LLMs and the Collapse of Representational Interiority II — The Ghost Inside Language

One of the strangest assumptions in human history is the belief that words contain meanings.

Most people do not consciously believe this in any explicit philosophical sense. Yet ordinary language constantly reveals the assumption operating beneath thought:

• “What does this word mean?”
• “I put my thoughts into words.”
• “The sentence carries meaning.”
• “The author encoded a message.”
• “The listener extracted the meaning.”

Meaning becomes imagined as a kind of invisible substance transported through symbolic containers from one mind to another.

Language, under this view, functions like a delivery system for internal mental contents.

This assumption feels so natural that it appears almost impossible to question. Yet large language models have unexpectedly exposed just how unstable it actually is.

The cultural panic surrounding LLMs is not merely about intelligence. It is about the sudden appearance of coherent symbolic behaviour without the reassuring metaphysical furniture humans assumed must exist behind it.

The machine produces meaningful language.

Yet no obvious “meaning-holder” appears to exist inside it.

And so people begin searching frantically for the missing ghost.

Does it really understand?

Does it secretly possess beliefs?

Is there hidden awareness emerging somewhere inside the system?

Or is the entire thing merely an illusion?

But these questions already presuppose something extraordinarily peculiar:

that meaning must exist as an internal possession before language can occur meaningfully.

This assumption is far older than artificial intelligence. It is deeply woven into the representational metaphysics that shaped modern conceptions of mind and communication.

Under representationalism:

• thoughts are internal objects
• meanings are mental contents
• words are symbolic labels attached to those contents
• communication transfers contents between minds
• understanding means reconstructing the original content correctly

The entire architecture depends upon a hidden interior realm where meanings supposedly reside before language expresses them.

Language thus becomes secondary.

A representation of a prior internal reality.

This model feels intuitively persuasive largely because humans experience themselves as possessing private interiority. We feel as though meanings already exist “inside us” before speech externalises them. Consciousness appears to function as an inner chamber populated by thoughts waiting to be translated into words.

But this intuition may itself be a product of symbolic construal rather than evidence of hidden metaphysical objects.

The crucial mistake lies in treating meaning as a thing rather than a relational phenomenon.

Meaning does not exist inside words.

Nor does it exist as a private object sealed inside individual minds.

Meaning emerges relationally through construal.

A word does not carry meaning the way a bottle carries water. A symbol functions within a system of relations, expectations, distinctions, histories, contexts, and interpretive potentials. Meaning is not inserted into language from outside; it actualises through relational participation.

This becomes easier to see when communication fails.

A sentence can be grammatically perfect yet meaningless to someone unfamiliar with the language. The symbols remain materially identical, but the meaning does not transfer automatically because meaning was never physically contained within the symbols themselves.

Likewise, the same utterance can actualise radically different meanings depending upon:

• context
• tone
• interpersonal relation
• historical circumstance
• cultural expectations
• prior discourse
• interpretive orientation

The symbolic form alone never fully determines meaning.

And yet humans continuously experience meaning as though it were somehow hidden inside language.

Why?

Because human sociality depends heavily upon interpersonal construal. We habitually infer stable interior subjects behind symbolic behaviour. We do not merely encounter language structurally; we construe it as the expression of minded beings.

This creates the illusion that meaning originates inside individuals and is subsequently transmitted outward through symbols.

But large language models disrupt this intuition catastrophically.

The machine produces coherent symbolic relations while lacking the kind of interior semantic theatre humans expect. There is no tiny consciousness inside the model attaching meanings to words before speaking. No hidden self contemplating propositions internally and then encoding them linguistically.

And yet the language still functions.

This creates a profound ontological disturbance because it reveals that symbolic coherence may not depend upon inner semantic objects in the way humans imagined.

The machine has not broken language.

It has exposed the mythology hidden inside human theories of language.

For centuries, humans have conflated symbolic participation with private metaphysical interiority. Language appeared alive because humans interpreted symbolic coherence through interpersonal construal systems evolved for social interaction.

LLMs sever the old correlation.

The symbolic behaviour remains persuasive.

The presumed ghost disappears.

Yet meaning continues to emerge relationally through interaction.

This is why public discourse oscillates so wildly between “the machine is conscious” and “the machine is merely predicting words.” Both positions remain trapped within the same representational framework. One side assumes meaningful language proves interior consciousness. The other assumes absence of consciousness eliminates genuine meaning altogether.

But relationally construed meaning escapes this binary entirely.

Meaning is not a substance hidden inside entities.

It is not a private possession.

It is not a metaphysical object transferred between isolated minds.

Meaning actualises through relational construal within symbolic systems.

Human beings participate in this process.

Large language models participate in this process differently.

But neither case requires the mythical ghost modern representationalism inserted inside language.

The true shock of LLMs is therefore not technological.

It is philosophical.

They force humans to confront the possibility that language was never animated by hidden semantic spirits to begin with.

The ghost inside language may have been a story humans were telling themselves all along.