Twelve Passes: A Structural Study of Artificial Intelligence and Natural Systems

Twelve Passes: A Structural Study of Artificial Intelligence and Natural Systems examines twelve structures drawn from contemporary neural-network language systems through twelve compressed poems composed according to the principles of Absolute Composition.

The sequence begins with text and its transformation into computational representations, proceeds through attention and internal-state transformation, and then moves into model training through sampling, forward propagation, loss, backpropagation, parameter adjustment, iteration, and convergence.

Each poem begins with a triad drawn from artificial-intelligence systems. These triads isolate structural relations such as Text - Tokenization - Tokens; Tokens - Embedding - Vectors; Tokens - Attention - Context Tokens; Signals - Weighting - Selected Signal; Prediction - Loss - Error; and Gradient - Update - Revised Weights. Their purpose is structural and compositional rather than encyclopedic.

Each computational relation is placed in correspondence with an independently observable natural process. Rock fractures. Water moves. Sediment accumulates. Plants bend and grow. Snow shifts. Sand cascades. Oscillation diminishes. These natural systems do not function as metaphors for artificial intelligence, and the computational operations do not explain nature. The two domains remain materially and causally distinct.

Their correspondence lies in behavior. A continuous structure may divide into discrete units. Separate elements may acquire relational position. Multiple influences may be differentially weighted. An earlier condition may persist within a later one. A system may generate an outcome, register discrepancy, propagate consequence, alter its internal configuration, repeat a process, and approach a more stable condition.

The poems treat terms such as attention, prediction, loss, error, and learning according to their technical functions within computational systems. They do not attribute consciousness, intention, desire, judgment, or awareness to artificial intelligence. The natural processes are likewise selected for behaviors that do not depend upon human decision.

Absolute Composition, the structural methodology used throughout the collection, begins with observable systems and asks how they behave. Composition emerges through arrangement, relation, pressure, threshold, transformation, recurrence, compression, and structural correspondence rather than symbolic substitution. Mechanisms may remain absent when their effects are physically recoverable from the conditions presented.

Twelve Passes does not argue that artificial intelligence is natural, that nature is computational, or that the two domains are equivalent. Instead, it examines whether materially different systems can exhibit verifiable correspondences in the ways they divide, weight, transform, transmit, revise, and stabilize information or matter.

This standard student edition includes a preface, reading guide, and appendix supporting classroom, scholarly, and independent study.