Excerpts from Norbert Wiener’s book “Cybernetics and Society, the human use of human beings”, Chapter “Rigidity and learning. Two models of communication behavior”. (The book was first published in 1949, in the USA).
“In the termite community, every worker performs his own mission. There may also be a separate caste of soldiers. Certain highly specialized individuals perform the functions of the King and the Queen. If man were to adopt this community as a model, he would live in a fascist regime with the ideal of each individual being defined from birth for special employment. The leaders are forever leaders, the soldiers forever soldiers, the peasant is nothing more than a peasant, and the worker is condemned to remain a worker.
The thesis of this chapter is that the fascist’s ambition for a human regime based on the termite model is the result of a profound misunderstanding of both termites and human nature. I want to emphasize that the natural development of the insect itself determines it to be an essentially stupid being, without the ability to learn, imprisoned in a mold that cannot be seriously modified. Also, I want to show how these physiological conditions make it a cheap mass-produced product, which has no more individual value than a paper plate that we throw away after use. On the other hand, I want to show that man, capable of tremendous learning and study, who can fill half his life, is by nature equipped with this ability, while the termite is not.
[…] Feedback is a method of controlling a system by re-injecting the results of its previous behavior back into it. If these results are simply used as numerical data to evaluate and adjust the system, then we have simple feedback, which is what control engineers use. However, if the information returned from the behavior is capable of changing the overall method and model of the behavior, then we have a process that can clearly be called learning.Another example of the learning process appears in relation to the problem of designing prediction machines. At the beginning of the Second World War, the relatively low efficiency of anti-aircraft weapons created the need to introduce devices that would follow the position of the aircraft, calculate its distance, determine the time interval it would take for the bullet to reach it, and calculate exactly where it would be at the end of the time interval.
[…] The adaptation of the general aiming plan and the adjustment of the weapon’s shot according to a certain system of movements made by the target is essentially an act of learning.
It will often be said that there is no Learning Theory that can be logically applied to machines. It will also be said that at our current stage of knowledge, any Learning Theory I may offer will be premature and most likely will not correspond to the actual functioning of the nervous system. I wish to follow a middle path between these two criticisms. On one hand, I want to provide a method for constructing learning machines, a method that will not only allow the construction of certain specific machines of this type, but will provide a general mechanical technique for constructing a large category of such machines. Only if I reach this level of generalization will I have defended myself, in a way, from the criticism that the mechanical process I claim is, in reality, essentially different in nature from learning. I also wish to describe such machines in terms that are not very different from those used for the nervous system and human and animal behavior. I am fully aware that I will not be accurate in the details when presenting the actual human mechanism and that I may initially be wrong. Nevertheless, if I use a mechanism that can be formulated in terms of concepts used for the human mind, it will be a starting point for criticism and a level of comparison for the expected results with other theories.”