Adaptive systems can be distinguished according to the number of traverses that they posses. If n indicates the number of traverses, then we can refer to different types of systems using Tn notation:

A T0-system does not have any traverses. An example is a book.

A T1-system has one traverse. An example is a heating system with a thermostat, but also a PC computer without any artificial intelligence algorithms implemented. Also, a reflex is a T1-system, much like any other simple homeostatic mechanism.

Also, a neural network without any learning mechanisms implemented is a T1-system. A neural network remains a T1-system irrespective of its complexity, number of layers, or a presence of reentrant connections. A mere increase in the complexity of a system does not make it more adaptive.

A T2-system has two traverses. If we add to a neural network a supervised learning mechanism, it becomes a T2-system. Usually, machine learning systems can be classified as T2.

T3-systems are arguably found only in biology (for now). Behaving animals are T3-systems.

A T4-system is created by adding one more traverse to a behaving animal. In nature, this additional traverse corresponds to the process of evolution by natural selection. Therefore, while an individual is a T3-system, the entire evolving species is a T4-system.

Back to: Key concepts of practopoietic theory