E. A. Yankovskaya – Ph.D. (Eng.), Assistant Professor, Department of Social Sciences and Technologies of National University of Science and Technology «MISIS». E-mail: email@example.com
Connectionism was based on trying to reproduce the key principles of the structure and functioning of the real nervous system. It was necessary to create an artificial cognitive systems with intellectual abilities. W. McCulloch and W. Pitts contended that the basic principle of functioning of the nervous system is the interaction of simple elements arranged in complex networks. According to they approach the basic property of neuron to create discrete electric impulse is enough to create a formalized model of a neuron.
This property of the neuron becomes a source of ideas that basic level of human thinking is very simple. One of the modern theorists of artificial intelligence D. Hawkins notes that it's also necessary to take into account variability and feedbacks in nervous system. These aspects have been partially taken into account, but cognitive actions still are attributed to the activity of the simple elements whose connections allows logical operations.
It's obvious that the formal neuron is too reductionist comparing with a real neuron. The simple scheme of nervous activities doesn't take into account real complexity of synaptic relations. It was also suggested that the thickness of the axon can play a crucial role in information processing because it determines the ability of axon to pass signal.
H. Dreyfus emphasizes that basing nerve impulses on the principle of "all or nothing " does not mean that all cognitive operations have such digital form. J. von Neumann pointed out that in addition to digital processes in the neuron in the nervous system, there are also processes based on different principles of operation. For example blood chemistry and other humoral fluids plays an important role in regulation of nervous activity. Thus treating nervous system independent of the other systems of the body it is impossible to adequately describe the principles of operation including performing "higher" intelligent functions.
Developing formal neuron W. McCulloch drew attention to the special properties of neural networks that do not allow the algorithm to create the nervous system and can be expressed in heterarchical principle of its action. Considering the nature of the non-algorithmic thinking W. McCulloch pointed intransitive character of cognitive processes. According to heterarchical organization of cognitive processes an element of a complex system cannot be too simple. Thus, the complexity of the cognitive involves the complexity of the elements that constitute the system and the complexity of their interactions. Such complexity does not allow to formalize the nervous system so that the artificial system completely reproduces the actual cognitive processes.