The Art of Memory
In The You On AI Encyclopedia
The technique is simple in principle. To remember a list, the practitioner imagines a familiar building — a house, a church, a palace — and places a vivid image associated with each item on the list in a specific location within the building. To recall the list, the practitioner mentally walks through the building in a fixed sequence, encountering each image in its location and retrieving the associated item. The images are most effective when they are striking, strange, emotionally charged; the locations are most effective when they are well-established, spatially distinct, stably ordered.
The technique's power was recognized by every classical and medieval rhetorician. Orators memorized speeches of hours through memory palaces. Lawyers memorized case law through imagined libraries. Medieval monks memorized scripture through architectural scenes. The Renaissance extended the technique into metaphysical ambition: Bruno's memory theaters encoded the entire structure of the cosmos, proposing that the trained memory could hold the universe.
Calvino's invocation of the technique in the memo on visibility is precise. The art of memory demonstrates that the visual imagination is not a decorative faculty but a cognitive architecture — a spatial scaffolding on which knowledge is built and from which it is retrieved. The inner cinema Calvino wished to preserve is structurally identical to the memory palace: an internal space, generated by the consciousness that inhabits it, whose architecture gives shape to what the mind can hold and what it can find.
The AI moment poses a specific threat to this architecture. A large language model has no architecture in this sense. Its memory is distributed across billions of parameters in a way that has no spatial analogue. It does not place images in rooms. It does not walk through corridors. It retrieves through attention mechanisms over token sequences — extraordinarily powerful and entirely non-spatial. The model does not know where anything is, because 'where' is not a category its architecture supports.
The consequence for users of the model is that the external supply of descriptions, however vivid, does not provide the spatial architecture that the internal memory palace provides. The descriptions are retrievable only through external queries to the model. They do not become part of the user's internal architecture. They do not give the user an expanded memory palace. They give the user access to a vast retrieval system whose logic the user does not inhabit. The user's own memory palace — the internal architecture on which cognition is built — may atrophy through disuse, replaced by the externalized retrieval the model provides.
Origin
The technique is classical, with precursors in Simonides of Ceos (5th century BCE) and formalization in the anonymous Rhetorica ad Herennium (c. 80 BCE), Cicero's De Oratore, and Quintilian's Institutio Oratoria. It was elaborated through the medieval period (Thomas Aquinas's treatment in the Summa Theologica), the Renaissance (Bruno, Camillo, Ramus), and into the early modern period before being largely displaced by printed reference tools.
Key Ideas
Images in architectural locations. The technique exploits the mind's spatial imagination as a scaffolding for memory and retrieval.
Striking images, stable loci. The images should be vivid; the locations should be well-established. Effectiveness depends on both.
Memory as spatial architecture. The technique demonstrates that cognition is not merely abstract storage but structured inhabitation of an internal space.
The Renaissance memory theaters. Bruno and others extended the technique into metaphysical systems, proposing the trained memory as cosmological instrument.
The AI-age risk. The language model has no spatial architecture; reliance on external retrieval may atrophy the internal memory palace that cognition requires.