Anticipation, Induction, and Learning

A system is considered anticipatory if it has the ability to foresee the consequences of an event and act in a way it is adapted for. In order to make such judgments anticipatory systems must possess some kind of description of their surroundings, which is used in the calculation of an appropriate action. In many cases it is sufficient to have an algorithmic description to follow and some anticipatory systems do choose their actions in a completely algorithmic way.

A more developed anticipatory behavior is displayed by systems, which not only possess a description but also a model of the surroundings. Those systems have an intrinsic conception of their surroundings, which they are able to reason about. This kind of anticipation is called model-based contrary to the description-based behavior, which characterizes those systems that slavishly follolw algorithmic rules.

In order to take advantage of model-based behavior it is necessary to be able to properly describe the surroundings in terms of how they are perceived. Such description processes are inductive and not recursively describable. That a system can perceive and describe its own surroundings means further that it has a learning capability. Learning is the process of making order out of disorder and this is precisely the most distinguish quality of inductive inference. Genuine learning without inductive capability is impossible.

The implication of this is that systems that have a model of the surroundings are not possible to implement on computers nor can computers be leaming devices contrary to what is believed in the area of machine learning.