Also known as: feedback systems
Cybernetics is the interdisciplinary study of systems that regulate themselves through feedback, using their own outputs to shape their inputs in pursuit of a goal. The term was coined by Norbert Wiener in his 1948 book Cybernetics: or Control and Communication in the Animal and the Machine, from the Greek κυβερνήτης (kybernētēs, "steersman", the same root as "governor").
The cybernetic frame unified streams of work that had previously been disconnected: Wiener's wartime research on automatic gun-laying systems, Walter Cannon's physiological notion of homeostasis, McCulloch and Pitts's logical model of neural networks, Shannon's information theory, and von Neumann's game theory and theory of automata. The Macy Conferences on Cybernetics (1946–1953) were the central forum.
Cybernetics directly anticipated reinforcement learning (a feedback agent maximising reward), control theory (a feedback regulator stabilising a plant), robotics (Wiener's prosthetic-limb proposals predated the field), and much of cognitive science (the brain as an information-processing feedback system). The second-order cybernetics of Heinz von Foerster and others extended the analysis to systems that observe themselves, foreshadowing modern reflective AI architectures.
As a named discipline, cybernetics largely faded by the 1970s, its concerns absorbed into AI, control theory, systems biology and complexity science, but the conceptual framework remains influential and is enjoying a revival in modern AI safety discussions of feedback loops between learners and their training data.
Related terms: Feedback Loop, Reinforcement Learning, Control Theory
Discussed in:
- Chapter 1: What Is AI?, A Brief History of AI