Ashby's idea is that a mechanism which can alter its internal configurations can do a random search for a configuration which achieves some desired "goal." The objective of an organism is to maintain a vital quantity in a stable equilibrium, like body temperature, by a complex set of mechanisms such as sweating and shivering. For a machine, the goal is to keep the values of certain "essential variables" within a desired range, and when these fall outside that range, to randomly vary the non-essential variables it can control until the values of the essential variables are restored. He called this mechanism of trial and error a "functional circuit" because it responded to its own success or failure, but later recognized it to be identical to the concept of feedback.

The theory thus offers a way to explain learning and biological adaptation, in terms of a single type of physical mechanism. In 1947, Ashby built an analog computer to demonstrate his idea. Called the Homeostat, it consisted of four interconnected units which sought to establish a pattern of electrical currents between them such that the whole ensemble would resist various external disturbances. The model of a goal-directed search which it embodied has become central in Artificial Intelligence.

Many significant technical developments have been inspired by cybernetics. Among these are genetic algorithms and evolutionary programming. Genetic algorithms were first devised by John Holland (1975), and attempt to simulate the self-organizing properties of biological evolution. They do this by dividing the possible solutions to a problem into pieces called alleles, which are analogous to the pieces which make up biological genes. Various combinations of alleles are combined into hypothetical solutions which are then tested against one another in a fashion analogous to Darwinian natural selection by competition. The "fittest" solutions are then recombined into a new population with minor mutations in a process analogous to sexual reproduction. The processes of recombination and selection are repeated many times until a near-optimal solution is found. This technique is often used as a method of non-linear optimization in computer science and engineering.

. Cybernetic concepts and theories continue on, reconstituted in various guises, including the fields of self-organizing systems, dynamical systems, complex/chaotic/non-linear systems, communications theory, operations research, cognitive science, Artificial Intelligence, artificial life, Robotics, Human-Computer Interaction, multi-agent systems and artificial neural networks.  Cybernetics slowly dissolved as a coherent scientific field during the 1970's, though its influence is still broadly felt