by Brian Dear
JETS exhibiting REGITIAC, 1963: Carolyn Leeb, Mike Walker, Maurice Miller, Nick Altenbernd Credit 14
—
Bitzer’s house became the new meeting place for JETS. With Hanson graduating from high school, the handmade JETS computer, by now christened REGITIAC (the Urbana High School sports mascot was the tiger, so they reversed the letters and added “IAC” as a homage to the ILLIAC), was no longer welcome on his family’s Ping-Pong table in the basement. Mike Walker gathered it all up and moved it to another location—Bitzer’s basement—and then took over the reins as the leader of the JETS. “I ended up,” says Walker, “basically living in Bitzer’s basement for a couple hours every evening after I delivered my papers and before I went home to have dinner.”
Walker had been introduced to JETS by his “terrific” high school math teacher, Miss Brannon, but he says it was Bitzer who drew him into the world of electrical engineering and computing. “It was Don Bitzer’s enthusiasm and acceptance of young people and bringing them in and putting them to work. I mean, truly. The guy’s got absolutely infectious enthusiasm. He’d bring you in and he would take the time, set you up to doing something, and there I was, winding coils on a lathe, and doing this sort of stuff, and hey, he wasn’t too proud to do some of that himself.”
Some parents of the JETS kids were growing suspicious of what exactly was going on in Bitzer’s basement. “One day, I was in a book club,” says Bitzer’s wife, Maryann, “and one of the members came and she said to me, ‘Maryann, I’ve got to apologize to you, I just couldn’t stand it anymore, but I came to your house, and I was looking in the windows.’ ” She told Maryann that each day after high school, her son would tell her that he was heading over to the Bitzers’ to work on the computer. “I feel terrible admitting this,” the woman told Maryann, “because I should trust him, he’s always honest.” The woman went up to the windows of the Bitzer home and peered in. The basement was mostly at ground level so it wasn’t hard to look inside and see what was going on. “I felt so terrible,” the lady confessed to Maryann, “because there they all were. All around this computer in your basement.”
—
From the start, Bitzer knew that PLATO was going to need a more powerful computer than the ILLIAC. CSL also realized it needed more computing power for the variety of projects under its roof. In time the lab invited hardware manufacturers in to discuss what machines were available. “It turned out,” says Bitzer, “that the Midwest salesman from CDC [Control Data Corporation], by the name of Harold Brooks, visited CSL to show us the CDC computer. Before he joined CDC he used to sell heavy equipment—cranes or the like. What a salesman.”
Getting that CDC 1604 was not easy. CSL took competitive bids from a number of vendors, including CDC and IBM. CDC’s bid was about $995,000. IBM’s, for a 7090 computer, was about $1.9 million. That led to a second round of bidding. CDC came in again at $995,000. IBM this time lowered their price to about $850,000. CSL still wanted the CDC machine. IBM was furious, and contacted the state legislature in Springfield to force CSL to accept the IBM bid. CSL didn’t want it. The legislature threw the whole thing out and told them both to rebid. For the third time, CDC came in around $995,000. IBM, for the exact same 7090 computer, came in around $200,000. This time, CSL was furious. Such were the ways of IBM in those days. After a long fight CSL got their CDC 1604, and IBM’s shenanigans would be remembered for many years to come by everyone at CSL, including Don Bitzer.
The first thing they did on the 1604, for which the PLATO project was granted about one to two hours a day, was create a simulator for the ILLIAC. Ten years of algorithms and libraries of subroutines created on the ILLIAC were too valuable to reinvent on the 1604, so it was decided to emulate the ILLIAC on the new CDC machine as a way to jump-start its use and be productive more quickly. Wayne Lichtenberger wrote the program, which they named SIMILE for “Simulate the ILLIAC Exactly.” “All of our programs in the laboratory and PLATO were written for the ILLIAC,” Bitzer says. “SIMILE was a program that made the 1604 simulate the ILLIAC and the simulation mode ran at roughly the same speed as ILLIAC, which was very slow. Nevertheless, the first thing we did was run PLATO in the SIMILE mode in order to get more clock time than the one or two hours a day on the ILLIAC that was available for the whole laboratory.”
Bitzer corralled Andy Hanson and Mike Walker to work on PLATO, continuing the tradition he had begun with Frampton and Singer. Hanson graduated in 1962 from Urbana High, and went on to Harvard. But even as a college student, during the summers he would come home to Urbana and work for Bitzer on PLATO. “What I recall,” says Hanson, “is that the work of at least a core group of Bitzer’s Boy Scouts in ’62–’63 seemed essential to Bitzer’s plan to make PLATO into a viable project on the 1604 in the face of zero resources, active opposition from some in CSL, and a power struggle with his collaborator. He needed massive amounts of programming and experimentation and he managed to get what he needed from us for practically nothing, and on top of that built the rest of the empire for which he is justly famous.” His “collaborator” was Peter Braunfeld, who was slowly detaching himself from Bitzer and PLATO. Too many cooks in the kitchen, it seemed, and Bitzer had established himself as head chef.
The challenge Bitzer presented to Hanson was extraordinary: write a multiuser, multitasking resident operating system for the 1604 that would support thirty-two simultaneous users. Not even CDC had been able to build such a thing on the 1604, but such was Bitzer’s unstoppable optimism that he was confident Hanson, all of nineteen, would be able to pull it off. “My work on the resident program,” says Hanson, “was built around Bitzer’s explaining a finite-state machine to me, and then expecting me to figure out the rest.” A “finite-state machine” is a concept used in electrical engineering and computer science to describe a “machine” or program that can be in one of several states until something changes causing the state to change. A simple analogy would be a traffic light, with its red, yellow, and green lights. The signal will show one light at a time, starting with red, then going to green after a certain amount of time has elapsed, then yellow, then green after another amount of time has elapsed, then back to red.
PLATO III diagram Credit 15
The “resident program” Hanson built was called CATORES, short for “CATO resident.” Bitzer was right. Hanson’s code was a work of art, and would keep the PLATO III system running until 1973, when it was finally retired to make way for PLATO IV.
Even though Bitzer’s Boy Scouts were doing great work, particularly during the summers when he would pay them and give them specific projects, some of the senior staff at CSL expressed consternation and doubt about the value of having a bunch of kids running around in their fancy laboratory. “There were some senior people,” says Bitzer, “who objected to these students being paid to come in to do this creative work which they didn’t think was productive.” Alpert shared Bitzer’s views on bringing in bright people regardless of age—his own daughter was one of them. Alpert defended Bitzer’s kids but other senior lab people continued to object. “They insisted that these kids give a seminar defending the work that they were doing,” says Bitzer. “These kids would have to defend their work. The kids loved it. They got together and gave this four-hour presentation on the things they had done with computers, and it was impressive. It was impressive. Everybody was impressed, and never again did we have a question about our summer students.”
—
Sometime around 1964, while Hanson was an undergrad at Harvard, he heard about an upcoming seminar at MIT. The presenter: Don Bitzer. Hanson attended and witnessed the same sort of skepticism against Bitzer and PLATO that he himself had faced back at the JETS science fair, a disbelief that would haunt the PLATO project for its entire history. Bitzer began his presentation and encountered heckling from the MIT audience. “His trademark enthusiasm was mistaken for ‘used car salesman’ misrepresentation,” says Hanson, “and the elderly—MIT engineering professor?—man sitt
ing literally next to me in the third or fourth row kept muttering out loud about how everything Don said was completely theoretically impossible. He objected in particular to the idea that you could handle as many students at once as Bitzer claimed using a single processor; we all know now that this is easy using task-swapping in multitask operating systems. The 1604 resident program of which I was—or was at least allowed to believe I was—a principal architect accomplished all this by brute force in a very primitive way, but it did work; the fellow next to me paid me no attention when I suggested that not only was our speaker, Bitzer, not lying, but that I had written the code that did it….I was probably not very convincing as a very young-looking twenty-year-old.”
6
Gas and Glass
The PLATO project launched just six years after Skinner unveiled his first teaching machine, made out of a wooden box, at a convention of psychologists in 1954. What people thought a teaching machine could be—should be—had evolved at a lightning pace in the ensuing years. First there were the mechanical boxes. Then there were the scrambled textbooks. Then IBM toyed with using a computer to simulate a teaching machine. That was the starting point for PLATO’s team, whose stated goal was “to develop an automatic teaching system sufficiently flexible to permit experimental evaluation of a large variety of ideas on automatic instruction.” Such a system would need innovations in software and hardware. They were still grappling with late-night paper-coding sessions at Bitzer’s dining room table and pecking away in machine language on the ILLIAC and then the 1604. The 1960 PLATO I diagram depicted what components they needed to acquire or build. One implied aspect of their stated goal would remain elusive for years: scale. Bitzer’s view was not unlike the mantra of Silicon Valley start-ups decades later: Go big or go home. In three years, his fledgling team had built and demonstrated PLATO I and II, not only showing hints of what could be done with automatic teaching, but also attracting more professors to try out the system within their respective subject areas. PLATO II’s time-sharing feat demonstrated that two terminals could be plugged into the ILLIAC and used simultaneously by two students: a necessary milestone. PLATO III, now under way, was going to be another important milestone. But the dream lay beyond even PLATO III. It did not stop at two or even thirty-two connected terminals: the dream called for thousands, ultimately millions of student terminals scattered all over the country, in schools, colleges, universities, and homes.
Two components of the 1960 diagram, it was becoming clear, needed ditching: the “Storage Device” and the “TV Display.” Something had to be done about those things. As word got out about what Bitzer had in mind to replace them, some reacted by saying that he must be insane.
PLATO’s early design called for each student to use a makeshift keyboard and a cheap TV display fed by a closed-circuit video signal from an expensive, unreliable Raytheon storage tube. That design had two big problems. First: the video signal itself. Closed-circuit video signals require a lot of communications bandwidth. Every time you added another terminal to the system, that meant another display, which meant you needed that much more video bandwidth. Unlike 1960s broadcast television, where millions of people tuned in to one of a mere handful of channels to see one of a handful of programs, with PLATO every student would be working on something different, working at their own pace, typing their own unique answers to questions, interacting seemingly one-on-one with their automatic teacher. In essence, every PLATO user would be tuned in to their own “show.” That was fine on PLATO I and PLATO II, and would be tolerable with PLATO III, but what about a future PLATO with thousands, maybe millions of terminals? Video signals were not a viable solution for that kind of scale.
The second problem had to do with a consequence of the way TV displays worked. Cathode ray tubes (CRT) constantly redraw themselves, thirty times per second, because the CRT has no memory. Instead, every thirtieth of a second, it requires a fresh signal of information describing the complete picture, which the electron beam draws, line by line, over and over. Lose that signal, and the picture instantly goes away. When a TV screen was used as a student’s PLATO terminal display, the Raytheon storage tube acted as a temporary “memory” containing any text and graphics that were mixed together in the tube, then blasted over the closed-circuit cable to the TV’s CRT. But storage tubes were expensive, unreliable, and required one per student terminal. This approach would not support a very large-scale PLATO system. It was hard enough to support a handful of terminals: sometimes the images would begin to drift while the student was interacting with the material.
Lezlie Fillman had been hired in August 1961 as an educator to observe students working with some of the mathematics lessons, see how they were interacting and responding, and suggest areas for improvement. “We had to put everything that was text on film,” Fillman says, “and try to display that to students as the image was drifting across the screen, and the students would put their answers in, and by the time they got their answer in after they’d pondered the problem, the answer had kind of drifted out of the box.” The storage tube arrangement was tolerable as a proof of concept, but it was impractical for thousands of terminals in the future. An alternate solution had to be found.
When PLATO began, the notion of Moore’s Law was still five years out. Not until 1965 would Gordon Moore, then of Fairchild Semiconductor and later one of the founders of Intel Corporation, write his famous article, “Cramming More Components onto Integrated Circuits” in Electronics magazine, from which emerged the principle known as Moore’s Law, which stated that the number of transistors and integrated circuits etched into a silicon wafer doubles every two years. As long as this trend continued, Moore argued, computer components would continue to get faster and cheaper. He even believed that this trend would lead to computers cheap enough for home use. Moore only had a few years’ worth of actual data upon which to base his prediction, since integrated circuits had only come onto the scene in the late 1950s. But the trend was so clear, so obvious, he believed that there was nothing to stop it, so it would continue for at least ten years. It turns out that trend would continue all the way until today.
It was not yet apparent to the PLATO team in 1960 that Moore’s Law was already under way. To appreciate what they were up against, consider how prohibitively expensive random-access memory (RAM) was for a computer then: $2 per bit. Take the typical 8 gigabytes of RAM found in a single present-day laptop: at $2 per bit, that much RAM in the early 1960s would cost $68 billion in 1962 dollars, or nearly $500 billion in 2017 dollars. That’s for one laptop.
Consider video RAM. Instead of using storage tubes, Bitzer could have built a box with some RAM in it that served as the memory of the display fed by the computer and then sent out to the TV screen. There was a simple reason why Bitzer did not go this route. If each PLATO student workstation were to rely on video RAM, each terminal would have cost over half a million dollars (nearly $5 million in 2016 dollars). Video RAM was therefore out of the question. They knew the storage tube route was temporary; something else would have to be done for the long haul. What other technologies were there to consider? There was no existing affordable digital memory and display technology.
Computer graphics displays were already a hot topic by 1963, but they were typically relegated to massive government-backed projects. For example, the Cold War–era SAGE project to detect incoming intercontinental ballistic missiles had gigantic graphics displays but that system cost about $10 billion—more than the Manhattan Project—to design and build. Much was learned from SAGE, including the numerous benefits of interactive graphical displays. CSL itself had worked on some for its air traffic control projects. But neither SAGE nor Cornfield solved how to put twenty interactive graphics terminals in a classroom in a school, and then repeat that many times over in other schools, and do it in a way that didn’t bankrupt the school, the town, the state, or the nation.
So, how were they going to scale PLATO but keep the cost of the terminal displays
manageable?
As early as 1961, Bitzer began looking around for alternatives. There was one that had promise: gas plasma, where you send a current of electricity strong enough to charge a gas to the point where it glows. If you created a grid of super-thin wires and ran each intersection of wires over tiny “cells” of gas, in theory you could at any one moment charge any cell you wanted in that matrix, with the combined glowing dots forming a picture. The idea had been kicked around for years. But success had so far eluded everyone.
Causing an enclosed area of gas to glow by running a sufficiently high voltage through it was an old idea. Neon signs, invented by Georges Claude in France fifty years earlier, worked on this principle. Just before the turn of the century, Sir William Ramsay of Scotland and Morris W. Travers had discovered the element neon. They injected neon into a tube that had a metal anode (the positive piece where the electrical charge flows in) and cathode (the negative piece that the electrons flow to, across the gas), powered it up, and discovered that the gas glowed the color of fire. By filling the tubes with different mixtures of neon and other gases, and then running voltage through them, you could control which color would glow. But there was something else about neon. In his 1904 Nobel Prize lecture, Ramsay said that neon “showed a spectrum characterized by a brilliant flame-colored light.” Travers remarked that “it was a sight to dwell upon and never forget.”
They’d discovered the Orange Glow.
—
By the 1950s, Burroughs Corporation brought out a product called the Nixie tube, a small glass tube with a single wire mesh anode and a number of separate cathodes, each ingeniously shaped into numbers, letters, or other symbols. Run a current through the anode and a selected cathode, and suddenly a visible number or letter would blaze to life with an Orange Glow. Over the coming years, Nixie tubes would appear in all kinds of devices, from industrial and technical equipment to indicators in elevators to let you know what floor you’re on.