Is There Meaning Without Anyone Meaning Anything? A Structuralist Proposal for Artificial Semantics

Recent developments in the field of deep neural networks (DNNs) and large language models (LLMs) have led to the spread of systems capable of producing coherent texts, drawing inferences and acting in a contextually situated manner. Such systems appear to exhibit semantically relevant behaviour, despite lacking features traditionally considered essential for meaning, such as sensory experience or human-like intentionality. This situation has revived the philosophical debate on the nature of meaning in intelligent machines. Seeking to contribute to this debate, this article proposes a change of perspective, linking semantic organisation in artificial systems not to symbols or embodiment, but to the concepts of structure and network . Through a trajectory that draws on the historical epistemology of AI, this paper explores a structural conception of semantic organisation in artificial systems. Our claim is that LLMs may exhibit a form of structurally organised semantic competence based on the internal relational organisation of networks rather than on symbols or bodily experience. Through the analysis of distributed representations and relational spaces embedded in LLMs, this paper aims to clarify in what sense such systems can exhibit forms of semantic organisation without possessing human understanding or phenomenal intentionality. This perspective has significant epistemological implications, suggesting a reconsideration of traditional concepts of representation, knowledge, and meaning. In this sense, this article aims to contribute to contemporary debates on the nature of semantic organisation in artificial systems.