What is cybernetics, and what did Norbert Wiener's 1948 book say?

Cybernetics is the science of how systems steer themselves. A thermostat that keeps a room warm, a ship's autopilot holding a course, and your hand reaching for a cup all work the same basic way: they compare what is happening with what should happen and adjust. The American mathematician Norbert Wiener gave this field its name and its first big book, Cybernetics, published in 1948. The full title spells out the program: "or Control and Communication in the Animal and the Machine."

Wiener's bold claim was that the same mathematics describes both living bodies and machines. Nerves and wires, muscles and motors, brains and calculating machines could all be studied as systems that send messages and react to them. That idea shaped engineering, biology and the early dreams of thinking machines.

Where the book came from

Wiener was a mathematics professor at the Massachusetts Institute of Technology (MIT). During the Second World War he worked with the engineer Julian Bigelow on a hard military problem: predicting where an enemy airplane would be a few seconds in the future, so that anti-aircraft guns could aim at that point. The plane was flown by a person, and the gun was often aimed by a person, so the machine had to take human behavior into account.

That work led Wiener and Bigelow to a central concept: feedback. When you want a motion to follow a plan, the difference between the plan and the actual motion is fed back as a new input that pulls the motion closer to the plan. They took a question to Wiener's longtime collaborator, the physiologist Arturo Rosenblueth: is there an illness where a person reaching for something overshoots and starts to swing back and forth? Rosenblueth said yes. It is called purpose tremor and is often linked to injury of the cerebellum, a part of the brain. To the three of them, that was strong evidence that the nervous system itself runs on feedback loops.

Through the 1940s these ideas spread at small interdisciplinary meetings, including a series sponsored by the Josiah Macy Foundation in New York, where Warren McCulloch and Walter Pitts, authors of the 1943 model of the neuron, also took part. Wiener says the group coined the word "cybernetics" in the summer of 1947, from the Greek word for steersman. During a trip to France that year, the Paris publisher Hermann et Cie asked him for a book. He signed its introduction in November 1947 at Rosenblueth's institute, the Instituto Nacional de Cardiología in Mexico City, and the book came out in 1948 in Paris and in Cambridge, Massachusetts.

How feedback works: a worked example

Wiener's own example is picking up a pencil. You do not consciously command each muscle. You simply want the pencil, and at every moment your eyes and your sense of your own body report how far your hand still is from it. That shrinking gap drives the motion. Here is the same loop in a thermostat:

This is negative feedback: the system acts to shrink the error. Wiener also explained what goes wrong when feedback is too strong or too slow. The system overshoots, corrects too hard, overshoots the other way and can swing wildly, as a badly tuned ship's steering engine does, or as a hand with purpose tremor does.

What else the book covers

Cybernetics is a mix of philosophy, heavy mathematics and bold speculation. It discusses time and prediction, information and noise in messages, and the nervous system. In a chapter on computing machines, Wiener argued that future calculating machines should be digital rather than analog, electronic rather than mechanical, and binary, and he compared them with the brain. He even discussed whether a machine could play chess, and whether one could learn from its own past games.

He also wrote frankly about the dangers. The new science, he warned, came with "great possibilities for good and for evil," including the automation of human labor. The book reached far beyond engineers and became a bestseller, and Wiener followed it in 1950 with The Human Use of Human Beings, written for general readers.

Why it mattered

Cybernetics gave scientists across many fields a shared vocabulary: feedback, control, information, signal and noise. It influenced control engineering, computing, neuroscience and communications. It also made "machines that behave purposefully" a respectable scientific subject a few years before the term "artificial intelligence" existed. When John McCarthy named AI in 1955, he picked a fresh label partly to keep his new field separate from cybernetics, with its focus on analog feedback, and from Wiener's strong personality.

What does cybernetics have to do with AI in 2019?

More than the name suggests. The word "cybernetics" is rarely used in AI labs today, but its central idea runs through modern systems. Training a neural network is a feedback loop: the network makes a prediction, the error is measured, and that error is fed back to adjust its internal settings so the next prediction is a little better. Reinforcement learning, the method behind DeepMind's AlphaGo, is learning from feedback about which actions lead to good results. Self-driving cars and drones rely on classic control loops that Wiener would recognize at once.

Wiener's warnings also sound current. In 2019 people are debating automation and jobs, the power of those who control intelligent systems, and the risk of machines pursuing goals that do not match what we really meant. Wiener raised each of these in the 1940s.

For a beginner, the most useful takeaway is his way of looking at any smart system: ask what it is trying to achieve, what it measures, and how it corrects itself. That simple loop explains a surprising amount of what today's AI does, and where it can go wrong.