13/22 AN INTRODUCTION TO CYBERNETICS 



and that a particular one of the possible forms is sought. This 

 selection is just about 1 megapick, and we know that the attempt 

 to select seriatim is hopeless. Can the selection be made by groups? 

 We can if there can be found some practical way of grouping the 

 input-states. 



A particular case, of great practical importance, occurs when the 

 whole machine is reducible (S.4/14) and when the inputs go separately 

 to the various sub-systems. Then the sequence: select the right 

 value for part 1, on part I's input; select the right value for part 2, 

 on part 2's input; and so on — corresponds to the selection being 

 conducted by groups, by the fast method. Thus, if the machine 

 is reducible the fast method of selection can be used. 



In fact, reducibihty is extremely common in our terrestrial 

 systems. It is so common that we usually take it for granted, but 

 he who would learn how to regulate the very large system must 

 become fully aware of it. 



To get some idea of how much the world we live in shows re- 

 ducibihty, compare its ordinary behaviour with what would happen 

 if, suddenly, the reducibihty were lost, i.e. if every variable had an 

 effect, immediate or delayed, on every other variable. The turning 

 over of a page of this book, instead of being just that and nothing 

 more, might cause the lights to change, the table to start moving, the 

 clock to change its rate, and so on throughout the room. Were the 

 world really to be irreducible, regulation would be so difficult as to 

 be impossible, and no organised form of life could persist (S.7/17). 



The subject must be left now, but what was said in Design ... on 

 "Iterated systems", and in the chapters that followed, expands the 

 thesis. Meanwhile we can draw the conclusion that if a responsible 

 entity Q (S. 13/10) is to design (i.e. select) a machine to act as 

 regulator to a very large system, so that the regulator itself is 

 somewhat large, the achieving of the necessary selection within a 

 reasonably short time is likely to depend much on whether the 

 regulator can be made in reducible form. 



13/22. Whence the Regulator"} Now at last we can answer the 

 question that has been latent throughout Part III: how is the desired 

 regulator to be brought into being? The question was raised in 

 S. 13/10, but since then we have explored a variety of topics, which 

 had to be discussed before the threads could be pulled together. 

 Let us now survey the position. 



The process of arriving eventually at a particular machine with 

 desired properties implies selection, and it also implies that the 



262 



