Sleeping Beauties

In the AI Story

In biological systems, dormant innovations are well documented. Genes encoding functional proteins may persist in a genome for millions of years without contributing to the organism's phenotype, silenced by regulatory mechanisms or rendered irrelevant by environmental conditions. When the environment shifts — a new food source, a disappearing competitor, a climate boundary — the dormant gene is activated and its innovation becomes the basis for a new adaptive radiation. The gene was functional throughout its dormancy; the organism carried the capability without expressing it.

The cultural parallel is exact in its structural logic. Ignaz Semmelweis demonstrated in 1847 that hand-washing dramatically reduced maternal mortality. The medical establishment rejected his findings; he died in an asylum in 1865. Two decades later Louis Pasteur's germ theory provided the mechanistic explanation that made the empirical observation scientifically legible, and hand-washing became standard practice. The innovation existed for twenty years before its environment was ready to receive it. The idea was sound. The evidence was available. None of this mattered because the conceptual landscape lacked the adjacent knowledge structures that would have made acceptance topologically accessible.

As Wagner observed, 'the awakening depends on the environment, and is beyond the innovator's control. It also cannot be predicted, and often calls the impact of the quality of an innovation into question, because this quality may often matter less than the environment.' The quality of the perceptron did not improve between 1969 and 1986. What changed was the landscape into which it was received — the adjacent knowledge structures, the available resources, the institutional readiness to support activation.

The framework introduces a specific form of uncertainty into assessments of AI capability. Current large language models have vast parameter spaces whose topology guarantees the existence of dormant capabilities — adjacent configurations that have not yet been activated because the appropriate contexts, prompts, or fine-tuning procedures have not been encountered. The relevant question is not only what models can do now but what they could do under conditions that have not yet been explored. The sleeping beauties are there. The question is whether the civilization receiving them will be robust enough to absorb the surprise.

Origin

Wagner developed the framework across a decade of research into dormancy phenomena in biological and cultural domains, culminating in the 2023 book Sleeping Beauties: The Mystery of Dormant Innovations in Nature and Culture. The bibliometric concept of 'sleeping beauties' in scientific literature — papers that receive little attention for years before becoming highly cited — was introduced by Anthony van Raan in 2004, providing one empirical thread that Wagner wove into a broader theoretical synthesis.

Key Ideas

Arrival does not guarantee activation. An innovation may be fully functional from the moment of its emergence and still require a receptive environment for its effects to register.

Environmental readiness is topological. The landscape into which an innovation arrives must contain the adjacent knowledge structures that make acceptance accessible from the positions the receiving community occupies.

Dormancy duration correlates with transformative impact. The longest-sleeping beauties, upon awakening, often produce the largest transformations — because the distance between the innovation and its environment's original position was greatest.

Quality matters less than context. The quality of Mendel's genetics, continental drift, and the perceptron did not change during their dormancies; what changed was the receiving landscape.

AI systems harbor dormant capabilities. The parameter spaces of current models contain adjacent configurations whose behaviors have not yet been elicited — beneficial and harmful capabilities awaiting activation.