This is why—driving a radar dish following the position of the missile’s reflection through the sky—these aligned motors do not only track the position of the bomb but also, through a web of relays and further connected motors, align and position the barrel of an artillery cannon, calculating and anticipating the missile’s trajectory. Triggering the launch of a supersonic artillery shell, the web of signals creates an improbable, microsecond encounter of destruction.^▶2

This assembly did not have a single name or creator. It was a novel combination of interconnected technologies—among them the SCR-584 radar and M-9 gun director—produced by an equally interconnected set of academic and industrial organizations over the course of World War II. The work of its creation centered around the Radiation Lab (Rad Lab) at MIT but also involved the Sperry Gyroscope Company, the Ford Instrument Company, Chrysler, and Bell Telephone Laboratories, and the US Military’s own feuding branches and technicians.

The result was a ballet of systems, mechanisms, and organizations—but also human senses. A radar operator was a crucial observer of a cathode ray projection of the radar dish’s signals, sorting real “pips” from noise in the system and initiating the tracking process by “pip matching” horizontal and vertical tracking mechanisms, thus distinguishing the target signals from background noise.

Locks, Interlocks, and the Legibility of Control

The history of this electromechanical choreography leads back not just to the extensive logistics of air defense but also to a miniature world an ocean away, lodged in the space between two oceans. 

In 1914, a visitor to one of the three control rooms of the locks of the Panama Canal would have been confronted with a gargantuan miniature: a scale model of the enormous, concrete canal lock in refined marble, brass, silk and steel. As The New York Times (1914) explained:

The locks are operated by electricity and the controlling switchboards reproduce in miniature on the board, by synchronous indicators, every detail of operation so that the man in charge sees the complete movement of all gates, valves, fender chains, &c., reproduced before his eyes, eliminating any errors which might otherwise occur. And in addition to this the control switches are so interlocked that an improper sequence of operations is impossible.

At the time, electric power was a rarity, even in factories. With the world still mostly lit by fire, it had been just a year since the first manufacturing plant in the United States was converted from steam and mechanical power. Yet the far greater precision that electronic controls offered over driveshafts and steam governors was already apparent. As the popular historian of the canal, David McCullough (1977, 601), concludes, “The chief virtue of electricity was in the degree of control it afforded.”

The Task of Theory

As the interchange and overlay of self-synchronized motion became more complex, so too did the delicate connections and adjustments necessary to keep their operation accurate and responsive. Unexpected behavior crept into control systems as an emergent property of their growing complexity; an example would be “hunting” an unexpected, shivery back-and-forth movement in a large mechanism as it seeks equilibrium through over-complicated feedback loops  

As technicians developed increasingly sophisticated practical techniques to manage electromechanical controls, their expertise collected into new theoretical frameworks on the way. As argued by historian David Mindell, however, these theories—information theory and cybernetics—did not presage the invention of control systems or the mindset they created. Rather, they emerged, fittingly, in a kind of constant feedback with the systems’ growing technological complexity (see Mindell 2002, 6.; a related argument is found in Galison 1994).

Even as they followed the developing physical systems’ logic, the abiding goal of theories of control and communication that emerged from the wartime milieu—Norbert Wiener’s Cybernetics, or Control and Communication in Human and Machine of 1948, and Claude Shannon and Warren Weaver’s Mathematical Theory of Communication of 1949—was to try to separate the signals conveyed by interconnected devices from their physical circumstance, or even presence, in a framework that was electronic, mechanical, or neurological (Shannon and Weaver 1949; Wiener 1948). Particularly when so liberated, these ideas were free to exert a deep influence on the rest of postwar reality.

The project did not literally involve electromechanical inputs and outputs—though it was influenced by them and would influence projects that did. Rather, it reconceptualized the urban pedestrian as a radar operator or even guided device themselves, receiving inputs from the environment and, if adequate signals were interpreted correctly, effectively navigating to a target. The project was conceived by two MIT faculty, artist György Kepes and architect and urban designer Kevin Lynch. It would appear in published form seven years later as a book authored by Lynch, The Image of the City (Lynch 1960).^▶4

Camouflage and Cambridge

Just as in the history of electromechanical motor-control systems, the concerns of Kepes and Lynch can be traced to their direct experiences of movement, control, and administration in wartime. Kepes’s seminal 1944 book Language of Vision is exemplary. Instrumental to his invitation to launch a program in “visual studies” at MIT, the book’s analysis of pictorial form and arguments for resulting approaches to “coherence” in the postwar environment deeply shaped The Image of the City (Kepes et al. 1944). Kepes traced his first book’s origin to a single experience during his service on a “camouflage committee” advising the Army in 1942: He and other Chicago Bauhaus faculty joined a fateful tour above Chicago in a military airplane that Kepes credited with inspiring the book. “I became newly alert to the . . . environment,” he later recalled, “by looking at camouflage from the air, at what light patterns could do” (ADC 1981).

Lynch spent the years 1941 to 1946 in a more typical military environment— in the US Army Corps of Engineers, supporting the work of aircraft throughout the Pacific Conflict, from Peleliu through the Philippines and Japan. Mustered out in 1946, Lynch capped an eclectic assembly of incomplete undergraduate training (Yale and Taliesin for Architecture, RPI for engineering) with MIT’s new BA in city planning in 1947 (Southworth and Banerjee 1995). Decamping to Greensboro, NC, for a brief period as a city planner in public service, Lynch was called back to MIT by an offer to join the faculty in 1949. But, in an important sense, he still had not left the Army. During Lynch’s studies at MIT, when he returned in 1949, and through the 1950s, the Institute was not just a military-tinged environment; it was in essential ways indistinguishable from the encampments that Lynch had lived in and helped construct in the previous decade. Uniforms were ubiquitous—MIT dropped its Reserve Officers’ Training Corps requirement for all male undergraduates only in the mid-1950s. Temporary military structures filled the central spine of the campus and housed the continued work of the Rad Lab and other military-industrial collaborations.

Trumbull was hired initially as an animator for 2001, and his first months in England were spent mass-producing the flashing displays of the Discovery spacecraft interior. But it was Trumbull’s familiarity with machining and control systems that proved essential to his growing role in the film’s complex visual effects. This was not a formal education but the product of a childhood spent in the workshop of his father, Donald Trumbull. Who had his own history working in visual effects—including 1939’s The Wizard of Oz—before wartime found him a new career in LA’s burgeoning aviation industry, where he crafted pneumatic and hydraulic systems, as well as epoxy-based part manufacturing. His tinkering resulted in seventeen patents and an indelible influence on his son, which can be traced, quite literally, in the movement of stars and spacecraft in 2001.^▶7

Creeping Toward Infinity

Kubrick’s insistence on detailed clarity and detail across the 70 mm film frame meant that as many visual effects as possible needed to be produced within 2001’s “Cinerama” camera itself—each frame a long exposure across a ruthlessly sharp aperture. The Discovery was modeled at the scale of a building: fifty-five feet long, with extensive foundations—long concrete pilings cast into the floor of Borehamwood’s Studio 3, all to ensure its complete stillness. Surrounding the model was a deep-black velvet curtain on velvet-covered scaffolding—the inky void on which star fields would be introduced by double exposure. The shot’s motion was provided by the camera, slowly driven by a Trumbull-engineered assembly of self-synchronizing motors, inching forward between countless individual frames.

Unlike previous, more primitive versions of in-camera visual effects, this electromechanical system enabled consistent, repeatable motions between takes. This was essential for the multiple passes needed to expose different parts of the film frame—incorporating separate exposures, for example, for the ship’s illuminated body and tiny scenes of interiors projected onto its windows. The system also had specific limitations. As revealed during the use of Selsyn motors in military control systems, the accuracy of a synchronized, analog motor-control system declines with increasing degrees of motion. As a result, Kubrick and Trumbull agreed on a simple, linear trajectory for the Discovery, produced by driving the camera motion on a long, straight track set obliquely to the giant model, propelled by a twenty-foot drive gear machined in Detroit and airfreighted to the set. Reporting on 2001’s technical innovations, American Cinematographer would term the result “The Mechanical Monster with the Delicate Touch” (Lightman 1968).

The long exposures and low aperture needed to capture the ship model in focus also limited the interval between frames: with any substantial speed, the lack of motion blur in the crisp, individual film frames began to undermine the illusion of motion. Steady, languid progress became the film’s signature. A further contributor to the stately speed of the Discovery were the moving star fields that were added to the dark, velvet background of each frame of the Discovery before it was developed.^▶8 With their sideways shift a subtle contrast to the movement of the spacecraft, these bright pinpricks (which were in reality a fine mist of Trumbull-applied airbrush paint on black cardboard) gave much-needed depth to the frame. Yet, beyond a ponderous speed, the fine paint stars would unexpectedly shimmer and separate, giving an illusion of doubling or tripling across the transition between frames. In Trumbull’s words, the “speed limit” of the film was set (Lightman 1968).

Sausages and Stargates

Trumbull’s Selsyn-guided machinery would leave its mark on many more of the film’s key sequences. A so-called “sausage machine” enabled the synchronized compositing of separately recorded sequences on a film-animation stand—including the background starscapes to model shots.^▶9 Apart from featured shots where Discovery moved obliquely, it was the “sausage” machine that slowly moved 8x10" photographic transparencies of models and planets across hundreds of film exposures, producing motion in languid elevation.

Aloft on Organ Pipes

The simulators of Apollo have as their first point of origin an improbable lineage of pipe organs and pianolas. These mechanisms were repurposed by Edwin Link, scion of a Binghamton, NY, pipe-organ company, to create what he patented in 1931 as “An apparatus . . . in simulation of an airplane” (Link Jr. 1931). Link successfully marketed his “trainer” to the Navy and Army Air Forces in the 1930s, and by World War II the devices were a ubiquitous part of military flight training. Yet the complex organ-pipe systems of such devices meant each trainer needed to be kept in “tune” manually, and each aircraft simulated needed to have a dedicated pneumatic system to capture its motion and behavior. As a result, the wartime research establishment explored improving pneumatic simulators with the same synchronized controls as guided artillery.

In 1943, the head of the Navy’s “Special Devices Desk” in charge of simulators (whose use was termed “synthetic training”), Commander Luis de Florez, went one step further, commissioning a group of MIT faculty to study the creation of an electronically driven trainer that could be switched between the behaviors of different aircraft—or even anticipate the behavior of aircraft not yet constructed (Redmond and Smith 1980). In 1944, MIT’s Servomechanisms Laboratory—a group whose pioneering work earlier in the war had solved fundamental problems of feedback and tracking in the application of self-synchronizing motor controls to aircraft defense—was commissioned to produce the device (see Mindell 2002, 207–30). In December, laboratory director Gordon S. Brown assigned the construction of “Device 2-K, Aircraft Stability and Control Analyzer” (ACSA) to one of his associate directors, Jay Forrester (Redmond and Smith 1980, 13).

Sowing Whirlwind

By the time Forrester stepped down from leading the project over a decade later, his allotted task had changed completely. Yet it would also have returned to the Lab’s air-defense roots. In 1946, Forrester abandoned the analog control methods that had defined the Servo Lab’s expertise in favor of new techniques of digital computing; at the time the province of punch cards, digital methods had never been engineered for the real-time computing needed for simulation (Redmond and Smith 1980, 41). Two years later, the surplus B-24 cockpit that had been set aside for the simulator’s use was itself scrapped; the project had a new name, “Whirlwind,” and a new focus—the real-time digital computer alone (Redmond and Smith 1980, 60).

After another two years, the Navy’s funds and patience near-exhausted, Forrester found new funding for the project in the Air Force’s Air Defense System Engineering Committee. This group was charged with protecting the United States against the perceived threat of Soviet nuclear bombing. Now, the real-time computer would become the core of America’s continental defense, the so-called Semi-Automatic Ground Environment or SAGE (Redmond and Smith 1980).

Harvesting SAGE

Deployed at twenty-three sites across North America from the 1950s to the 1980s, SAGE was built to protect the United States against airborne attack by Soviet bombers. Each installation interpreted radar signals from across a large sector of the United States to create a digital map of moving aircraft and sort and identify potential hostile aircraft. Through its digital interface, SAGE operators interacted with this digital model of the airspace and could deploy surface-to-air defenses against any potential target. Initially, these commands traveled by automated teleprinters to surface-to-air missile batteries; later in the system, the launch systems were fully connected to SAGE’s consoles.

As a system designed to protect against bomber attack, SAGE was rendered practically obsolete as soon as the Soviet Union began to produce a ballistic missile that could reach the United States, the R-7, in 1956: the speed and near-vertical approach of the ballistic missile rendered it invulnerable to radar-based defense.^▶11 Nevertheless, the system remained operational, including all its computers and mainframes, until the 1980s.

Twin Spaces

To better integrate the hundreds of hours of simulation needed for each mission, three of these simulators were located adjacent to and wired into Houston’s Mission Control. There, what often seems a singular environment was really two stacked, identical control spaces in MSC Building 30, with one housing the simulation of upcoming missions in the Gemini and Apollo sequence even while actual launches were being supervised in its twin (see de Monchaux 2011, chap. 12 and 18).

Starting from Gemini, the IBM 7094 mainframes running Mission Control were programmed to run calculations on active missions and simulate upcoming missions at the same time. When IBM’s new System/360 mainframes replaced the 7094 for Apollo, the computers would simulate not only their own operation but would contain virtual versions of other computers—from the circuits girdling the vast Saturn V rocket to the small Apollo Guidance Computer built by MIT and Raytheon. And it worked. “Throttle down. Better than the simulator,” remarked Buzz Aldrin just minutes before Apollo 11 landed on the lunar surface (National Aeronautics and Space Administration 1969, 313).This focus on safety and mission success that created NASA’s simulators indirectly created another key movement in this history: the need to simulate the lunar landing for the global television audience.

Safety Signals

In 1961, when design work on Apollo’s communications infrastructure—the Unified S-Band or USB—commenced, engineers allocated only 500 kHz for TV broadcast, a fraction of that apportioned to safety-related signals like telemetry, or biomedical monitoring, and only a tenth of the scale of an earthbound TV signal. It would be several years before a “desperate” NASA administration realized that the decision profoundly compromised their ability to broadcast what had been conceived from the outset as a televised event.

This myopia had two consequences. First, while there is no inherent technical reason that a high-resolution color TV image could not be transmitted from the moon, the images viewed by millions on earth were grainy and ghostlike, produced on special, low-resolution, low-framerate cameras developed by the Radio Corporation of America (RCA) and Westinghouse; only the black-and-white RCA version was ready for Apollo 11. Second, CBS’ production team was so unconvinced by the visual quality of broadcast graphics provided by NASA’s press office, they decided it needed to be prepared to “go it alone” and prepare their own simulated visuals for the TV broadcast of Apollo—where “going it alone” ironically required commissioning NASA’s own technical suppliers of simulator content.^▶16

Which brings us back to HAL 10000 and the CBS lunar stage set. In the same month when Walter Cronkite’s evening news had recently expanded from fifteen to thirty minutes, the thirty-one-hour Apollo 11 broadcast was an extended visual feast—with a healthy serving of simulation. This included views of actors in space suit costumes on a stage set at LEM manufacturer Grumman in Bethpage, Long Island, with mining slag standing in for the lunar surface. Broadcast views also combined transparencies and mechanisms sourced from NASA contractors with new techniques in noise reduction and chromakey (“greenscreen”) technology for live broadcast. A prime example was the view broadcast as the Eagle capsule descended to the moon’s surface, which combined a transparency of the rising earth, another of a hovering lunar lander, and a scrolling, latex-belt lunar surface—all under the prominent caption “CBS NEWS SIMULATION.”

The roadway and its adjoining buildings had been built two years previously. They formed part of an elaborate model of two square miles of Marin County at 1"=40'0" scale. While sheltered in an architecture school, the model’s precision and materials resembled a model train set, with the “emphasis on foreground detail” (Institute for Urban and Regional Development 1975). Tiny model cars populated the curbside; signs from supermarkets and gas stations were photo-etched in thin, colored plexiglass. And around the horizon, slopes of sawdust stood in for the sun-bleached grasses of the Marin headlands.

Road Shows

While more sophisticated, the model was a direct descendant of the experiments that Gyorgy Kepes and Kevin Lynch had undertaken at MIT in the 1950s, filtered through new possibilities in digitally driven control systems. The idea of reproducing urban motion in film—particularly new kinds of roads and highways—had existed since the earliest plans for Kepes and Lynch’s study of urban perception. While the simulation of perception had given way in The Image of the City to its plan-based representation, the project’s original seed of inspiration would flower on these model hillsides.

“With a better understanding of the motion picture as a tool of ‘visual recording,’ ” Georgy Kepes had written in 1955, “one can hope to tackle . . . new aspects of our present urban environment” (Kepes 1955a, 4). In 1958, Lynch’s student Donald Appleyard began work on a separate study of urban highways, focusing on multiple techniques to picture and analyze the cinematic experience of automotive travel through the city. The resulting book, A View from the Road (1964), featured both the use and an explanation of a new language of graphic notation for speed, view, and direction, as well as a flip-book-style filmstrip of Appleyard’s drawings along one page corner (Appleyard et al. 1964). By 1966, Appleyard was able to conduct tests with a true, small-scale film setup, including an Optex “modelscope,” a 16 mm Bolex film camera, and bicycle and model train tracks (see Young 1968).

Cable and Community

By 1967, Appleyard was a new faculty member at UC Berkeley, where he declared his goal “to develop a way of simulating extensive urban areas on small-scale models and recording journeys through them” (Appleyard 1967). For Appleyard, this goal was important for the same reason Luis de Florez had sought to simulate unbuilt airplanes in 1944—to allow new designs to be vetted and evaluated before the expense and investment of construction. As Appleyard explained in a letter to simulator pioneer Edwin Link’s younger sister, the result would be a “Communications Model of the Planning Process” in which “proponents,” “opponents,” and “evaluators” would be connected by “simulation media” through “questions, evaluations, decisions, perceptions and [again] evaluations” (Appleyard 1974, 1977, 47). In the context of Berkeley in the late 1960s and early 1970s, Appleyard believed such an approach was essential to remove undue political concerns from the planning process. In his vision of the future, the use of urban simulation would expand and combine with new technologies like two-way cable television to create a full “communications model of community participation” (Appleyard 1977, 47). As Appleyard imagined, we would come to regularly confront simulated versions of our daily reality—but in a controlled and calibrated manner, to envision and shape a collective process of urban design. While simulation in a military-industrial context was coming to represent a lack of agency, Appleyard in 1960s and 1970s Berkeley sought to demonstrate its possibilities for democratic cocreation.

To facilitate this vision, Appleyard proposed and received the 1973 National Science Foundation (NSF) grant “Environmental Dispositions and the Simulation of Environments.” Totaling $348,000 (or over $2.5 million today), the funds paid for the enormous Wurster Hall gantry—or, as it became officially known, the Berkeley Environmental Simulation Laboratory (BESL). As outlined by Appleyard and his coauthor, psychologist Kenneth H. Craik, the grant proposed not just the creation of an elaborate simulation apparatus but also its validation through psychological research. “Research participants,” it was proposed, would arrive in Marin County by car from Berkeley. After arrival in the project area, they would be subjected to either a real drive, or a simulated journey along the identical route. Afterward, “extensive debriefing sessions will examine how the travelers describe the environment they pass through, what they notice in it, what they recall of it, how they feel about it, how they evaluate elements in it, and how they mentally organize its subareas and features” (Appleyard et al. 1973, 2). Only when the precise difference between our experience of a real and simulated version of the same architectural context was “calibrated,” Appleyard proposed, could a scientific process of the consideration of urban design proposals by prospective publics be undertaken.

At the time of the NSF grant, Appleyard was not satisfied with any efforts at producing simulated driving footage. His early Boston experiments were “jerky” and “unrealistic”^▶18 A surplus driving simulator purchased from Yale and the Connecticut Department of Transportation produced only a single, unstable sequence in 1970 (Institute for Urban and Regional Development 1975). With the NSF grant, Appleyard was determined to produce his own, superior filming apparatus from scratch. With 2001 still the gold standard of visual model photography, Appleyard arranged a meeting with Douglas Trumbull in fall 1973.

Silent Signals

Trumbull was at a professional loose end. After obtaining wide admiration (if not, in his view, appropriate screen credit) for his contributions to 2001, Trumbull had produced visual effects for lower-end science fiction films: the Ringo Starr-starring alien sex comedy Candy and the (moderately) higher-brow microbiological caper The Andromeda Strain. On the strength of these outings, he was commissioned by Universal Pictures to direct his own feature, Silent Running (1972).

An outer-space eco-fable with Bruce Dern as an isolated astronaut, accompanied by original Joan Baez songs and robot mascots operated by bilateral amputees, Silent Running imagined Earth’s remaining supply of plants in isolation beyond the rings of Saturn. The film was a low-budget, technical tour de force. Trumbull and his team built and filmed a twenty-four-foot model of Dern’s starship, the Valley Forge, with an ingeniously updated version of the linear camera-control system from 2001. Crafted by Trumbull and a young visual effects supervisor, John Dykstra, the system combined front-and-rear camera projection with linear motor control. Instead of a combination of synchronized Selsyn motors, however, the Silent Running rig combined newer, incremental stepper motors with custom transistor-based controls, enabling an additional axis of movement at the same stately pace of interplanetary motion that had defined 2001. For all its technical wizardry, however, Silent Running was a commercial flop. As a result, when he spoke to Appleyard in 1973, Trumbull was without work for himself or his assistants. Reluctant to leave Los Angeles, Trumbull recommended the young John Dykstra to Appleyard. Dykstra would spend the next year in Berkeley devising the control system for Appleyard’s gantry-mounted camera.^▶19

Maneuvering Minicomputers

At Berkeley, Dykstra looked for help. He found it in the staff of an aging Federal Aviation Administration landing simulator located seven miles north at the UC’s Richmond Field Station. Constructed in 1961 using synchronized motor systems and a full-size cockpit on overhead tracks, the simulator building sported an enormous, slanted profile, covering a 1:10-scale model runway. As a war-surplus cockpit slid down tracks hung from the building’s roof, the space could be filled with fog to simulate low-visibility landings and test a variety of visual safety systems. The system’s “Principal Electronic Technician,” Karl Mellander, had consulted already with Appleyard on the NSF proposal; Dykstra would collaborate even more with two of the simulator’s staff, Jerry Jeffress and Alvah Miller.

While an analog, Selsyn-controlled system was extensively studied for BESL’s gantry, the system Dykstra, Jeffress, and Miller developed would draw instead from the digital legacy of SAGE.^▶20 Beginning in 1955, members of Forrester’s team at MIT’s Lincoln Laboratory had experimented with a series of SAGE’s components—a combination of magnetic-core memory, the scale and modular construction of SAGE’s small console machines, their interactive displays, and newly available transistor electronics. The (relatively) small and capable machine they built, the Lincoln Labs TX-0 and its successor TX-1, inspired Forrester deputy Ken Olson and his lieutenant Harlan Anderson to form the Digital Equipment Corporation (DEC) in 1957. From 1960 into the 1970s, DEC marketed a series of increasingly sophisticated so-called “minicomputers,” the PDP series. (The term “minicomputer” was a deliberate play on the contemporary “miniskirt”: small, modern, and self-consciously revolutionary; see Cerruzi 2003.)

If IBM’s mainframe dominance was grounded in its experience with the massive AN/FSQ-7, then the DEC PDP sustained the interactive modularity of SAGE’s consoles (Cerruzi 2003, 53, 140). Where IBM dispatched its own technicians to perform maintenance and upkeep, DEC instead delivered reams of newsprint-printed manuals, instructions, and suggestions for modifications. Appleyard’s own reflections on the purchase of a computer for the project reveal the two firms’ contrasting approaches: “IBM was not interested in our ‘small requirements,’ our funding being too modest from their point of view. DEC on the other hand continually contacted us and helped with design questions” (Institute for Urban and Regional Development 1975, 36).

Driving Targets

When the BESL system was completed in 1974, its operator would first use a downward-facing TV to record a two-dimensional “rail” along one of the model’s roads, centering on roadways using a “bombsight” graphic that quietly recalled the system’s military genealogy. This “rail” would be recorded to a disk drive as a series of 16-bit x-y-coordinates (a sensor on the bottom of the model scope would ultimately control the z-coordinate, keeping the eye, as it were, on the road). Using a second series of controls, the speed and orientation of the camera would be added, including its vertical position, rotation, and angle of the view, all centered on the “pupil”—the technical term for the very center—of a tiny lens sitting just one-tenth of an inch above the model’s surface. This second set of controls was crucial to the idea of vehicle-based simulation, recording not just the pathway of the vehicle but also the view cone of the driver, swiveling to anticipate turns and new trajectories.

Sound and Light(sabers)

A last control-system cameo is worth recording. Just as the surfaces of Star Wars’ models were filled with found objects and kit-model parts taken out of context, so too were the soundscapes of the revolutionary film. Designed by Walter Murch from a series of found recordings by Ben Burtt, the soundtrack of Star Wars incorporated a legion of found and repurposed recordings: from leaky air conditioners to the hum of high-tension electricity wires, they created a dense, consistent environment as part of a place-making process Murch called “Worldizing.”^▶23

Two of the essential, unidentifiable sounds of the film were made from recordings of control systems. The sound of the Millenium Falcon was produced from a recording of motion-control systems of the same Dykstraflex camera that enabled its simulated movement. And, befitting the origins of simulation itself, the sound of lightsabers came from a pair of antiquated self-synchronizing motors in a film projection room at the University of Southern California (Rinzler 2010, 66).

Deeds and Boxes

On August 28, 1940, before the entry of the United States into World War II but after the United Kingdom had begun to fight the Battle of Britain, a small group of scientists joined more than a thousand sailors on the Duchess of Richmond from Liverpool to Halifax, Nova Scotia. With the scientists was a closely guarded “black, metal deed box” containing cavity magnetron E1189, serial number 12, which had been developed by the British General Electric Corporation (GEC). While radar technology was then being developed in the US, Germany, UK, and Japan, the E1189 microwave tube was the first to produce a signal powerful enough, and at a high enough frequency, to make detailed, mobile microwave radar possible.

Through a complex process of technological diplomacy, the black box was taken to Washington from Halifax, and thence to MIT, where the device it contained formed the basis of developments in the newly chartered Rad Lab (Von Hilgers 2011). There, whether via this specific container or the similar, black speckled equipment containers that enclosed much of the facility’s complex equipment, the phrase “black box” became a term of general use in the Rad Lab (Galison 1994, 246 46n). By 1945, Wiener was using the term in a theoretical context, contrasting technological “black boxes” (whose inputs and outputs were understood but whose inner workings were not open to view) with a “white box” whose interior could be easily revealed (Galison 1994, 245).

In the systems-engineering context of SAGE and the ICBM defenses that superseded it, the “black box” became, once more, a practical term. In a complex assembly of controls and subsystems (often literally enclosed in dark metal), a “black box” was a portion of the whole (like a guidance system) defined not by its workings but by its inputs and outputs; how it arrived at these could be changed and upgraded as long as the relationships between the device and the larger supersystem (of, say, a missile) were maintained (see de Monchaux 2011, 47). Thus, the “black box” described not so much Wiener’s notion of unknowability, as it did the new definition of any technology as a series of defined relationships between mutable things—in contrast to the tangible objects that had mostly defined technological reality to date. Yet, like a black, metal deed box, a mechanism can become not only difficult to physically enter; it can become impossible to understand.

The Loss of Finding Our Way

The way-finding problem that The Image of the City analyzed—the use of landmarks, edges, signs, and pathways to negotiate complex urban reality—was viewed anew through a cybernetic lens, but also as (seemingly) immutable, like the idea of streets themselves. Decades later, however, a visit to the Boston Common finds no one at all asking for directions or navigating by the spire of a church; instead, they hold small boxes with screens, on which an algorithm predicts an optimized route. We are no longer reading signals from the urban landscape but in the urban landscape, preoccupied with signals on the surface of a box we cannot enter. And divorced from our environment even so.

In his 1968 review of 2001: A Space Odyssey, American Cinematographer editor Herb Lightman found only one element of the technological development depicted in the film difficult to believe—the erratic artificial intelligence of HAL-9000 (Lightman 1968). Fifty years on, however, a reality-challenged AI reads instead as the most credible part of 2001’s vision. Yet, unlike the illuminated, weightless interior of the film’s second-act finale, where HAL’s Perspex-slotted brain is deconstructed by Dave Bowman, the interior of today’s responsive algorithms are, as a rule, closed to view—even as they model key components of human cognition.

This presents the key irony of this tale. Our current interaction with civic and digital systems is defined by simulation—of attention, of traffic patterns, of the filling, defining, and interpolation that defines the current mathematical approach of machine intelligence. We are surrounded by these black boxes, but we cannot enter them; simulation has become like a locked-room murder mystery, with us, as detectives, on the outside.

Pneumatic Knowledge and the Paradox of Simulation

When Edward Link modeled the behavior of aircraft with pipe-organ parts in the 1920s, he continued a centuries-long tradition of airborne simulation. Through the eighteenth and nineteenth century, a series of “androids,” or humanlike automata, were presented to the scientific authorities as key insights into life itself. From the first such devices presented to the French Academy by Jaques Vaucanson in 1738—including a working flute player with glove-leather lips, lungs, and tongue—to a set of elaborate speaking heads produced by the Abbé Mical in 1783, through which, it was proposed, the pronunciation of French could be permanently fixed, these clockwork-pneumatic simulations sought to elucidate the most delicate parts of our existence: breath and song.

The most celebrated of these devices manipulated baser substances: the duck crafted by Vaucanson which, it seemed, ingested corn and produced a substance similar to feces—an achievement for which he was hailed by Voltaire as “Prometheus’ rival” (Riskin 2016, chap. 4; parts of this history are also considered in de Monchaux 2011, chap. 2). Like his Hungarian counterpart Wolfgang Von Kempelen’s chess-playing Turk of 1770, with its hidden operator, Vaucanson’s duck was an illusion—it did not alter its food into feces but simply squeezed brown paste from a hidden compartment. But Kempelen’s subsequent speaking machine, developed over twenty years, was again a pneumatic tour de force, accurately reproducing spoken language and providing key insights into the nature of speech. The notion that a model gave special insight into nature was a key belief in a world where—until the early nineteenth century—man himself was given by Diderot’s Encyclopédie as an example of a “machine,” instead of its opposite (Riskin 2016, chap. 4).

This history is relevant, because it outlines the opposite of our larger tracing. What we have followed above, since the 1930s, is the transformation of the model from an analog of, or insight into, its subject. It is now an active agent in shaping the reality it charts. In this arc, the line between simulation and real experience—from SAGE to soundstage—begins not just to resonate but, ultimately, to collapse. Yet, as can be traced in the movement of such models into smaller and more closed interiors—like the algorithms in an iPhone, ultimately closed to view—the more ubiquitous and successful such a model, the more likely it is to contravene, precisely, a model’s original purpose: to create, through its analog, a more complete understanding of the whole. We are caught in a paradox: the more powerful a model in shaping our reality, the more closed to view it, and its movements, seems to become.