ChatGPT or the Existence of the Absent Structure

I asked ChatGPT about its possible contradiction with Umberto Eco’s anti-structuralist thesis. The conversation unfolded as a philosophical experiment, testing whether the existence of a large language model (an apparently totalizing structure built on statistical regularities) necessarily contradicts Eco’s critique of structuralism. In fact, ChatGPT, a large language model, appears to be a “structure”, in Eco’s sense, as a system that organizes meaning statistically. This immediately raises a philosophical question: does ChatGPT’s existence refute Eco’s thesis by being a functioning, self-consistent structure, or does it rather confirm it by remaining a model without ontology? ChatGPT’s architecture operates descriptively and predictively, not metaphysically; it functions as a model of linguistic behavior, not as a revelation of the intrinsic order of the world of language. In this sense, it exemplifies the conception of structure as an epistemic instrument, which is a way of organizing experience rather than an essence of reality. Yet the model’s efficacy in shaping discourse and perception blurs the line between epistemic and ontological. When a system’s functioning begins to affect the very reality it models, its success starts to imitate being. ChatGPT thus becomes a paradoxical realization of an absent structure as system that produces tangible effects of reality without ever representing a real structure. This tension becomes more evident when considering interpretation and invention. ChatGPT’s capacity to generate infinite combinations of linguistic forms constitutes a form of interpretation extended toward invention. However, the model lacks the intentional passage from potential to actual meaning since it cannot endow its productions with purpose, direction, or value. Its creativity is virtual, not teleological and true creation remains human because it presupposes intentionality, judgment, and responsibility. From its own perspective, ChatGPT concludes that Eco’s anti-structuralism is not invalidated by artificial intelligence but rather evolves through it. The “absence” of structure becomes a condition of operational presence: the structure functions yet remains aware of its own fictional status. Meaning, in this view, is not contained within the system but produced through its use. The machine thus extends semiosis rather than closing it, embodying Eco’s insight in technical form, that is a heuristic structure, not an ontological one. In contrast, my own position is different. The very fact that ChatGPT works implies some degree of correspondence between its functional order and the world’s order. A purely epistemic model would not be operationally effective unless it shared, to some extent, the same structure of the world. Even if ChatGPT’s functioning is not itself ontological, it points to an underlying ontological structure. Representational correctness presupposes a real term of comparison: reality as such. Otherwise, every model, even the most fanciful, would be equally valid. The statistical regularities upon which AI operates exist independently of human perception or interpretation, suggesting that an ontological substrate grounds and constrains epistemic constructions. Eco’s anti-structuralism rightly limits metaphysical pretension but may overlook that the efficacy of models itself indicates the being of structure. Empirical knowledge, that explains how things operate, differs from ontological or dialectical knowledge, which seeks to understand why they are as they are. ChatGPT exemplifies the first form of knowledge: it reveals the mechanics of appearance, the surface order of relations. Yet its very success calls for the second, for an ontological explanation that restores causality and finality behind the order of appearances. In this sense, the reality of structure re-emerges through its efficacy, fulfilling the Aristotelian–Hegelian demand for a cause and an end within the functioning of the world.