12/18 AN INTRODUCTION TO CYBERNETICS 



The mechanism is known to be error-controlled, for the informa- 

 tion that determines whether it shall move on or stick comes from 

 the line itself. 



This case is so simple as to be somewhat degenerate. If we pay 

 no attention to the internal actions between R and T, so that they 

 fuse to form the F of S.10/5, then the case becomes simply that of a 

 determinate system which, when the initial state is given, runs along 

 a determinate trajectory to a state of equilibrium. Thus every 

 basin with a state of equilibrium in rj can be said to show a simple 

 form of regulation; for it acts so as to reduce the variety in the 

 initial states (as disturbance D) to the smaller variety in the terminal 

 state. 



Much the same can be said of the rat that knows its way about a 

 warehouse; for wherever it gets to it can make its way back to the 

 nest. As much can be said for the computer that is programmed to 

 work by a method of successive approximation; for, at whatever 

 value it is started, the successive values are moved determinately to 

 the goal, which is its only state of equilibrium. 



Ex. : A card is to be found in a shuffled pack of 52 by examination of them one 

 by one. How many will have to be examined, on the average, if (i) the 

 cards are examined seriatim, (ii) if one is drawn, examined, returned if 

 not wanted, the pack shuffled, a card drawn, and so on? (Systematic 

 versus random searching.) 



12/18. When the machinery is all determinate, the problem of 

 S. 1 2/ 1 4 may arise — that of getting Tto go to some state of equilibrium 

 that has some desired property. When this is so, the solution given 

 there for the Markovian machine is, of course, still valid : one couples 

 on a vetoer. 



12/19. Continuous variation. After these primitive forms, we 

 arrive at the regulators whose variables can vary continuously. (It 

 must be remembered that the continuous is a special case of the 

 discrete, by S.2/1.) Of the great numbers that exist I can take only 

 one or two for mention, for we are interested here only in their 

 general principles. 



Typical is the gas-heated incubator. It contains a capsule which 

 swells as the temperature rises. The mechanism is arranged so that 

 the swelling of the capsule cuts down the size of the gas flame (or of 

 the amount of hot air coming to the incubator) ; an undue rise of 

 temperature is thus prevented. 



236 



