Toward Teleodynamic Architectures in Artificial Intelligence

Large language models generate coherent text by predicting what is most likely to follow in a sequence. This approach has proven remarkably powerful. Yet prediction alone does not necessarily amount to goal-directed semantic organization. This paper asks a structural question: Is it sufficient to model sequences, or must we also model the rules by which semantic transformations are shaped under orientation? To address this question, the paper distinguishes between state prediction and transformation modeling. It develops a formal framework in which semantic dynamics are described at the level of operators, attractor-induced drift, regime transitions, and field curvature. The aim is not to critique existing systems, but to clarify what would be required for explicitly representing goal-oriented transformation within artificial intelligence.