Cybernetics

Cybernetics is the study of control and communication in animals, machines, and organizations. Its central idea is that purposeful behavior can be explained by feedback: a system acts, measures the effect of its action, and uses the difference between the result and the goal to correct what it does next. The field takes its name from Norbert Wiener's 1948 book Cybernetics: Or Control and Communication in the Animal and the Machine, and Wiener built the word from the ancient Greek kybernetes, meaning steersman ^[1][2]. For about a decade after the Second World War, cybernetics was the banner under which mathematicians, engineers, neurophysiologists, psychiatrists, and social scientists tried to construct a common science of brains and machines.

Its relationship to artificial intelligence is both ancestral and adversarial. The McCulloch-Pitts model of the neuron, early autonomous robots, and the perceptron all came out of the cybernetics milieu, and modern neural networks descend directly from that line of work ^[6][14]. Yet the researchers who founded AI as a named discipline in 1956 deliberately picked a new term to put distance between themselves and Wiener's program. John McCarthy later wrote that "one of the reasons for inventing the term 'artificial intelligence' was to escape association with 'cybernetics'" ^[12]. The two fields then diverged for decades: symbolic AI captured most of the funding and attention, while cybernetics turned toward biology, epistemology, and management before fading as an organized movement ^[2].

Wartime origins

The technical core of cybernetics came out of a Second World War fire-control problem. Wiener, a mathematics professor at MIT, worked for the US National Defense Research Committee on automatic aiming systems for anti-aircraft guns, which required predicting where an evasively maneuvering aircraft would be by the time a shell arrived. The statistical prediction and filtering methods he developed for this problem (the Wiener filter among them) later became standard tools in signal processing, and the experience convinced him that prediction, feedback, and communication were aspects of a single subject ^[5].

Two 1943 papers laid the conceptual foundations. In "Behavior, Purpose and Teleology," published in Philosophy of Science, Wiener, the engineer Julian Bigelow, and the Mexican physiologist Arturo Rosenblueth argued that goal-directed behavior in organisms and machines alike could be described in terms of negative feedback, giving the old philosophical notion of purpose an engineering definition ^[2][5]. The same year, the neuropsychiatrist Warren McCulloch and the logician Walter Pitts published "A Logical Calculus of the Ideas Immanent in Nervous Activity" in the Bulletin of Mathematical Biophysics, modeling neurons as threshold logic units and showing that networks of such units could compute any logical function. The McCulloch-Pitts neuron influenced John von Neumann's thinking about automata and computer design, and it remains the ancestor of the artificial neuron used in machine learning today ^[6].

Wiener's Cybernetics (1948)

Wiener coined the term cybernetics in the summer of 1947 and published Cybernetics: Or Control and Communication in the Animal and the Machine the following year, jointly through Hermann et Cie in Paris and MIT Press ^[1][2]. The book ranged across servomechanisms, computing machines, neurophysiology, and communication, and it treated information as a fundamental quantity in its own right; Wiener developed statistical ideas about communication in parallel with Claude Shannon's information theory ^[5]. The Greek root had a history Wiener acknowledged: Plato had used the steersman as a metaphor for governance, and in 1834 the French physicist Andre-Marie Ampere had used "cybernetique" for the science of government in his classification of knowledge ^[2].

Despite its mathematical density, the book became a surprise bestseller and made cybernetics a public topic far beyond engineering ^[1]. Wiener followed it in 1950 with The Human Use of Human Beings, a nontechnical book about the social consequences of automatic machinery, and he spent much of his remaining career warning about automation's effect on labor and about the militarization of science ^[1][5]. A second edition of Cybernetics appeared in 1961 with two added chapters ^[1].

The Macy conferences

The institutional home of early cybernetics was a series of ten small interdisciplinary meetings sponsored by the Josiah Macy Jr. Foundation, held in the United States from March 1946 to April 1953 and chaired throughout by Warren McCulloch ^[3][4]. The first meeting was titled "Feedback Mechanisms and Circular Causal Systems in Biological and Social Systems"; after Wiener's book made the new word famous, the group adopted it, and the series became known as Cybernetics: Circular Causal and Feedback Mechanisms in Biological and Social Systems. The renaming followed the arrival of the Austrian-born physicist Heinz von Foerster, who joined at the sixth conference in March 1949 and proposed using Wiener's term in place of the unwieldy original title ^[3].

The core group mixed disciplines to an unusual degree: Wiener, von Neumann, Pitts, Bigelow, and Rosenblueth from mathematics, engineering, and physiology sat alongside the anthropologists Margaret Mead and Gregory Bateson, the psychologist Kurt Lewin, the neurophysiologist Rafael Lorente de No, and others ^[3][4]. Claude Shannon took part as well, and two leading British cyberneticians appeared as guests near the end of the series: W. Ross Ashby presented at the ninth conference in 1952, and W. Grey Walter attended the tenth in 1953 ^[3]. Von Foerster edited the published transactions, with Mead and Hans-Lukas Teuber as assistant editors; only the sixth through tenth conferences were documented this way ^[3][4]. The conferences are also remembered as a seedbed for cognitive science, since they put the question of how brains process information at the center of several disciplines at once ^[6].

British cybernetics

Britain produced a parallel school with a more experimental, build-it-and-see temperament. Its social hub was the Ratio Club, an informal dining club founded in 1949 by the neurologist John Bates that met until 1958. Members included Alan Turing, who joined right after the inaugural meeting, along with Ashby, Grey Walter, Horace Barlow, and Donald MacKay; McCulloch spoke twice as a guest. The club drew psychiatrists, physiologists, mathematicians, and engineers, and it became the most influential cybernetics group in the UK ^[11].

Ashby's homeostat and the Law of Requisite Variety

W. Ross Ashby was a psychiatrist, director of research at Barnwood House Hospital in Gloucester from 1947 to 1959 and later a professor at the University of Illinois at Urbana-Champaign ^[7]. On 16 March 1948 he completed the homeostat, a machine assembled from four interconnected Royal Air Force bomb control units, each carrying a pivoted magnet. Disturb the magnets and the device would reconfigure its internal circuit parameters until the whole system settled back into equilibrium, a property Ashby called ultrastability ^[8]. He demonstrated it at the ninth Macy conference in 1952. Time magazine in 1949 called it "the closest thing to a synthetic brain so far designed by man," and Wiener described Ashby's work as one of the great philosophical contributions of the day ^[7][8].

Ashby's two books gave cybernetics much of its theoretical vocabulary. Design for a Brain (1952) worked out how a machine could adapt the way a nervous system does, and An Introduction to Cybernetics (1956) became the field's standard textbook. The latter introduced the Law of Requisite Variety, summarized in Ashby's phrase "only variety can destroy variety": a regulator can only hold a system steady if it commands at least as many distinct responses as the disturbances it must absorb ^[7]. The law is still cited in control engineering, management theory, and discussions of AI oversight.

Grey Walter's tortoises

W. Grey Walter was an American-born British neurophysiologist who worked at the Burden Neurological Institute in Bristol from 1939 to 1970. He was a pioneer of electroencephalography, the first to localize the strongest alpha rhythms to the occipital lobe, and in the 1960s the discoverer of the contingent negative variation, an electrical signal that appears in the brain before conscious awareness of an intended movement ^[9].

Between 1948 and 1949 he built two small three-wheeled robots, Elmer and Elsie, known as the tortoises; specimens in the Science Museum Group collection carry his mock-Linnaean species name M. speculatrix ^[9][10]. Each used a photocell, a touch contact, and a two-tube vacuum-tube amplifier driving steering and drive motors; Walter described them as having "only two sensory organs and two electronic nerve cells" ^[19]. From those few components came surprisingly lifelike behavior: the tortoises sought moderate light, backed away from obstacles, and found their way to a hutch with an internal light bulb to recharge ^[9][10][19]. In a famous demonstration, Walter mounted a pilot light on a tortoise and set it before a mirror; the machine, attracted and repelled by its own reflection, began flickering and jigging, behaving, he wrote, like a clumsy Narcissus. His point was deliberately economical: rich connections among a very small number of components can generate complex behavior ^[9]. Walter described the work in Scientific American articles in 1950 and 1951 and in his book The Living Brain (1953), and the tortoises are widely treated as ancestors of autonomous robotics ^[9][19].

Key works

Scroll sideways for more →

The split from symbolic AI

The 1956 Dartmouth workshop that launched artificial intelligence as a named field was, in part, a secession from cybernetics. McCarthy was explicit about it decades later: cybernetics' "concentration on analog feedback seemed misguided, and I wished to avoid having either to accept Norbert (not Robert) Wiener as a guru or having to argue with him" ^[12]. The new field bet on digital computers and symbol manipulation rather than analog feedback and neurophysiology, and as AI grew it pulled funding and students away from cybernetic research ^[2].

The neural-network strand of cybernetics survived the split for a while. Marvin Minsky had built SNARC, an early neural-network learning machine, in 1951 ^[6], and Frank Rosenblatt developed the perceptron at Cornell Aeronautical Laboratory, describing it in a 1957 report and a 1958 Psychological Review paper and building the Mark I Perceptron hardware in 1959 ^[14]. But Minsky and Seymour Papert's 1969 book Perceptrons, which proved that single-layer perceptrons cannot represent many classes of patterns, helped push neural-network research into a long stagnation while symbolic approaches and expert systems dominated ^[14]. The connectionist revival of the 1980s, and eventually deep learning, vindicated the McCulloch-Pitts lineage, though by then few of its practitioners called themselves cyberneticians. Histories of the AI winters usually treat the fate of the perceptron as a cautionary tale about how quickly a research program can lose its funding and its reputation together ^[14].

Second-order cybernetics

From the late 1960s the remaining cybernetics community took a reflexive turn. Margaret Mead, in a keynote to the American Society for Cybernetics delivered in 1967 and published in 1968, urged the field to apply its own ideas to itself, treating cybernetics "as a way of looking at things" rather than a toolbox ^[13]. Heinz von Foerster, whose Biological Computer Laboratory at the University of Illinois at Urbana-Champaign became the movement's base, drew the canonical distinction: first-order cybernetics is "the cybernetics of observed systems," while second-order cybernetics is "the cybernetics of observing systems," the study of how observers, including scientists, construct the systems they describe ^[13]. The 1974 volume Cybernetics of Cybernetics, produced from his course materials, gave the program its name.

Second-order cybernetics developed alongside Humberto Maturana and Francisco Varela's theory of autopoiesis, which describes living systems as self-producing, and Ernst von Glasersfeld's radical constructivism ^[13]. These ideas found audiences in biology, family therapy, sociology, and design theory, but they moved the field further from the engineering mainstream, and by the 1980s cybernetics had largely dissolved as a unified discipline even as its concepts spread ^[2][13].

Beyond the laboratory

Cybernetics traveled unusually far outside research institutes. In the Soviet Union it was attacked as a reactionary pseudoscience in the early 1950s, then rehabilitated by the mid-to-late 1950s and embraced on a large scale, with "cybernetics" becoming the Soviet umbrella term for a wide sweep of computing, control, and information research ^[2][18]. In Chile, the British operations researcher Stafford Beer, founder of management cybernetics, led Project Cybersyn for Salvador Allende's government between 1971 and 1973: a telex network linking state-run factories to Santiago, statistical software to flag production anomalies, and a purpose-built operations room for decision-making, all organized around Beer's Viable System Model. The system proved its worth during the October 1972 truckers' strike; the military coup of September 1973 destroyed the operations room and ended the project ^[15].

The word itself proved as durable as any theory. Manfred Clynes and Nathan Kline compressed "cybernetic organism" into cyborg in 1960, in work on adapting humans for spaceflight ^[16], and the "cyber-" prefix went on to name cyberspace, cybersecurity, and much else, long after most users forgot the steersman underneath.

Legacy

As an organized movement, cybernetics peaked early and faded; as a set of ideas, it never left. Control theory absorbed the mathematics of feedback and grew into a core engineering discipline. The McCulloch-Pitts neuron and the perceptron sit at the root of the neural networks that power modern machine learning ^[6][14]. Grey Walter's tortoises prefigured behavior-based and embodied approaches to AI and inspired later roboticists, including the BEAM robotics tradition of simple analog machines ^[9]. Ashby's requisite variety and Wiener's warnings about automation both circulate in current debates about regulating and aligning AI systems, and Wiener's 1950 argument that society must decide how to live with automatic machines reads as an early entry in the history of artificial intelligence policy discussion ^[1][5].

Institutionally, a smaller cybernetics community persists. The American Society for Cybernetics, founded in the 1960s, still documents and develops the field ^[3][13], and the IEEE Systems, Man, and Cybernetics Society continues to publish IEEE Transactions on Cybernetics among its journals ^[17].

See also

• History of artificial intelligence
• Control theory
• Feedback loop
• Neural network
• Information theory
• Symbolic AI

References

1. ^Wikipedia: Cybernetics: Or Control and Communication in the Animal and the Machine. en.wikipedia.org/..._in_the_Animal_and_the_Machine
2. ^Wikipedia: Cybernetics. en.wikipedia.org/...Cybernetics
3. ^American Society for Cybernetics: Summary of the Macy Conferences. asc-cybernetics.org/...MacySummary
4. ^Wikipedia: Macy conferences. en.wikipedia.org/...Macy_conferences
5. ^Wikipedia: Norbert Wiener. en.wikipedia.org/...Norbert_Wiener
6. ^Wikipedia: A Logical Calculus of the Ideas Immanent in Nervous Activity. en.wikipedia.org/...s_Immanent_in_Nervous_Activity
7. ^Wikipedia: W. Ross Ashby. en.wikipedia.org/...W._Ross_Ashby
8. ^Wikipedia: Homeostat. en.wikipedia.org/...Homeostat
9. ^Wikipedia: William Grey Walter. en.wikipedia.org/...William_Grey_Walter
10. ^Science Museum Group Collection: cybernetic tortoise objects (M. speculatrix). collection.sciencemuseumgroup.org.uk/search
11. ^Wikipedia: Ratio Club. en.wikipedia.org/...Ratio_Club
12. ^John McCarthy, review of B. P. Bloomfield (ed.), The Question of Artificial Intelligence, Stanford University. www-formal.stanford.edu/...bloomfield
13. ^Wikipedia: Second-order cybernetics. en.wikipedia.org/...Second-order_cybernetics
14. ^Wikipedia: Perceptron. en.wikipedia.org/...Perceptron
15. ^Wikipedia: Project Cybersyn. en.wikipedia.org/...Project_Cybersyn
16. ^Wikipedia: Cyborg. en.wikipedia.org/...Cyborg
17. ^IEEE Systems, Man, and Cybernetics Society: Publications. ieeesmc.org/publications
18. ^Wikipedia: Cybernetics in the Soviet Union. en.wikipedia.org/...Cybernetics_in_the_Soviet_Union
19. ^W. Grey Walter, "An Imitation of Life," Scientific American, Vol. 182 No. 5, May 1950. scientificamerican.com/...an-imitation-of-life