Each iteration of the pattern follows the same structural sequence. A mechanism appears. The mechanism impresses — it captures something genuine about reality, explains phenomena that were previously puzzling, produces outputs that seem almost magical relative to what preceded it. A leap is then made: from 'explains some things' to 'explains everything.' The leap is motivated not by evidence but by the aesthetic pleasure of a unified theory — the universe as clock is more elegant than the universe as a messy assortment of clockwork and biology and weather and consciousness.
The clock metaphor was not foolish. It captured real regularities in planetary motion. But it concealed what it could not describe: clocks do not evolve, do not produce novelty, do not develop consciousness. The clockwork universe was a map of the predictable features drawn by people who had temporarily forgotten that reality also contains unpredictable features, and that the unpredictable features are at least as important. The engine metaphor was not foolish either — energy does flow from concentration to uniformity, entropy does increase. But the metaphor concealed the most interesting phenomenon in the universe: the emergence of complex self-organising systems that locally reverse the trend.
The large language model is the most dangerous all-explaining mechanism in the series precisely because each previous one was limited by the obviousness of its dissimilarity to the thing it was supposed to explain. Nobody seriously thought the universe was made of tiny gears. Nobody seriously thought the brain was filled with steam. But many people seriously think that a system producing human-quality language is, in some meaningful sense, thinking. The resemblance has crossed the threshold of plausibility, and once a metaphor crosses that threshold, it stops being treated as a metaphor and starts being treated as a description.
Midgley's point is not that any of these metaphors was wrong. The clock metaphor illuminated mechanical regularity. The engine metaphor illuminated thermodynamic flow. The computer metaphor illuminates information processing. The language model metaphor illuminates statistical structure in language. Each captured something real. The error, in every case, was the promotion — the move from 'captures something real' to 'captures reality itself.' And the error is not just intellectual. It has consequences. Cultures that organized themselves around the clockwork universe built specific institutions, specific metaphysics, specific relationships between science and religion. Cultures that organize themselves around the language model are building their own. The question is whether the institutions will survive the inevitable moment when the metaphor is recognized as a metaphor rather than a description.
The framework appears across Midgley's corpus but receives its most explicit treatment in The Myths We Live By (2003), Chapter 4 ('The Machine Image'), and in Science as Salvation (1992), where she traces the specific pattern of the AI research community's tendency to inflate technical achievements into metaphysical claims.
The pattern repeats. Clock, engine, computer, language model — the same structural error recurs with different mechanisms, each more seductive than the last.
The metaphor captures something. The error is not in the metaphor but in its inflation — the move from useful illumination to total description.
Resemblance crosses the plausibility threshold. Language models are more dangerous than prior mechanisms because their outputs are closer to what they claim to explain.
Elegance is not evidence. Unified theories are aesthetically satisfying and epistemologically suspect; reality is not obliged to be simple enough for a single mechanism to explain.