AMPLIFYING REGULATION 14/5 



not worth serious consideration. (It is not, however, more blatant 

 than the crane-driver's dependence on a suitable power supply.) 

 This case, however, is only somewhat extreme (having been selected 

 to show one end of the scale). Other cases he further along the 

 scale, and are of more general interest. The principle, however, 

 remains unaltered. 



Next consider the case in which Q wants a water bath to be restored 

 to a certain temperature; restorations will be required 100 times in 

 each day and over a whole year. This means that on 36,500 

 occasions the temperature must be corrected by a raising or a lower- 

 ing — a one-bit selection, say. The whole Grand Disturbance 

 (S.13/8) thus has variety of 2-^6^0° possibilities. Q probably could 

 transmit this in the year, but finds it inconvenient. If then his 

 resources are such that he can make a thermostat at a cost of, say, 

 1000 bits, then by using the fact that the Grand Disturbance is 

 repetitive (S.13/9), the act of selecting appropriately from 1000 bits 

 has as consequence the correct selection from 36,500 bits. So an 

 amphfication of about x36 (if measured on the logarithmic scale) 

 has occurred. 



This second example is more ordinary than the first. The fact 

 that its method is widely used in practice shows whether or not the 

 practical man thinks it worth while. 



There is, of course, not necessarily any amplification; and the 

 practical man, before he builds a machine to do a job, always makes 

 at least an intuitive assessment of the balance: 



Cost (in some sense) of making 

 the machine which will do the 

 job. 



Cost incurred by doing it 

 himself. 



What this chapter deals with are the actual quantities involved, 

 when our interest is centred on the amount of communication and 

 selection that is required. 



Finally let us consider an example in which the possibility of 

 amphfication is obvious and of practical use. Suppose twenty 

 men are given the task of keeping two thousand rooms constant in 

 temperature and humidity. If some means of control exists in each 

 room, the twenty may yet find the task beyond their capacity if 

 they try to compensate for all the atmospheric variations by manipu- 

 lation of the controls directly. It may happen, however, that 

 machines are available such that if the men become mechanics and 

 act as regulators to the machines, the machines can be made into 

 air-conditioners and maintained as such. And it may further 

 happen that the amount of regulation that the mechanics can supply 



269 



