The Friendly Orange Glow

by Brian Dear

  The technician was an accomplished engineer named Wayne Lichtenberger, who had not only also graduated from the University of Illinois in 1955, but was, like Bitzer, a winner of that year’s Bronze Tablet award. Lichtenberger had become involved with digital computing at CSL in 1957 while he was in graduate school, and had known Bitzer since freshman year.

  “The theory was,” Braunfeld recalls, “a) here we are at a university, so by definition we know how to educate, and b) we know how to do what was in those days called real-time data processing. Which I guess today there isn’t even a word for that, because it’s all real-time. But the notion of real-time data processing was a big deal in those days, because computers typically had as their inputs cards or paper tape, and it came in and it did its thing and it came back out. The notion of actually getting stuff in at random times was rather rare.”

  Having all come fresh from the Cornfield project, these CSL engineers immediately saw an analogy between Cornfield’s need to track the real-time comings and goings of multiple airplanes and naval vessels, identifying which was friend, which was foe, while they moved in different directions at different speeds, and the new project’s need to track the real-time interactions of multiple students each sitting at a terminal, each interacting with the computer at their own pace, each student having essentially an ongoing, private, live conversation with their digital tutor. The big breakthrough that gave Bitzer and his colleagues a surge of confidence was that they could see at an abstract level the similarities between these two very different problem spaces—air traffic control and student data. Step back far enough, and from a purely architectural point of view, naval destroyers’ radar data could be seen as data from student terminals. To the computer it was all just zeroes and ones. Students would be interacting with the system in real time, so the data would be pouring in from multiple sources in real time. No different, to the computer, from multiple aircraft radar pings being fed back to the central machine in real time. The notion of a central computer handling lots of random incoming data, processing it as fast as possible, and spewing it back out to graphical terminals? Just think of the students as aircraft and it all started making sense.

  —

  On June 3, 1960, Alpert finally sent a letter to Dean Everitt announcing his decision to launch a new project to build an automated, computerized teaching system, having found someone with the requisite enthusiasm, expertise, and energy to lead it. He described the project this way:

  As suggested by Professor Sherwin some time ago, the advent of the high speed computer makes possible a new approach to education in which the principle of feedback (from the student) may be applied to such traditional educational tools as the book, visual aids, etc. While this idea is in principle applicable to the teaching research objective of any course of study at various levels of sophistication, the first research objective will be toward the design of an “automatic teaching machine” to teach students at various levels how to use a high speed computer. This particular objective is especially suited to the talent and motivation of the people who would have to design such a machine and represents a unique educational need in the coming decade.

  Alpert explained why teaching how to use the computer itself was the ideal first subject for the automatic teaching machine:

  1. The design of the first machine requires decisions as to the philosophy of the approach, the design and use of a computer complex (the “machine”), and the programming of a course, all of which are intimately related. This relatedness means that the decisions must be made by people who not only know computers but also the proposed course of study. By selecting a subject which is (by definition) understood by any computer expert we have an obvious advantage in staffing a creative program.

  2. One of the important aspects in learning to use a computer is to learn its language. It seems an advantage rather than a disadvantage to omit the human teacher and thrust the prospective student into contact with a machine.

  3. This teaching objective is finite and specific. The effectiveness of the teaching program will be measurable in an objective way.

  4. The research program will be uniquely suited to studying man-machine relationships. Both the language of the machine and new methods of input-output will necessarily be related areas of investigation.

  5. While it is expected that the techniques developed can be applied to other subjects, it is evident that teaching the use of the computer to people in all walks of life is of itself a broad educational objective with far-reaching implications. It is particularly suited to an educational institution.

  Such was Alpert’s confidence in his laboratory’s expertise and intellectual power that they would propose to embark on such a project that had no precedent and no certainty that it would work, and that indeed in all likeliness would fail. It is also noteworthy that the project would not directly involve educational psychologists, learning theorists, and others from the social sciences. Alpert seems to have become so fed up with them that when the idea of using Bitzer came to him, it must have looked like the light at the end of the tunnel. With Bitzer, Alpert had found a way to keep an eye on the project and keep it entirely within CSL. Over the next five decades, this decision would fuel criticisms that PLATO was too engineering centric: built by electrical engineers and physicists with little input or direction from real educators, when, considering PLATO’s mission, it should have been led, designed, and run by educators.

  —

  Peter Braunfeld did his undergraduate work at the University of Chicago, and decided in 1949 to pursue a graduate degree in mathematics at the University of Illinois. He became a teacher’s assistant in the math department, but needed a job for the summer. “In those days they had guys going around with sticks with spikes in them to pick up pieces of paper in the Quad,” says Braunfeld. “I thought that it would be a great job. It would be outdoors, it would be mindless, and it would bring in a little bit of income.” He went to the student employment office to apply for the trash-picker’s job, but was surprised when they called him to say, “We want you to go over to a place called the Control Systems Laboratory because they’re looking for people.” A physicist named Lloyd Fosdick hired him “essentially as a gofer,” Braunfeld recalls.

  Braunfeld had a reputation for being an excellent mathematician. “He was an extremely good mathematician, extremely good,” says Jeannine Leichner, one of only a few women who worked at the lab. “Peter was a really nice guy and very, very bright.” But it took time for the lab to appreciate how good Braunfeld was.

  Fosdick envisioned CSL building a computer that could track many airplanes in the sky simultaneously, in real time. A Navy aircraft carrier, escorted by destroyers, cruisers, supply craft, and other fleet vessels, would collect the radar data from many of these vessels, which would be scanning the skies for airplanes. The radar data would then be fed into a central computer aboard the carrier, which would sort out the data, identify directions, speeds, and altitudes of aircraft, evaluate threats, and, when foes were found, provide targets to eliminate them. This was the secret project Cornfield developed on the fourth floor of CSL.

  “As time went on,” Braunfeld recalls, “I got more and more interested in this stuff and I guess I got to be pretty good at it, and literally, kind of a Horatio Alger story, from just doing dog-work, they kept giving me more and more difficult things to do, and ultimately I was involved in doing a lot of programming for them.”

  Eventually, CSL tried not only to track airplanes but also to shoot them down. “There was a big project,” says Braunfeld, “that involved the interactions of humans at screens and what functions they could perform best and what functions could the computer. Did the computer really help at all?”

  Cornfield was beginning to wind down by the time of Sputnik, “and it was clear,” says Braunfeld, “the military wasn’t so inclined to support these things anymore.” While the benefits of a Cornfield-type system appealed to the Navy, the actual implementation
of Cornfield turned out to be not viable.

  A quick overview of the design of the system may provide some insights into the early thinking of PLATO. When one considers a 1950s-era digital computer, of which there were only a few in the world, it is important to remember that they were hulking machines full of vacuum tubes requiring enormous electrical power to operate and considerable physical space in a climate-controlled building. Cornfield’s designers felt that installing such enormous computers on every ship in the fleet, or at least the destroyers that escorted an aircraft carrier, was simply impractical. Instead they proposed that the aircraft carrier serve as a “mainframe,” or data center, a central processing facility that could crunch through the data fed from the various destroyers’ radar systems and then feed back the results to each ship. CSL’s engineers proved that such a system could work, but the requirements for ILLIAC-style digital computers were so extreme that the Navy concluded the Cornfield concept was not suitable for operational use. The Navy saw the centralization of the data processing as one of Cornfield’s less attractive design attributes: in engineering-speak, it’s what’s known as a SPOF, or “single point of failure.” They would have preferred that each ship have its own computer and be able to assess its own radar data rather than depend on the aircraft carrier’s computer. (Perhaps the Navy’s memory of World War II and the heavy losses of ships accounted for their reluctance to put all their computing eggs in one basket.) In the 1960s, as the cost and electrical and physical space requirements of digital computers became more reasonable with each successive year, thanks in large part to the tiny, more efficient semiconductors replacing big, bulky vacuum tubes, some aspects of the Cornfield concept did eventually go to sea, in the form of the Naval Tactical Data System, which featured a more distributed model with computers on multiple ships in a carrier fleet.

  “There was a feeling,” Braunfeld says, “that the people who were involved in this Cornfield system were about the best bunch of people that had ever been assembled in any kind of work….We were an incredible, elite group.” This feeling may have contributed to Bitzer’s confidence in accepting the challenge from Alpert to organize and lead a team to build a teaching computer. Bitzer and Lichtenberger would handle the engineering for PLATO, and Braunfeld would handle the software. Their shared experience on Cornfield guided their thinking for the new PLATO project. But it didn’t take long for them to see that computer-based education was a wholly different kind of challenge.

  In June 1960, coming up with a name for their new computer project “took the first week of research,” Bitzer recalls. “We had a meeting the first week. Peter Braunfeld and Chal Sherwin and I met and…we each came with names. I came in with the name ‘PLATO,’ and I said ‘Programmed Logic for Automated Teaching Operations.’ ” Bitzer recalls Braunfeld saying, “No, no, we can’t say that, ‘Automated’ means self-mated, we don’t want to give that image, we’ll call it ‘Automatic.’ So we made that change, and that was it, and then we got down to work.”

  4

  The Diagram

  Bitzer was frugal. “Don never spent a nickel if a penny would do,” says Garrie Burr, a longtime technician on the PLATO project. Consider Bitzer’s legendary attire—the man was widely believed to own just one suit and that one suit was widely believed to have been a hand-me-down bought at a Goodwill store. “Clothes meant very little to Don,” says his stepmother, Ruth. Even back in his undergraduate years, Don’s clothes were a thing. Says one Beta Theta Pi brother, “His clothes, he’d just drop them on the floor, didn’t care if they were rumpled up or what, and would hop back in ’em the next morning and you frequently wondered if they ever got washed, because he just didn’t care.” Judy Sherwood, who would work on PLATO a few years later, vividly remembers, “He would go out jogging at lunch time. And he had as usual these raggedy awful-looking jogging clothes.”

  One time, Bitzer threw his white T-shirt and white shorts into the washer with some red towels, and they came out pink. “My dad gets pretty attached to just one item of clothing,” says his son, David. “He went right ahead and wore them, and jogged in this pink shirt and these pink shorts, and I think this was probably at the time that the Peter Sellers movies were out and therefore he got the nickname, jogging through campus, of the ‘Pink Panther.’ ”

  He wore the pink running outfit for a long while afterward, and at least one time it became an issue. After a run, he arrived back at the office and found his secretary still out at lunch. There was a man sitting in the outer office. The conversation unfolded like this:

  BITZER: [in glorious raggedy pink running outfit] Hello, can I help you?

  GUEST: I’m Professor _______ and I have an appointment with Don Bitzer. And I’m just waiting. He’s at lunch and I’m just waiting. People in the hallway told me I could sit here, he’d be back in a little while.

  BITZER: Well, I’m Bitzer, can I help you? Why don’t you come on into the office? And I’m glad you got here.

  GUEST: You’re not Don Bitzer.

  BITZER: Well, I am.

  GUEST: No, no, no, you’re not Don Bitzer.

  BITZER: How do you know I’m not Don Bitzer?

  GUEST: Well, I’ve never met Bitzer and I don’t know what he looks like, but he doesn’t look like you.

  “That was pretty typical of Bitzer’s attitude and approach towards people and his garb,” Sherwood says. “He goes into the office and he finds somebody there and he didn’t know him from anybody, and he welcomes them heartily, you know, let me show you around, in his raggedy pink jogging shorts, I mean, somebody with more sense of style would have stopped in the john or gym or someplace and changed his clothes before he went back to the office. But not Bitzer. He probably had his clothes in his own office and he’d shut the door and he’d change clothes. Oh, he was a character.”

  He was frugal with cars as well. There was the Volkswagen so past its prime, so decrepit, that the floor (“What floor?” quipped one relative) rusted out and you could see the ground while riding—hopefully you did not ride over a puddle, as the splash came right up into the car. Later he got a Dodge Dart (from Bitzer Motors, naturally), for which he was reluctant to get power steering, considering that an unnecessary luxury. He ran that Dart for many years, again long past its prime. “Every car Don had was into the well-after’s,” says a relative.

  —

  That same frugality came in handy with PLATO as it was getting started—and it no doubt pleased Alpert that he’d picked someone to lead the project who was not going to burn a lot of CSL’s budget while building the system. PLATO had to get by on whatever spare change CSL was willing to toss at it, plus whatever Bitzer could scrounge together. CSL had priority access to the ILLIAC. That machine might be ugly, might be slow, might be severely limited in capability, and was, in 1960, already eight years old and aging fast—recent advances in integrated circuits and solid-state electronics went far beyond the ILLIAC’s short-lived vacuum tubes, and would soon become the standard for new computers—but that’s all they had to start with.

  Bitzer, Lichtenberger, and Braunfeld already knew how to write code for it, if you can call it that: it was an act requiring long hours of thinking and planning, large amounts of stamina, determination, and, above all, paper: one didn’t simply march over to the machine and type the code into it and watch it run. Instead one tested out ideas on paper, wrote and rewrote the routines, met with others, and commented on them, and only after a thorough vetting would one march over to the machine—assuming you had booked some time beforehand—and type in the magic incantations that then generated punched holes in a special paper tape.

  To make the ILLIAC do anything required first getting yourself into a special place mentally, getting into a “zone,” a place that put you, in the author Ellen Ullman’s words, “close to the machine.” It required the equivalent of a Swiss watchmaker’s intimacy with tiny gears, wheels, springs, rollers, and escapements, only in this case the intimacy was with logic circuits,
vacuum tubes, and a lot of binary and hexadecimal mental gymnastics. It required thinking like the machine, one that was extremely limited in memory and speed, and understanding that a machine only ever did what it was told. If you gave it erroneous instructions, it faithfully went and made errors, because as far as it was concerned, that’s what you told it to do.

  Modern computers take advantage of what are called “interrupts,” signals sent either by the program running at the moment or by some component of the hardware, to the central processor to tell it to stop what it’s currently doing, set aside its current task and the memory associated with that task, and go deal with another task using that task’s memory until told otherwise (that is, until it receives another interrupt). Once the higher-priority task has been dealt with, usually in a tiny fraction of a second, the processor can resume working on its original task. Every computer today uses interrupts. The ILLIAC’s 1952 vintage afforded no interrupts whatsoever. “You had to be very, very careful how you wrote the code so that ILLIAC could respond suitably,” Lichtenberger says. “I remember fooling around with that for quite a while until we worked out appropriate techniques.”

  Despite the ILLIAC’s numerous quirks and limitations, the PLATO staffers were confident they could deal with them. They had experience connecting peripheral devices to the ILLIAC, back during the Cornfield project. It was already wired up, ready to go. So the ILLIAC was going to be the engine that powered PLATO, whatever PLATO turned out to be. There were other pieces of the PLATO system that they would have to build from scratch. And the ILLIAC knew nothing about teaching, or computer science; all kinds of software would have to be written. The amount was in fact absurdly ambitious, and the chances of a tiny team getting it all done were hardly confidence-inducing.