The Computer at Dartmouth

A New Essential of Liberal Education

PROFESSOR OF MATHEMATICS

DURING the past year we saw the twentieth anniversary of the first dropping of the atomic bomb. I am not going to discuss the bomb, but I am going to tell you a little story about some computing that went into the bomb.

When I was sent to Los Alamos, they assigned me to what was then called "the computing center," which really consisted of some rather primitive IBM bookkeeping machines. There were seventeen of them in a large room. A staff of between fifteen and twenty of us who worked in three shifts - 24 hours a day, six days a week — tried to solve some mathematical problems going into the construction and dropping of the bomb. A typical problem took us about three weeks to do, working 24 hours a day, six days a week, and we worked about a year and a half on various kinds of problems.

I recently made an estimate as to what this would mean in terms of the present Dartmouth Computation Center. I have come to the conclusion that any sophomore at Dartmouth College starting from scratch - that is, where he has to start writing instructions for the machine - could do all the work that was done in a year and a half at Los Alamos in one afternoon. And he can do it in one afternoon at a time when there may be thirty other people using the same computer. This is the system that I would like to tell you about.

I shall start with a word of motivation. When the new Computing Center was proposed to the Trustees of Dartmouth College, our major argument was the following. Computers are beginning to have an increasing effect on the lives of all of us. Almost all large businesses today are influenced by computers. Twenty years from now all businesses, and most private lives, will be influenced by computers. Whether this is going to be a fully favorable effect, as it could be, or a very harmful one will depend on whether the people who make the policy decisions know what computers can do and what they can't do, or whether they blindly trust the people who run the machines. An institution like Dartmouth College that trains many of the leaders of the nation in business, in government, and in research, has an obligation to make sure that our liberal arts graduates understand one of the most significant factors that will influence their lives. Therefore, we have made it a requirement that everyone who takes at least a year of mathematics at Dartmouth (which includes 80% of each freshman class) must learn something about using the computer. Not just indirectly, but must actually use the computer and show that he understands how to put it to good use.

Now this raises a very serious problem. Computers are very expensive, and human beings, especially when they first try using a computer, tend to be very slow. So the problem is: "How do you enable hundreds of students to use the same computer?" The answer is what is called a time-sharing system - a system under which many different people can all use the same machine at the same time.

With this system each user sits at a teletype machine, which is a pretty ordinary typewriter except that it can send what it types over telephone lines. It is connected by a telephone line to the computer. There are 40 teletypes hooked to our present computer. A number of them are at the Computing Center. A few of them are at departments that make heavy use of computing, some at the Associated Schools, and some of them are at remote locations. People have called in to our Computing Center from all over the country. Dean Tribus gave a demonstration in Scotland: he fed in problems there, had them solved in Hanover, and the answers were typed out seconds later in Scotland.

With any teletype, by calling in, you have complete control of the computer and absolute freedom to work with it as if you were the only person using the machine. You don't ever have to see the machine. You don't have to know how it works. All you have to understand is an extremely simple language, called BASIC, that we have constructed for this purpose. It is something our students typically learn in three hours. And then you can sit down and have one of the most powerful research and educational tools at your complete disposal.

At the moment, the heart of the computing equipment is located in an ugly basement in College Hall. But the new Peter Kiewit Computation Center is already under construction, on the corner of Elm and Main Streets. This will provide the kind of elegant surroundings such a magnificent system deserves.

THERE is a famous story about the trouble with computers. Twenty years ago the difficulty was that they never did what you told them to do. The trouble with computers these days is that they do exactly what you tell them to do, not what you meant to tell them to do.

As this story indicates, the major problem in computer usage is learning how to instruct the machine - or programming. The human user must in a few instructions explain to the machine what he would like it to do. This requires that he anticipate all the difficulties that may arise in thousands or even millions of computations. And since the machine has no "common sense," the instructions must be absolutely precise and complete. It is to this skill that we want to introduce our students.

We are currently putting about 650 freshmen through the training program each year. During the second semester of their mathematics sequence they have to write four programs having to do with mathematics and they must work on them until they get the right answer. Now this is not an enormous chore. It is a small part of their course. On the other hand, it makes all the difference in the world whether you have ever used the machine or not used the machine. Once they have passed this requirement, they never, if they wish, have to touch a machine again in their entire lives.

On the other hand, if they want to, the machine is there any time they can get their hands on one of the typewriters, day or night. So a student can go there and do his homework in a physics course, or do a research project in an engineering course. There is a statistics course for psychologists which is built around computers. The student may be a research assistant for a faculty member and solve a significant research problem. All these things have happened during the past year and should happen with increasing frequency in the future.

Of course the Computing Center is even more important for the faculty. All of a sudden we have, literally at our fingertips, a tremendously powerful computer. In the past, if you had a big research project, it was worth while going through all the labor of writing programs, punching up cards, submitting them, and going back a dozen times until you got it to work. Now we are at the stage where, with most research projects that require numerical work, you can sit down at the typewriter and not get up from that typewriter until you have your problem solved.

This is a common experience. And I assure you that if this sounds fantastic to laymen it sounds much more fantastic to people who are experts on computers. For example, we had a consultant visit us last summer. Very fortunately he was here for two weeks, because for the first week we could not get him away from the typewriter. And he had been working with computers for years! Our timesharing system was a revolutionary new experience for him.

How does this marvelous system that allows 40 different people to use the machine at the same time work? The Computation Center uses equipment manufactured by the General Electric Company. The heart of the system is not one, but two computers. The "computer" in the ordinary sense is a GE-235, which is a nice high-speed computer. It can do about one million multiplications a minute. Its memory is large enough to hold 8000 nine-digit numbers, or 16,000 instructions. While this is quite remarkable, it is far from the fastest or largest machine now available. We are proud of the fact that our system works without needing a multi-million-dollar computer.

The second machine is the Datanet 30. Its availability was one major reason why we bought General Electric equipment. It handles for us all the communication with teletypes, and makes all the key decisions. In effect, it is the boss, and the GE-235 is its slave. It talks back and forth to you. And it is a very special machine that can talk to all 40 teletypes at the same time in such a way that it could be typing out 40 different sets of answers on 40 different teletypes, some of them in California, all at full speed.

There is also a very large memory which is connected to both computers. This memory looks like a gigantic jukebox, and works somewhat on that principle. It can store a few million numbers (or instructions) and find any block of 1000 of these in a quarter of a second. These magnetic disks are a tremendous improvement over magnetic tapes, since it does not matter whether we want information from the "beginning" or the "end."

Let us now follow a typical program through the system. A student sits down at a teletype and types a short list of instructions. These enter into the Datanet, simultaneously with everyone else's program. The Datanet deciphers the teletype signals, and collects them on one of the disks. When the student types RUN, the Datanet signals to the GE-235 and tells it to go get the program from the disk and work on it. Let us suppose that the work is complete within five seconds, as most requests are. Then the computer places the answer (or an errormessage) on one of the disks and goes on to the next problem. While the Datanet types the answer on the student's teletype, the GE-235 may be solving a dozen other problems.

Let us suppose that there were errors in the original program (and there usually are). Then the student simply retypes those lines that had mistakes in them, or types a couple of new lines, and types RUN again. His old program is still on the disk and will remain there as long as he sits at the teletype, so that corrections can be entered simply and quickly. An experienced person may be able to get a dozen runs in a quarter of an hour — usually enough to make all necessary adjustments in his programs.

But what if his program requires more than five seconds? Then the Datanet signals "time is up," and the GE-235 moves all its computations to the disk. Each customer waiting in line gets a shot at the machine and then the student is allocated further computing time. His problem will still be done, but he is not allowed to hold up the bulk of requests, which require only seconds of computing. Our experience is that if we need a quick run we get service within ten seconds. And a long problem, requiring a minute of computation, will be completed within five, minutes.

THE disk memory is also used for saving programs. This has many important applications. For example, a student who is unable to complete his work in a single session may save his program and return to it at his leisure. A faculty member may develop a program that he wants to use repeatedly for a research project. He can save it on the disk, and retrieve it in seconds any time he needs it. Altogether between five and ten thousand programs can be saved. Among these are the so-called "library programs." These are programs of general interest, written by experts, available to anyone. For example, there is a program for computing correlation coefficients. A social scientist needs to know practically nothing about computers to use this. He simply calls the program from the "library," types in his data, types RUN, and within a few seconds he has his correlation.

WHAT we have learned is that we have been asking the wrong questions about computers in the past. We used to ask, "Is man better at doing this or is the machine better at doing this?" Usually, if the machine could do it at all the machine was better. But this is the wrong question to ask. The right question to ask is, "What is the best way of getting a given result when man and machine work together?" And what we have found out is that man and machine form a team that is of course much faster than the human being alone but much smarter than the machine left to its own devices.

So instead of trying to write an enormous program and let the machine calculate for hours - and heaven only knows what happens in between - you let it work for a while, look at it, and make a suggestion to the machine, and let it go on. This gives a new feeling to computing. Many of the leaders in computing feel today that this kind of time-sharing is the future of all computing, not just because many more people can use it, but because of this tremendous new feeling of man working together with a device that magnifies his own mental resources.

If you think this physical equipment is fabulous - and it is - what you really should appreciate are the students who wrote 90% of the instructions for the machines. For example, the heart of the executive program was written by two students who, at the time, were a sophomore and a junior at Dartmouth College. Working as student assistants, they earned about $500 a year and maybe $60 a week during the summer, which helped them pay for their college education. Now the same students can earn in two weeks as much as they earned in a year with us. And believe me, they are worth every penny of it, because, right at this moment, they know more about time-sharing than most of the "experts" in the country.

We have some of the most fantastically able undergraduate students in the country. And if our system is better - as we think it is - than the systems of other institutions, it is not because the equipment is better, but because the brainpower of a small number of faculty members and a large number of hardworking students has made the difference for Dartmouth College.

Professor Kemeny at a computer teletype in the Bradley Mathematics Center.

Diagram of the components in Dartmouth's remarkable time-sharing system.

Prof. Thomas E. Kurtz of the Mathematics Department is Director of theComputation Center, for which a new building is now being constructed.

PROFESSOR KEMENY'S article is based on a talk he delivered to a group visiting the campus as guests of the College at a Dartmouth Horizons program.