In 1948, the same year that Claude Shannon published his theory of information, an MIT mathematician named Norbert Wiener published a book called Cybernetics: or Control and Communication in the Animal and the Machine. The title was new (Wiener coined “cybernetics” from the Greek kybernetes, meaning helmsman or governor), and the idea behind it was ambitious: to create a unified science of feedback, control, and communication that applied equally to machines, living organisms, and social systems.
Cybernetics became one of the most influential intellectual movements of the twentieth century. It shaped the development of computer science, artificial intelligence, robotics, neuroscience, systems theory, and management science. It also produced something unusual for a mathematical theory: a set of ethical warnings. Wiener was among the first scientists to foresee that intelligent machines would transform society, and he spent the last years of his life warning that the transformation might not be benign.
The Child Prodigy
Norbert Wiener was born in Columbia, Missouri, in 1894, the son of Leo Wiener, a professor of Slavic languages at Harvard. Leo Wiener was a demanding father who educated his son at home with an intensity that bordered on cruelty. Norbert entered Tufts College at eleven, graduated at fourteen, and received his PhD from Harvard at eighteen. He was, by any standard, a child prodigy, and his early years left psychological scars that he carried for the rest of his life.
After postdoctoral work in Europe (studying with Bertrand Russell at Cambridge and David Hilbert at Göttingen), Wiener joined the MIT mathematics department in 1919. He spent his entire career there, becoming one of the most original mathematicians of his generation. His contributions to pure mathematics included fundamental work on Brownian motion, harmonic analysis, and stochastic processes. But it was his wartime work on anti-aircraft fire control that led him to cybernetics.
The Anti-Aircraft Problem
During World War II, Wiener worked on the problem of automatically aiming anti-aircraft guns at fast-moving enemy aircraft. The challenge was to predict where the aircraft would be by the time the shell arrived (a process that took several seconds, during which the aircraft could move a significant distance). The prediction had to account for the aircraft’s current position, speed, and acceleration, and it had to be updated continuously as new radar data arrived.
Wiener recognized that this was fundamentally a problem of feedback. The gun’s aim had to be continuously adjusted based on the difference between the predicted and actual positions of the target. If the gun was aimed too far to the left, the error signal (the difference between aim and target) would cause the mechanism to correct to the right, and vice versa. The system corrected itself by feeding the output (the gun’s current aim) back into the input (the control mechanism) and using the error to drive corrections.
This insight, that self-correcting behavior can be achieved through feedback loops, was not new. Engineers had been building feedback mechanisms since James Watt’s centrifugal governor for steam engines in the 1780s. Physiologists knew that the body uses feedback to regulate temperature, blood pressure, and posture. What Wiener saw was that feedback was a universal principle, common to machines and organisms, and that it could be analyzed using a single mathematical framework.
The Cybernetic Vision
Wiener’s Cybernetics (1948) presented this framework. The book drew connections between engineering, biology, psychology, and mathematics that had never been made so explicitly before:
- Feedback in machines: A thermostat measures the temperature, compares it with the desired temperature, and turns the heating on or off to reduce the error. This is negative feedback: the output (room temperature) is fed back to the input (the thermostat’s controller) to maintain stability.
- Feedback in organisms: When you reach for a cup, your visual system monitors the position of your hand, compares it with the cup’s position, and sends corrective signals to your muscles. This is the same feedback loop that governs a thermostat, implemented in neurons and muscle fibers rather than wires and switches.
- Communication in both: Machines process information through electrical signals. Organisms process information through nerve impulses. In both cases, the fundamental quantity is information (in Shannon’s sense), and the fundamental process is the transmission, processing, and feedback of information.
Wiener argued that the similarities between machines and organisms are not superficial analogies but deep structural parallels. Both are information-processing systems that use feedback to maintain stability, adapt to changing conditions, and pursue goals. The mathematics of feedback, control, and communication applies to both. A unified science of these phenomena, cybernetics, could illuminate both engineering and biology.
Influence on Artificial Intelligence
Cybernetics had a direct and profound influence on the development of artificial intelligence. Several of the founders of AI, including John McCarthy, Marvin Minsky, and Warren McCulloch, were deeply influenced by cybernetic ideas. McCulloch and Walter Pitts published their model of neural networks (artificial neurons that compute logical functions) in 1943, directly inspired by the cybernetic framework of comparing brains to computing machines.
The cybernetic idea that intelligence is a property of information-processing systems, not a uniquely biological phenomenon, was the conceptual foundation that made AI thinkable. If the brain is a feedback system that processes information, and if machines can process information and implement feedback, then there is no reason in principle why a machine cannot exhibit intelligent behavior. This argument, implicit in Wiener’s work and explicit in the work of his followers, opened the door to the entire field of AI.
However, the relationship between cybernetics and AI was not always harmonious. By the mid-1950s, the AI community had largely split from cybernetics. The AI researchers (led by McCarthy and Minsky) focused on symbolic reasoning and logical computation. The cyberneticians (led by Wiener and the neural network researchers) focused on adaptive systems, learning, and feedback. The two approaches competed for decades, and it was not until the recent triumph of deep learning (which is essentially a descendant of the cybernetic tradition of neural networks) that the cybernetic approach was vindicated. The full arc from Colossus to ChatGPT owes more to cybernetics than is commonly recognized.
The Human Use of Human Beings
Wiener’s most prescient contribution may not have been technical at all. In 1950, the same year Alan Turing published his landmark paper on machine intelligence, he published The Human Use of Human Beings: Cybernetics and Society, a book aimed at a general audience that explored the social and ethical implications of automation and intelligent machines.
Wiener warned that machines capable of performing tasks previously reserved for humans would displace workers on a massive scale. He compared the coming automation revolution to the Industrial Revolution and argued that unless society planned carefully, the benefits would accrue to machine owners while the costs would fall on workers. “Let us remember that the automatic machine is the precise economic equivalent of slave labor,” he wrote. “Any labor which competes with slave labor must accept the economic conditions of slave labor.”
He also warned about the dangers of delegating decisions to machines. If a machine is programmed to pursue a goal, and if the goal is specified imprecisely, the machine may pursue the goal in ways that are harmful or absurd. A machine told to maximize production might work humans to exhaustion. A machine told to win a war might use methods that no human commander would consider acceptable. The danger is not that machines are malicious but that they are literal: they do exactly what they are told to do, and what they are told to do may not be what was meant.
These warnings, written in 1950, anticipated by seventy years the current debates about AI alignment, algorithmic bias, automation and employment, and the risks of artificial general intelligence. Wiener saw the problem more clearly than almost any of his contemporaries, and his warnings were largely ignored.
A Forgotten Prophet
Norbert Wiener died of a heart attack in Stockholm on March 18, 1964, at the age of sixty-nine. By then, cybernetics as a distinct discipline was already fading, absorbed into the more specialized fields it had spawned: control theory, information theory, systems engineering, neural network research, and artificial intelligence. The word “cybernetics” fell out of fashion in the English-speaking world (though it remained influential in the Soviet Union, where it was applied to economic planning and management theory).
Today, Wiener is less famous than Shannon, less celebrated than Turing, and far less known than the AI researchers who built on his ideas. The rivalry between von Neumann and Turing tends to overshadow Wiener’s equally foundational contribution. This is partly because cybernetics was a broad, interdisciplinary vision rather than a specific technical achievement, and broad visions are harder to credit to a single person. It is also because Wiener’s ethical warnings were uncomfortable. Scientists prefer to celebrate possibilities, not to dwell on dangers.
But Wiener’s central insights remain as relevant as ever. The idea that intelligent behavior emerges from feedback and information processing is the foundation of modern AI. The warning that machines pursuing poorly specified goals can cause harm is the defining concern of AI safety research. And the observation that automation creates wealth for some while displacing others is the central social question of the twenty-first century. Norbert Wiener saw all of this in 1948. The world is still catching up.
No products in the cart. 