Control has always been at the center of the human relationship to machines.
The steam engine existed as a toy in Alexandrian-era Egypt. Harnessing steam for work proved difficult and dangerous until 1788, when James Watt added a âcentrifugal governor,â applying the principle of âfeedbackâ to the steam engine, allowing the engine to monitor and regulate its own output. Although the modern steam engine had been invented by Thomas Newcomen three quarters of a century before, Wattâs addition of self-regulation saw him credited as the âfatherâ of the steam engine.
Wattâs steam engine produced a new tension: As it became possible to build machines vastly more powerful than human beings, it became ever more important to rein in their power. The catalog of the ills of the Industrial Revolution feature many entries that demonstrate the destructive power of machines: Their capacity to rip limbs from bodies, overwhelm their operators, or catastrophically fail to self-regulate â then explode.
In the mid-nineteenth century, James Clerk Maxwell â who gave the world the eponymous equations unifying the electromagnetic forces â wrote an influential paper, On Governors, framing the centrifugal governor as a key example of a control system, a system capable of self-regulation.
With the growing capacities of steam-driven machinery â embodied in the locomotive â the idea of control became synonymous with the idea of safety; with the right controls, delivered both automatically and through human oversight (in this case, a locomotive engineer), inhuman forces could be deployed for human benefit.
The word governor has roots that stretch back into Latin â gubernator â and further back into Greek â ÎșÏ
ÎČΔÏÎœÎźÏÎ·Ï â âkubernetes,â derived from the word for âsteersman,â âpilot,â or âguide,â that is, the person who responds to the boat, the current, and the wind, integrating all of that into subtle shifts and changes in direction. In the 1940s, that governing metaphor of self-control â a pilot guiding a boat through dangerous waters â would find its modern form, in weapons.
A weapon focuses power toward a destructive end, but must not destroy whoever deploys that force, and so exists in that uneasy tension between power and control. Killing power must be balanced rationally, and by design.
In the immediate post-war period, the immense technical advances made by both sides during the conflict slowed and deepened. Although the Messerschmitt jet fighter had been used to great effect by the Luftwaffe, only after the war did aeronautical engineers have the luxury to experiment, expand, and improve on those early efforts. Rapid innovation led to a jet that flew so fast it could approach the speed of sound, with fatal consequences as it encountered its shock wave; or could bank into a turn so rapidly that it slammed the pilot with intense gravitational forces, driving them to unconsciousness as blood pulled away from the brain. The jet fighter had capacities embodied in its design that could prove fatal.
In order to minimize the opportunities for disaster, these early jet fighters crowded their cockpits with an exhaustive array of instrumentation needed for flight, navigation, and combat operations. This stretched the pilot in another direction â toward a condition of âcognitive overload.â So much important data, streaming into so many displays, can easily avalanche into confusion. Many pilots tried to master these early jet fighters; only a few had the necessary capacity to safely integrate all of the information they provided, achieving âsymbiosisâ with the machinery, where aircraft and pilot behaved as a single unit.
That symbiosis of man and machine emerged as one of the central themes of academic discourse in the early post-war era. The âMacy Conferences,â gathering the best and brightest from disciplines as diverse as mathematics, neurology, and anthropology,1 worked toward symbiosis using concepts garnered from both âhardâ and âsoftâ sciences: The 300-year-old Cartesian split between res cogitans, our inner life, and res extensa, the world of things. Symbiosis, for the Macy conferees, pointed toward a new, active relationship between the inner and outer worlds, one that could be embodied in new kinds of machines.
Each participant, eminent in their own discipline, came with an open mind â and a set of proposals. None came more prepared than Norbert Wiener, a true polymath whose talents extended to mathematics, electronics, systems theory, computing â and a new discipline, which heâd both invented and named after that Greek word for âsteersmanâ: cybernetics.
Wiener aimed for nothing less than a complete description of the theory of control in any system â living or mechanical. Laid out in elegant though nearly inaccessible mathematical detail in his 1948 volume Cybernetics,2 Wiener introduced several concepts that have, in the years since, become so commonplace they seem almost obvious. First among these, feedback: The idea that the output of a process can be used to help control that process. For Wiener, obsessed with the theory of information, this meant every process generates information that can be used to control it.
Two years later, in The Human Use of Human Beings, Wiener used far more accessible language to explain his cybernetic hypotheses, writing, âInformation is the name for the content exchanged with the outer world as we adjust to it, and make our adjustment felt upon it.â3 In this economy of language Wiener unites res cogitans with res extensa via the exchange of information, describing a system where messages from the world change the internal landscape, while interiority sends messages that change the world, each process informing the other in a dance of feedback.
During the war, Wiener embodied his theories of control and feedback in the design of electronic circuits controlling anti-aircraft guns, allowing them to automatically track their targets. Five months after VJ Day, ENIAC, the first true digital computer, gave mathematicians like Wiener a tool that could âprocessâ information along lines determined by the logic of a âprogram.â Yet, for all its brute force as a numerical calculator, the computer had no significant connections to the outside world, beyond a few switches and relays. All mind and no body, these first computers had res cogitans without res extensa. They could think, but they could not do, nor could they sense the way their thoughts shaped the world, and respond.
To bridge that divide, the computer would have to become a cyber-physical system, a combination of mind plus senses paired with an ability to affect change in the world, then sense those changes. Taking in data continuously generated by a range of sensors, integrating it into an assessment of state, the cyber-physical device could compute decisions that would then be translated into actions. These would in turn generate new sensor data, requiring further integration and assessment, producing still further decisions and actions, generating still more data â on and on and on in a continuous flow, an exchange of content between the outer and inner worlds that seemed to perfectly embody Wienerâs conception of cybernetics.
At its essence, a cyber-physical device must be responsive in real time; sensing, deciding, and acting without delay. This real-time operation had not been considered in the design of early digital computers. In 1951, WHIRLWIND, the first computer designed for cyber-physical operations,4 proved that a computer program could respond in real time to a range of sensor inputs, and, although far too large, heavy, and power hungry to find its way into a jet fighter, WHIRLWIND acted as a key âtool for thinking,â a canvas upon which a new generation of computing pioneers could inscribe their own ideas for a symbiosis between a cyber-physical system and a human operator.
Research into âmanâcomputer symbiosisâ took a deadly serious turn as the Cold War commenced and development of megaton thermonuclear weapons created the specter of imminent planetary apocalypse. Military commanders needed to operate effectively and accurately within an increasingly narrow window between the launch of a surprise nuclear attack and its culmination. Real-time decision support systems could assist with command and control in these emergency circumstances, demanding the integration and display of vast amounts of often conflicting data, without producing additional cognitive overload among already-stressed decision makers.
The design of such decision support systems owed as much to psychology as to physics or mathematics. At MIT, a post-war center for both computing and sensory psychology, psychologist J. C. R. Licklider consulted on machine-to-human interfaces for SAGE, the Semi-Automatic Ground Environment, a vast continent-wide system designed to provide military decision makers data needed to make decisions about nuclear threats and attacks. SAGEâs radar sensors scanned the airspace above and beyond the United States, looking for inbound threats, eventually spanning some 78 early warning âDEW Lineâ radar stations, networked into a nationwide complex of supercomputers,5 each staffed with an entire floor of cutting-edge SAGE graphic terminals, where military operators could see, classify, and potentially elevate any perceived threats to their commanders.
Licklider consulted on the design of the SAGE graphic terminals, developing some of the first efforts toward a symbiosis of man and computer, then shared a vision for the possibilities of symbiosis in a brief but hugely influential 1960 paper, âManâComputer Symbiosisâ: