaviour in a computer simulation in 1986, based on only three corresponding principles:

Separation:
Steer to avoid crowding local flockmates
Alignment:
Steer towards the average heading of local flockmates
Cohesion:
Steer to move toward the average position of local flockmates

Human individuals in large groups behave like swarms aswell. Therefore, computer simulations based on flocking algorithms are also used to simulate situations like the evacuation of buildings or the flow of road traffic.

Example: Weather Reports & Flow Fields

Agent-based systems are used for weather forecasts. The animations shown in TV weather reports are based on a particle system in which air masses are subsituted by a large number of elements, visualising the paths of air currents and pressure areas over land and water. Each element is displayed as an arrow, its original direction influenced by the underlying topography, and of course all the dynamic factors involved, such as temperature, humidity, etc., in relation to time. This allows to simulate the develoment of cloud fields in the near future.