In the early 1960s, interactive computing began to spread out from the few tender saplings nurtured at Lincoln Lab and MIT – spread in two different senses. First, the computers themselves sprouted tendrils, that reached out across buildings, campuses, and towns to allow users to interact at a distance, and to allow many users to do so at the same time. These new time-sharing systems blossomed, accidentally, into platforms for the first virtual, on-line societies. Second, the seeds of interactivity spread across the country, taking root in California. One man was responsible for sowing those first transplants, a psychologist named J.C.R. Licklider.
Joseph Carl Robnett Licklider — known to friends as “Lick” — specialized in psychoacoustics, a field that bridged the gap between imaginary states of mind and the measurable physiology and physics of sound. We met him briefly before, as a consultant in the 1950s FCC Hush-a-Phone hearings. He had honed his skills at the Psycho-Acoustics Laboratory at Harvard during the war, devising techniques to improve the audibility of radio transmissions inside noisy bombers.
Like so many American scientists of his generation, he found ways to continue to meld his interests with military needs after the war, but not because he had a special interest in weaponry or national defense. The only major civilian sources of money for scientific research were two private institutes founded by the industrial titans of the turn of the century: the Rockefeller Foundation and Carnegie Institute. The National Institutes of Health had only a few million dollars to spend, and the National Science Foundation was created only in 1950, with a similarly modest budget. To get funding for interesting science and technology in the 1950s, your best bet was the Department of Defense.
So, in 1950, Licklider joined an acoustics lab at MIT directed by the physicists Leo Beranek and Richard Bolt, and funded almost entirely by the U.S. Navy. Once there, his expertise on the interface between the human senses and electronic equipment made him a natural early recruit to MIT’s new air defense project. As part of the Project Charles study group, tasked with figuring out how to implement the Valley Committee air defense report, Licklider pushed for the inclusion of human factors research, and got himself appointed co-director of radar-display development for Lincoln Laboratory.
There, at some point in the mid-1950s, he crossed paths with Wes Clark and the TX-2, and instantly caught the interactive computing bug. He was captivated by the idea of being in total control of a powerful machine that would instantly solve any problem addressed to it. He began to develop an argument for “man-computer symbiosis,” a partnership between human and computer that would amplify humankind’s intellectual power, in the same way that industrial machines had amplified its physical power. He noted that some 85% of his own work time2
…was devoted mainly to activities that were essentially clerical or mechanical: searching, calculating, plotting, transforming, determining the logical or dynamic consequences of a set of assumptions or hypotheses, preparing the way for a decision or an insight. Moreover, my choices of what to attempt and what not to attempt were determined to an embarrassingly great extent by considerations of clerical feasibility, not intellectual capability. …the operations that fill most of the time allegedly devoted to technical thinking are operations that can be performed more effectively be machines than by men.
The overall concept did not stray too far from Vannevar Bush’s Memex, an intellectual amplifier that he sketched in his 1945 “As We May Think,” though Bush’s mix of electro-mechanical and electronic components gave way to a pure electronic digital computer as the central intellectual engine. That computer would use its immense speed to shoulder all the brute-force clerical work involved in any scientific or technical project. People would be unshackled from that drudgery, freed to spend all of their attention on forming hypotheses, building models, and setting goals for the computer to carry out. Such a partnership would provide tremendous benefit to researchers such as himself, of course, but also to national defense, by helping American scientists stay ahead of the Soviets.
Soon after this Damascene encounter, Lick brought his new devotion to interactive computing to a new position, at a consulting firm run by his old colleagues, Bolt and Beranek. As a sideline from their academic physics work, the two had dabbled with consulting projects for years; reviewing, for instance, the acoustics of a movie house in Hoboken, New Jersey. Landing the acoustics analysis for the new United Nations building in New York City, however, brought them a slew of additional work, and so they decided to leave MIT and consult full-time. Having acquired a third partner in the meantime, architect Robert Newman, they now went by Bolt, Beranek and Newman (BBN). By 1957, having grown into a mid-sized firm with dozens of employees, Beranek felt that they risked saturating the market for acoustics work. He wanted to extend their expertise beyond sound to the full range of interaction between humans and the built environment, from concert halls to automobiles, across all the senses.
And, so, naturally, he sought out his old colleague Licklider, and recruited him on generous terms as the new vice-president of psychoacoustics. But Beranek had not reckoned with Licklider’s wild enthusiasm for interactive computing. Rather than a psycho-acoustics expert, he had acquired… not a computer expert, exactly, but a computer evangelist, eager to bring others to the light. Within the year, he had convinced Beranek to lay out tens of thousands of dollars buy a computer, a meager little thing called the LGP-30, made by a defense contractor called Librascope. Having no engineering expertise himself, he brought on another SAGE veteran, Edward Fredkin, to help configure the machine. Despite the fact that the computer did little but distract Licklider from his real work while he tried to learn to program it, he convinced the partners to put down still more money3 to buy a much better computer a year-and-half later: DEC’s brand new PDP-1. Licklider sold B, B, and N on the idea that digital computing was the future, and that somehow, sometime, their investment in building expertise in the field would pay off.
Shortly thereafter, Licklider, almost by accident, found himself in the perfect position for spreading the culture of interactivity across the country, as head of a new government computing office.
In the Cold War, every action brought it’s reaction. Just as the first Soviet atomic bomb had spurred the creation of SAGE, so did the first Soviet satellite in orbit, launched in October 1957, trigger a flurry of responses from the American government. All the more so because, while the Soviets had trailed the U.S. by four years in exploding a fission weapon, in rocketry it seemed to have leaped ahead, beating the Americans to orbit (by about four months, as it turned out).
One of the responses to Sputnik to create, in early 1958, an Advanced Research Projects Agency (ARPA) within the Defense department. In contrast to the more modest sums available for civilian federal science funding, ARPA was given an initial budget of $520 million, three times the budget of the National Science Foundation, which had itself been tripled in size in response to Sputnik.
Though given a broad charter to work on any advanced projects deemed fit by the Secretary of Defense, it was initially intended to focus on rocketry and space – a vigorous answer to Sputnik. By reporting directly to the Secretary of Defense, ARPA was to rise above debilitating and counterproductive inter-service rivalries and develop a unified, rational plan for the American space program. But in fact, all of its projects in that field were soon stripped away by rival claimants4: the Air Force had no intention of giving up control over military rocketry, and the National Aeronautics and Space Act, signed in July 1958, created a new civilian agency to take over all non-weaponized ventures into space. Having been created, ARPA nonetheless found reasons to survive, acquiring major research projects in ballistic missile defense and nuclear test detection. But it also became a general workshop for pet projects that the various armed services wanted investigated. Intended to be the dog, it had instead become the tail.
The first foray by ARPA into computing was, in a sense, busy work. In 1961, the Air Force had two idle assets on its hands and needed something for them to do. As the first SAGE direction centers neared deployment, the Air Force had brought on, RAND Corporation, based in Santa Monica, California, to train personnel and prepare the twenty-odd computerized air defense centers with operational software. RAND spun off a whole new entity, System Development Corporation (SDC), just to handle this task. SDC’s newly acquired software expertise was a valuable resource for the Air Force, but SAGE was winding down and they were running out of work to do. The Air Force’s second idle asset was a (very expensive) surplus AN/FSQ-32 computer which had been requisitioned from IBM for SAGE but turned out to be unneeded. The Department of Defense solved both problems by assigning ARPA new research task of command-and-control, to be inaugurated with a $6 million grant to SDC to study command-and-control problems using the Q-32.
ARPA soon decided to regularized this research program as part of a new information processing research office. Around the same time, it had also received a new assignment to create a program in behavioral science. For reasons that are now obscure, ARPA leadership decided to recruit J.C.R. Licklider to oversee both programs. The idea may have come from Gene Fubini, director of research for the Department of Defense, who would have known Lick from his time working on SAGE.
Like Beranek, Jack Ruina, then head of ARPA, had no idea what he was in for when he brought Lick in for an interview. He thought he was getting a behavioral science expert with a dash of computing knowledge on the side. Instead he got the full force of the man-computer symbiosis vision. Computerized command-and-control required interactive computing, Licklider argued, and thus the primary thrust of ARPA’s command-and-control research program should be to push forward the cutting edge of interactive computing. And to Lick that meant time-sharing.
Time-sharing systems originated with the same basic principle as Wes Clark’s TX series: computers should be convenient for the user. But unlike Clark, the proponents of time-sharing believed that a single computer could not be used efficiently by a single person. A researcher might sit for several minutes pondering the output of a program before making a slight change and re-running it. During that interval the computer would have nothing to do, its great power going to waste, at great expense. Even the hundred-millisecond intervals between keystrokes loomed as vast gulfs of wasted time for the computer, in which thousands of computations could have been performed.
All of this processing power need not go to waste, if it could instead be shared among many users. By slicing up the computer’s attention so that it could serve each user in turn, the computer designer could have his cake and eat it – provide the illusion of an interactive computer completely at the user’s command, without wasting most of the capacity of a very expensive piece of hardware.
The concept was latent in SAGE itself, which could serve dozens of different operators simultaneously, each monitoring his own sub-sector of airspace. After meeting Clark, Licklider immediately saw the potential to combine the shared user base of SAGE with the interactive freedom of the TX-0 and TX-2 into a potent new mix, and this formed the basis of his advocacy for man-computer symbiosis, which he proposed to the Department of Defense in a 1957 paper entitled “The Truly Sage System, or Toward Man-Machine System for Thinking.” In that paper described a computer system for scientists very similar in structure to SAGE, with a light-gun input, and “simultaneous (rapid time-sharing) use of the machine computing and storage facilities by many people.”
Licklider, though, lacked the engineering chops to actually design or build such a system. He managed to learn the basics of programming at BBN, but that was as far as his skills went. The first person to reduce time-sharing theory to practice was John McCarthy, an MIT mathematician. McCarthy wanted constant access to a computer in order to craft his tools and models for manipulating mathematical logic, the first steps, he believed, towards artificial intelligence. He put together a prototype 1959, consisting an interactive module bolted onto the university’s batch-processing IBM 704 computer. Ironically, this first “time-sharing” installation had only one interactive console, a single Flexowriter teleprinter.
By the early 1960s, however, the MIT engineering faculty as a whole had become convinced that they should invest wholesale in interactive computing. Every student and faculty member with an interest in programming who got their hands on it, got hooked. Batch-processing made very efficient use of the computer’s time, but could be hugely wasteful of the researcher’s – the average turnaround time for a job on the 704 was over a day.
A university-wide committee formed to study the long-term solution for the growing demand for computing resources at MIT, and time-sharing advocates predominated. Clark fought a fierce rearguard action, arguing that the move to interactivity should not mean time-sharing. As a practical matter, he argued that time-sharing meant sacrificing interactive video displays and real-time interaction, crucial features of the projects he had been working on with the MIT biophysics lab. But more fundamentally, Clark seemed to have a deep philosophical resistance to the idea of sharing his workspace. As late as 1990, he refused to connect his computer into the Internet, and stated outright that networks “are a mistake” and “don’t work.”5
He and his disciples formed a sub-sub-culture, a tiny offshoot within the already eccentric academic culture of interactive computing. But their arguments in favor of small, un-shared computer workstations did not find purchase with their colleagues.6 Given the cost of even the smallest individual computer at the time, such an approach seemed economically infeasible to the other engineering faculty. Moreover, most assumed at that time that computers – the intellectual power plants of a dawning information age – would benefit from economies of scale, in the same way that physical power plants did. In the spring of 1961, the final report of the long-range study committee sanctioned large-scale time-sharing systems as the way of the future at MIT.
By that time, Fernando Corbató, known to colleagues as “Corby,” was already working to expand the scope of McCarthy’s little experiment. A physicist by training, he learned about computers while working on Whirlwind in 1951, while a grad student at MIT. 7 After completing his doctorate he became an administrator for MIT’s newly formed Computation Center, built around the IBM 704. Corbató and his team (initially Marge Merwin and Bob Daley, two of the best programmers in the Center) called their time-sharing system CTSS, for Compatible Time-Sharing System – so-called because it could run simultaneously with the 704’s normal batch-processing operations, seamlessly snatching computer cycles for users as needed. Without this compatibility the project would indeed have been impossible, because Corby had no funding for a new computer on which to build a time-sharing system, and shutting down the existing batch-processing operation was not an option.
At the end of 1961, CTSS could support four terminals. By 1963, MIT hosted two instances of CTSS on 3.5 million dollar transistorized IBM 7094 machines, with roughly ten times the memory capacity and processing power of their 704 predecessor. The system’s supervisor software passed through the active users in a roughly round-robin fashion8, servicing each for a fraction of a second before moving on to the next. Users could store programs and data in their own private, password-protected area in the computer’s disk storage, for later use.9
Each computer could serve roughly twenty terminals. That was enough to not only support a couple of small terminal rooms, but also to begin spreading access to the computer out across Cambridge. Corby and other key individuals had office terminals, and, at some point, MIT began providing home terminals to technical personnel so that they could do system maintenance at odd hours without having to come on-campus. All of these early terminals consisted of a typewriter with some modifications to support reading from and writing to a telephone line, plus a continuous feed of perforated paper instead of individual sheets. Modems connected the terminals via the telephone system to a private exchange on the MIT campus, via which they could reach the CTSS computer. The computer thus extended its sensory apparatus over the telephone, with signals that went from digital to analog and back. This was the first stage in the integration of computers into the telecommunications network. The mixed state of AT&T with respect to regulation facilitated this integration. The core network was still regulated, and required to provide private lines at fixed rates, bu a series of FCC decisions had eroded its control over the periphery, and thus it had very little say over what was attached to those lines. MIT needed no permission for its terminals.
The desired goal of Licklider, McCarthy, and Corbató had been to increase the availability of computing power to individual researchers. They had chosen the means, time-sharing, for purely economic reasons – no one could imagine buying and maintaining a computer for every single researcher at MIT. But this choice had produced unintended side-effects, which could never have been realized within Clark’s “one man, one machine” paradigm. A common file area and cross-links between users accounts allowed users to share, collaborate, and build on each other’s work. In 1965, Noel Morris and Tom Van Vleck facilitated this collaboration and communication, with a MAIL program that allowed users to exchange messages. When a user sent a message, the program appended it to a special mailbox file in the recipient’s file area. If a user’s mailbox file had any contents, the LOGIN program would indicate it with the message “YOU HAVE MAIL BOX.” The contents of the machine itself were becoming an expression of the community of users, and this social aspect of time-sharing became just as prized at MIT as the initial premise of one-on-one interactive use.
Lick, having accepted ARPA’s offer and left BBN to take command of ARPA’s new Information Processing Techniques Office (IPTO) in 1962, quickly set about doing exactly what he had promised – focusing ARPA’s computing research efforts on spreading and improving time-sharing hardware and software. He bypassed the normal process of waiting for research proposals to arrive on his desk, to be authorized or rejected, instead going into the field himself and soliciting the research proposals he wanted to authorize.
His first step was to reconfigure the existing SDC command-and-control research project in Santa Monica. Word came down to SDC from Lick’s office that they should curtail their work on command-and-control research, and instead focus their efforts on turning their surplus SAGE computer into a time-sharing system. According to Lick, the basic substrate of time-shared man-machine interaction must come first, and command-and-control would follow. That this prioritization aligned with his own philosophical interests was a happy coincidence. Jules Schwartz, a SAGE veteran, architected the new time-sharing system. Like its contemporary, CTSS, it became a virtual social space, including among its commands a DIAL function for direct text messaging between on-line users, as can be seen in this example exchange between John Jones and a user identified by the number 9:
Next, to provide funding for the further development of time-sharing at MIT, Licklider found Robert Fano to lead his flagship effort: Project MAC, which lasted into the 1970s.10 Though the designers initially hoped that the new MAC system would support 200 simultaneous users or more, they had not reckoned with the ever-escalating sophistication and complexity of user software, which easily consumed all improvements in hardware speed and efficiency. When launched to MIT in 1969, the system could support about 60 users on its two central processing units (CPUs), roughly the same number per CPU as CTSS. However, the total community of users was much larger than the maximum active load at any given time, with 408 registered users in June 1970.11
Project MAC’s Multics system software also embodied several major advances in design, some of which are still considered advanced features in today’s operating systems: a hierarchical file system with folders that could contain other folders in a tree structure; a hardware-enforced distinction between execution in user and system mode; dynamically linked programs that could pull in software modules as needed during execution; and the ability to add or remove CPUs, memory banks, or disks without bringing down the system. Ken Thompson and Dennis Ritchie, programmers on the Multics project, later created Unix (a pun on the name of its predecessor) to bring some of these concepts to simpler, smaller-scale computer systems.
Lick planted his final seed in Berkeley, at the University of California. Project Genie12, launched in 1963, begat the Berkeley Timesharing System, a smaller-scale, more commercially-oriented complement to the grandiose Project MAC. Though nominally overseen by certain Cal faculty members, it was graduate student Mel Pirtle who really led the time-sharing work, aided by other students such as Chuck Thacker, Peter Deutsch, and Butler Lampson. Some of them had already caught the interactive computing bug in Cambridge before arriving at Berkeley. Deutsch, son of an MIT physics professor and the prototypical computer nerd, implemented the Lisp programming language on a Digital PDP-1 as a teenager before arriving at Cal as an undergrad. Lampson, for his part, had programmed on a PDP-1 at the Cambridge Electron Accelerator as a Harvard student. Pirtle and his team built their time-sharing system on a SDS 930, made by Scientific Data Systems, a new computer company founded in 1961 in Santa Monica.13
SDS back-integrated the Berkeley software into a new product, the SDS 940. It became one of the most widely used time-sharing systems of the late 1960s. Tymshare and Comshare, companies that commercialized time-sharing by selling remote computer services to others, bought dozens of SDS 940s for their customers to use. Pirtle and his team also decided to try their hand in the commercial market, founding Berkeley Computer Corporation (BCC) in 1968, but BCC fell into bankruptcy in the 1969-1970 recession. Much of Pirtle’s team ended up at Xerox’s new Palo Alto Research Center (PARC), where Thacker, Deutsch and Lampson contributed to landmark projects such as the Alto personal workstation, local networking, and the laser printer.
Of course, not every time-sharing project of the early 1960s sprung from Licklider’s purse. News of what was happening at MIT and Lincoln Labs spread through the technical literature, conferences, academic friendships, and personnel transfers. Through these channels other, windblown, seeds took root. At the University of Illinois, Don Bitzer sold his PLATO system to the Department of Defense as a means of reducing the cost of technical education for military personnel. Clifford Shaw created the JOHNNIAC Open Shop System (JOSS), which the Air Force funded in order to improve the ability of RAND employees to perform quick numerical analyses.14 The Dartmouth Time-Sharing System had a direct connection to events at nearby MIT, but was otherwise the most exceptional, being a purely civilian-funded effort sponsored by the National Science Foundation, on the basis that experience with computers would be a necessary part of a general education for the next generation American leaders.
By the mid-1960s, time-sharing had not taken over the computing ecosystem. Far from it. Traditional batch-processing shops predominated in sales and use, especially outside university campuses. But it had found a niche.
In the summer of 1964, some two years after arriving at ARPA, Licklider moved on again, this time to IBM’s research center north of New York City. For IBM, shocked to have lost the Project MAC contract to rival computer maker General Electric after years of good relations with MIT, Lick would provide some in-house expertise in a trend that seemed to be passing it by. For Lick, the new job offered an opportunity to convert the ultimate bastion of conventional batch computing to the new gospel of interactivity.15
He was succeeded as head of IPTO by Ivan Sutherland, a young computer graphics expert, who was succeeded in turn, in 1966, by Robert Taylor. Licklider’s own 1960 “Man-Machine Symbiosis” paper had made Taylor a convert to interactive computing, and he came to ARPA at Lick’s recommendation, after a stint running a computer research program at NASA. His personality and background formed him in Licklider’s mold, rather than Sutherland’s. A psychologist by training and no technical expert in computer engineering, he compensated with enthusiasm and clear-sighted leadership.
One day in his office, shortly after taking over the IPTO, a thought dawned on Taylor. There he sat, with three different terminals, through which he could connect to the three ARPA-funded time-sharing systems in Cambridge, Berkeley, and Santa Monica. Yet they did not actually connect to one other – he had to intervene physically, with his own mind and body, to transfer information from one to the other. 16
The seeds sewn by Licklider had borne fruit. He had created a social community of IPTO grantees, that spanned many computing sites, each with its own small society of technical experts, gathered around the hearth of a time-sharing computer. The time had come, Taylor thought, to network those sites together. Their individual social and technical structures, once connected, would form a kind of super-organism, whose rhizomes would span the entire continent, reproducing the social benefits of time-sharing on the next higher scale. With that thought began the technical and political struggle that would give birth to ARPANET.
Richard J. Barber Associates, The Advanced Research Projects Agency, 1958-1974 (1975)
Katie Hafner and Matthew Lyon, Where Wizards Stay Up Late: The Origins of the Internet (1996)
Severo M. Ornstein, Computing in the Middle Ages: A View From the Trenches, 1955-1983 (2002)
M. Mitchell Waldrop, The Dream Machine: J.C.R. Licklider and the Revolution That Made Computing Personal (2001)