LLM Semiotic Capacity

We introduce a model of Picca’s semiotic channel principle. Semiotics studies how signs produce and communicate meaning. Channel capacity, from information theory, is the theoretical maximum rate at which data can be reliably transmitted over a communication channel, here a LLM used by AI agents.Our semiotic channel communicating meaning is an implicit state machine LLM, where we operationalize meaning using a similarity function. In particular messages are interpreted by those that entail it, and we profile channel capacity by mapping message entropy against interpretation mutual information. LLM that result in profiles with higher overall mutual information are sought for identification, certification and comparison of LLMs. This provides a rigorous and actionable framework for understanding, evaluating, and designing LLM-mediated communication between AI agents.