The Uses of Computers in Technology

大佬60年前的论文。September 1, 1966

by Steven Anson Coons

考虑做一个整理。

In most technological applications computers have been used to execute a specific program of instructions. Now they are beginning to fulfill their promise of interacting directly with men in engineering design。

Traditional (if so new a field can be said to have a tradition), the other is quite novel. The first category includes the multifarious applications in which the computer carries out a program of instructions with little or no intervention by human beings. This is a powerful way to use an information-processing machine, and it has dominated the early years of the computer era. The second category embraces a new class of applications in which the computer is an active partner of man. I believe that within the next few years this new way of using computers will bring about deep changes in the large segment of technology that might be called "creative engineering."

As the computer is traditionally applied to a technological task, it acts as it is told to act. This is not to say that a machine so instructed cannot accomplish impressive tasks. Its program can be quite elaborate—so complex that no human being could follow it in a reasonable length of time (even, in some instances, in an unreasonable length of time). In obeying instructions, a computer often deals appropriately with changing circumstances and adjusts to variations in its environment, achieving its purpose by a process so subtle as to give the impression of adaptive intelligence. The machine is nonetheless acting as an automaton. Its behavior, although complex, is mechanical and predictable. Man's ingenuity is applied to presenting the problem or setting up the task; thereafter, the machine grinds away at the solution or execution.

This is not the case when the computer and man are linked in what J.C.R. Licklider of the International Business Machines Corporation calls a symbiotic relationship, a relationship in which each can perform the kind of activity for which it is best suited. Man is quite good at inventing and organizing ideas, making associations among apparently unrelated notions, recognizing patterns, and stripping away irrelevant detail; he is creative, unpredictable, sometimes capricious, and sensitive to human values. The computer is almost exactly what man is not. It is capable of paying undivided attention to unlimited detail; it is immune to distraction, precise, and reliable; it can carry out the most intricate and lengthy calculation with ease, without a flaw, and in much less than a millionth of the time that would be required by its human counterpart. It is emotionless, or so we suppose. It suffers from neither boredom nor fatigue. It needs to be told only once; thereafter, it remembers perfectly until it is told to forget, whereupon it forgets instantly and absolutely.

When man and machine work together, the shortcomings of each are compensated by the other, which leaves both partners free to exercise their individual powers in a common enterprise. The potential of such a combination is greater than the sum of its parts.

It was clear when the first electronic computers were being developed that the machines could by their nature deal easily with repetitive calculations. During World War II, computers worked out firing tables for artillery. Another early application was the calculation of logarithmic and trigonometric tables to a large number of significant figures. It was startling to find that some of the classic tables that had been calculated "by hand" contained errors that were discovered only after computer calculation. Computers have continued to specialize in bulky calculations, particularly those in which the procedure is either involved and complicated or repetitive.

It soon became apparent, however, that the computer could also maintain quite sophisticated control over its own procedures and could successfully attack problems of a more difficult kind. Specifically, the ability of the computer to compare two numbers and to elect any one of two or three courses of action based on the outcome of the comparison, although simple in principle, was nonetheless revolutionary.

Has led to some sophisticated applications.
A computer can, in fact, be relied on to carry out the most intricate processes in the manipulation and transformation of information, provided that these processes are understood well enough by humans to be described in complete detail to the computer.

A computer can, for example, control industrial processes. Not all "automated" industrial plants have computer systems, and not all computerized plants are equally automatic. It is possible to construct complex control systems based on continuous monitoring and feedback loops without including computers. Sometimes, computers are introduced to make calculations and inform a human operator what needs to be done. Moreover, a computer can on its own control an individual subprocess or regulate an important variable in a production line.

In some cases (still largely confined to the petroleum and chemical industries), a computer system actually controls the routine operations of the plant. The control of chemical plants is a good example of an application in which the computer can deal with a large amount of information, monitoring the many variables involved in such a way as to maintain optimum production and quality of the product.

The variables in a chemical process—temperature, pressure, flow, valve settings, viscosity, color, and many others—are interrelated in complicated ways, and usually, the relations are highly nonlinear. If two ingredients must flow into a reaction vessel in a certain ratio, and the flow rate of one ingredient is deficient for some reason, it does no good for the computer system to attempt to rectify the deficiency by opening the supply valve wider if the valve is already fully open. Instead, the computer should take account of the state of affairs and close the valve on the other supply line until the desired ratio of flows is achieved.

A computer is able to receive information from many measuring stations located at strategic places in the process plant, to perform the necessary calculations and comparisons of these detailed data, to make decisions on how to monitor the control mechanisms, and to send commands back to them in such a way as to maintain optimum operation. This capability is highly reliable, and since there is essentially no limit to the complexity of the information with which the computer can deal, industrial engineers can now devise processes so intricate that it would be difficult, if not impossible, for a human to follow or manage manually.

Impossible, to control them with human workers.
Another industrial application of computers lies in the numerical control of machine tools. A great many parts of machines are produced by either milling or routing, processes in which a cutting tool moves so as to cut some contoured shape out of sheet metal or heavier stock. In conventional methods, this demands the constant attention of a skilled machine operator, particularly if the contour to be formed is irregularly curved.

Under the control of a computer, the cutter can be made to move in any desired path, and it is, in principle, no more difficult to produce "sculptured" shapes bounded by complex curved surfaces than it is to produce objects with flat faces. The numerical control of machine tools has enjoyed extraordinary success because it guarantees the reliability and reproducibility of even the most elaborate shapes. The spoilage due to human error is reduced to the vanishing point, and many parts are now practicable that would be prohibitively expensive to produce if a human operator had to monitor the settings of the machine.

A striking example is the milling of airplane-wing "skins" from slabs of aluminum alloy. For structural reasons, these sheet metal skins need to be thicker near the wing roots, where the bending stress is high, than they do near the wing tips, where it is less. For a long time, this has been accomplished by assembling an elaborate laminated structure, with sheets of varying thickness fastened together by hundreds of rivets and stiffened by bulkheads and frames. The assembly of such structures is complicated and time-consuming.

Now it has been found that much of the wing structure can literally be cut out of solid slabs—tapered thickness, stiffening members, and all—at a cost and in a time substantially less than is needed for conventional assembly methods. Wing skins cut from slabs two inches thick, 10 feet wide, and 40 feet long are not at all uncommon.

The increasing capabilities of modern computers suggested that a more direct partnership between the machines and their human operators would be effective, and several developments described in other articles in this issue combined to make this possible. First, the languages by which men communicate with computers have evolved rapidly. Language forms have now begun to appear that are much more "problem-oriented."

Another important development that makes the man-machine combination feasible is time-sharing [see *"Time-sharing on Computers," by R. M. Fano and F. J. Corbaté, page 128]. Computers can be operated economically only if they are kept constantly busy at productive work. A man working at a computer console cannot keep the machine busy because the machine can receive a command, interpret and act on it, and return a reply or a result in a few microseconds. Then, it must wait while the human operator digests the reply, thinks about it, and decides on his next action. Enough people at individual consoles can provide the time-shared computer with a workload that will keep it gainfully employed.

A third development is the display console, on which the computer can create symbols, graphs, and drawings of objects and can maintain the display statically or cause it to move, simulating dynamic behavior. Together with input devices such as the "light pen," the display console becomes a window through which information can be transferred between the man and the machine.

The comfortable and congenial combination of man and machine made possible by these three developments has found some of its first applications in computer-aided design. By "design," I mean the creative engineering process, including the analytical techniques of testing, evaluation, and decision-making, and then the experimental verification and eventual realization of the result in tangible form.

In science and engineering (and perhaps in art as well), the creative process is a process of experimentation with ideas. Concepts form, dissolve, and reappear in different contexts; associations occur, are examined and tested for validity on a conscious but qualitative level, and are either accepted tentatively or rejected. Eventually, however, the concepts and conjectures must be put to the precise test of mathematical analysis. When these analytical procedures are established ones (as they are in such disciplines as stress analysis, fluid mechanics, and electrical network analysis), the work to be done...

...is entirely mechanical. It can be formulated and set down in algorithms: rituals of procedure that can be described in minute detail and can be performed by a computer. Indeed, this part of the creative process should be done by the computer in order to leave man free to exercise his human powers and apply his human values.

There is much talk of "automated design" nowadays, but usually, automated design is only part of the design process, an optimization of a concept already qualitatively formed. There are, for example, computer programs that produce complete descriptions of electrical transformers, wiring diagrams, or printed-circuit boards. There are programs that design bridges in the sense that they work out the stresses on each structural member and in effect write its specifications. Such programs are powerful new engineering tools, but they do not depend on an internal capability of creativity; the creativity has already been exercised in generating them.

Potentialities can best give an idea of computer-aided design by describing one of the tools that makes it possible. One of the early and epochal instances of man-machine symbiosis was the program called Sketchpad, completed late in 1962 by Ivan E. Sutherland of the Massachusetts Institute of Technology [see "Computer Inputs and Outputs," by Ivan E. Sutherland, page 86]. Sutherland used the TX-2 computer, an experimental machine built at the Lincoln Laboratory of MIT, with the idea of providing direct man-machine interaction at the console long before such a notion had much currency in computer technology and long before the notion of multiple users of a machine was much more than a dream.

When Sutherland began work on Sketchpad, the TX-2 had a cathode-ray-tube screen and a light pen as existing rudimentary pieces of equipment, but little had been done to exploit their possibilities. Sutherland set out to develop a system that would make possible direct conversation between man and machine in geometric, graphical terms.

In the course of the development of Sketchpad, he would invite people in to try out his system so that he could observe their reactions. On one occasion, Claude E. Shannon, Sutherland's adviser on his doctoral thesis, wanted to perform a geometric construction. Rather than work out the construction on...