402 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1963 



What characterizes cybernetics, however, is the generality of its 

 ideas. Whereas much of engineering, or cell biology, or neurophys- 

 iology, or economics, is necessarily concerned with particular problems 

 of energy-transport, or biochemistry, or finance, the cyberneticist is 

 concerned only with those abstract features, common to systems in all 

 fields, by virtue of which effective action is organized. 



To bring out some of these essential features, let us first look at our 

 helmsman. Basically, the chain of events by which the ship's course is 

 governed has three stages, which we may call indication (of disparity 

 between actual and desired course), calculation (of form of action 

 required to diminish disparity), and selection (of the helm-position 

 calculated to be appropriate) . As a result of his action, the indication 

 will generally alter, so that we have here a "closed loop" of cause and 

 effect — the so-called "feedback loop" which is a feature of all self- 

 regulating, as distinct from merely physically constrained, mechanisms. 



Cybernetics is built upon the realization that each of the foregoing 

 stages can in principle be mechanized. The compass, the ball-float, 

 the centrifugal governor, and the bimetallic thermometer are long- 

 familiar examples of sensitive devices whose mechanical output can 

 indicate the disparity between the present state of some quantity 

 (direction, water level, speed, or temperature) and some preset "goal 

 state." Rudders, water cocks, steam valves, and furnace regulators 

 are even older examples of the third stage in tlie regulative chain. 



Wliat is new in scale (though not so new in principle) is the mech- 

 anization of the second stage — the process of calculation. Although 

 a mechanical general-purpose computer was devised by Charles Bab- 

 bage in the middle of the 10th century, it was only the advent of 

 electronics that gave such devices the speed and capacity needed to 

 replace the human brain in organizing complex regulative actions. 

 For the simplest type of ball-cock, steam-governor, or thermostat, the 

 only "calculator" required is a simple mechanical link between indi- 

 cator and control, so that supply is reduced when the indicator reads 

 "excess." 



For a self-guided missile, or a "self-optimizing" chemical process- 

 controller, on the other hand, the calculating mechanism must con- 

 tinually modify and supplement the link between the indicator and 

 the selector of action, in accordance with a multitude of other data 

 from auxiliary receptor-organs and from the past as recorded in its 

 storage devices. 



In still more complex automata noAv envisaged, the repertoire of 

 action includes means of changing the pattern of instructions or con- 

 nections within the automaton itself — a kind of gear-changing opera- 

 tion on its own logical macliinery. In this way by "trial and error" 

 (a form of intelligent natural selection), as well as by calculation, 

 it develops within itself the organizing routines and "conditioned 



