Electrical engineering firms sometimes need circuits which will have to perform a certain function, but about which nothing more can be specified in advance. No instructions are provided as to what kind of circuits they are to be, or how large their components. The "filters" used in telecommunications come into this category. They serve, for example, to unscramble a number of telephone conversations going on at the same time over a single wire, so that each subscriber hears nothing but the call destined for him. A ticklish I business! Planning a filter used to be a task that took Pup the time of two or three engineers and mathematicians for several weeks. Today, there are electronic computer programs to do the work for them. Time I needed, including setting down the results in the high-I speed printer: fifteen minutes! At eight o'clock in the morning the engineer brings the data referring to the I completed filter to the computer on duty. At noon he I can have the filter circuit, soldered up and ready for I use.
To check up on the result, the programmers have made a counter-calculation which determines, from the components for which calculations have just been made, exactly what the filter which is to be put together from them will be able to do. The results must be identical with the requirements specified by the engineer at 8 a.m. (The engineer can also obtain details from the ticket the high-speed printer delivers to him, which will inform him what will happen if a certain tolerance is exceeded in one or more of the components. That saves the filter technician the trouble of making intricate measurements.) The time needed for this second filter program is likewise about a quarter of an hour. A comparison with the amount of work the mathematicians would have to spend on it is impossible, because this program again concerns a calculation that nobody ventured to work on at all before the advent of the computer.
But even the programming of these calculations is no occupation for a lazy afternoon. Each of the two filter programs comprises about 5,000 instructions. A good programmer is allowed about an hour for each of them, to puzzle out the nature of the problem, draw up a flow chart and make any necessary corrections and revisions. So even the programmer, if he had to work on his own, would have to spend over two years at his desk until his computer mastered just one of the two jobs to be done. If we add up his salary, the rent of his office, the use of the computers for testing and so on, we would very soon arrive at a huge cost - about equal to the sum such a program with 5,000 instructions costs. Obviously the whole project will only be a paying proposition if the program is going to be used frequently (for it works out every circuit for one-seventieth of the cost of the two or three human experts we have mentioned) or if the calculation simply cannot be made except by electronic means.