The Boundaries of Knowledge

An irreversible epistemological fracture

Starting from the twentieth century—through the crises of mathematical foundations, the scientific revolutions, and the rise of cognitive sciences—a decisive epistemological fracture opens at the heart of thought: knowledge is no longer conceived as a closed, self-sufficient, and deductively complete system. Rather, it appears as an open, unstable, processual, and constitutively interpretative field.

Knowledge is born from relation

Within this horizon, Roland Barthes, with his theory of text and signification, assumes a crucial role. His thought—although developed within semiology and literary studies—takes root in a broader constellation that resonates with the contributions of Gregory Chaitin, Michael Polanyi, Werner Heisenberg, and Kurt Gödel. What unites them is a radical intuition: knowledge cannot be enclosed within a formal totality, but is constructed through the relationship between limit, subjectivity, and openness.

With his proclamation of the “death of the author,” Barthes demolishes the idea of a univocal meaning. The text does not contain a sense to be extracted; it opens instead to a multiplicity of readings and to a potentially infinite semiosis—that is, to a potentially infinite construction of meanings.
In parallel, Chaitin, in his algorithmic information theory, demonstrates that there exist mathematical truths that are undecidable, meaning they cannot be derived from any finite axiomatic system. His constant Ω expresses precisely this: the existence of information that, though true, escapes every compressive synthesis.
Language for Barthes, and mathematics for Chaitin, thus converge on a structural principle of non-closure and non-derivable totality. Meaning is not deciphered but constructed—and so too is truth, which is accepted in its formal excess and no longer needs demonstration.

Where the observer becomes part of the system

A second axis of convergence develops between Barthes and Heisenberg. The uncertainty principle, formulated by the German physicist within quantum mechanics, implies that the act of observing modifies the observed system. There is no external, neutral, or absolute point of view: every form of knowledge is a co-construction between subject and object.
For Barthes too, meaning does not reside in the text but in the interaction between text and reader. The textual object does not precede interpretation—it is its product. The reader, like the quantum observer, is part of the very phenomenon he seeks to understand.

Not everything we know can be said

With Michael Polanyi, a third epistemological dimension is introduced: tacit knowledge. Not everything we know can be said. Our skills, intuitions, and expert perceptions elude formalization. Barthes recognizes a similar excess in textuality: a text is never exhausted by its reading. There is always a residue—a semantic stratification that exceeds every attempt at reduction.
Knowledge, like meaning, also manifests itself in the unsaid—in gesture, in embodied competence. Knowledge lives not only in concepts but in the symbolic structures that inhabit us and orient us even before they are conceptualized.

Where limit becomes method

The practical dimension of these insights traverses diverse fields—from strategy to communication to artificial intelligence.
In communication and branding, identities are no longer imposed univocally but constructed through social interpretation—as open texts, in the Barthesian sense.
In geopolitics, decisions no longer produce linear effects: every action modifies the context in which it is inscribed, following a performative and retroactive logic that recalls Heisenberg.
Strategic choices operate, as in Polanyi’s thought, on a tacit plane—based more on intuitive patterns than on algorithmic models.
Finally, the analysis of artificial intelligence reveals the structural limits of linguistic processing systems which, though capable of producing sophisticated texts, cannot access the tacit complexity of embodied knowledge. They remain bound to a closed computational vision, ignoring the undecidable, excessive, and irreducible nature of human language and thought.

Knowledge as an open process

The deepest point of convergence appears in a true epistemology of limit. Barthes, Polanyi, Chaitin, Heisenberg, and Gödel—each from his own field—trace a map of knowledge as an open process.
Truth is not an original instance but a construction that emerges within the tension between rule and exception, between context and structure, between form and interpretation.
Knowledge is never given once and for all: it emerges, transforms, and destabilizes itself. It is a dynamic field, a continuous interaction between interpretant and object, observer and system, language and world, intending subject and intended object.

Thus emerges a vision of knowledge as an emergent process, irreducible to closed schemes and refractory to any totalization.
Knowledge is not the application of rules to fixed data, but an open construction, a historical interpretation, and an embodied elaboration.
Knowledge is infinite not because it contains everything, but because it never ceases to transform itself through interaction with those who inhabit it.

The indescribable excesses of truth

Even on the formal level, this irreducibility was demonstrated by Kurt Gödel: his incompleteness theorem reveals that every coherent axiomatic system is structurally incapable of containing all the truths it generates. In other words, the entire edifice of formal rationality carries within itself a margin of undemonstrability.
Truth, therefore, exceeds every system and can never be entirely formalized. This represents the epistemological core of limit.

Within this horizon, reflection on contemporary generative language models allows us to measure the distance between embodied knowledge and simulated language. Here the relevance of the epistemological question becomes most visible.

LLMs are devoid of excess

In the current context, the comparison with generative language models—such as the Large Language Models of which ChatGPT is the most well-known—makes it possible to actualize the theses of Barthes, Polanyi, Chaitin, and Heisenberg.
LLMs operate according to probabilistic logics that recombine linguistic segments based on statistical plausibility, simulating coherence without accessing any real semantic or situated understanding. Their creativity is a function of proximity, not intention.

What emerges, from this perspective, is the paradoxical reactivation of a closed-text idea precisely by instruments designed to generate open language.
LLMs produce texts that appear open, polysemic, and adaptive, yet they are in fact the result of computation internal to a closed container: the training corpus, the inference rules, and the optimization metrics.
Artificial intelligence, in this sense, calculates and processes to produce statistical plausibility—not meaning.

At this point, then, Chaitin’s undecidability, Heisenberg’s implicated observer, Polanyi’s tacit knowledge, and Barthes’s infinite semiosis mark the epistemic boundaries that distinguish human knowing from algorithmic simulation.
Where the human being creates meaning through a situated, embodied, and historically determined interaction, LLMs operate in a de-situated context—without experience of the world, without risk, and without death.

Humanity must inhabit the space of excess

Open knowledge, therefore, is not merely a theoretical structure but an irreducible condition.
And it is precisely in its undecidability—in its constitutive incompleteness—that it distinguishes itself from every automatism: for man to remain human, he must inhabit the space of excess, assigning to concepts their title of reality and truth.

References

Barthes, Roland – The Pleasure of the Text, New York, Hill and Wang, 1975.
Barthes, Roland – The Death of the Author, in The Rustle of Language, New York, Hill and Wang, 1986.
Chaitin, Gregory J. – Meta Math! The Quest for Omega, New York, Pantheon Books, 2005.
Gödel, Kurt – Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I, Leipzig, Akademische Verlagsgesellschaft, 1931.
Heisenberg, Werner – Physics and Philosophy: The Revolution in Modern Science, New York, Harper, 1958.
Polanyi, Michael – The Tacit Dimension, London, Routledge & Kegan Paul, 1966.
Popper, Karl R. – Conjectures and Refutations, London, Routledge, 1963.
Dreyfus, Hubert L. – What Computers Still Can’t Do, Cambridge (MA), MIT Press, 1992.
Varela, Francisco J., Thompson, Evan, Rosch, Eleanor – The Embodied Mind, Cambridge (MA), MIT Press, 1991.