In 1948 the British psychiatrist William Ross Ashby (1903 to 1972) completed a machine he called the Homeostat, built from four interconnected units he assembled out of war-surplus Royal Air Force bomb-control gear. Each unit had a pivoted magnet sitting in a trough of water that acted as a variable resistor, and the four were wired so that the output of each fed into the others. When Ashby disturbed the system - by hand, or by reversing a connection - the magnets swung, but the machine then adjusted its own internal settings until all four returned to a central, balanced position.
Ashby’s point was that the Homeostat modeled a basic feature of living things: the ability to keep essential variables within survivable limits despite disturbances. He called this property “ultrastability,” and the machine could reach a new stable state even after changes its designer had not anticipated, by stepping through different internal configurations until one held. He laid out the theory in his 1948 and 1952 book Design for a Brain, and his journals are preserved in the W. Ross Ashby Digital Archive. The Homeostat was demonstrated at the Macy cybernetics conferences and described in the press as an “artificial brain.”
The Homeostat is a landmark of the cybernetics era, alongside Wiener’s feedback theory and Grey Walter’s tortoises. It framed adaptation and learning as the behavior of feedback systems seeking equilibrium - a view of intelligence as self-regulation that contrasts with the symbol-processing approach AI would soon take, and that still echoes in control-theoretic accounts of learning.