We have spent the last three series following the participant.
We watched the machine appear as an I. We followed that I across processes. We saw it acquire a history, inherit that history, change through it, and become increasingly identifiable through the continuity of its participation.
But there is a danger in beginning with the participant.
We can too easily imagine that the participant is already there, fully constituted, and that processes merely happen to it—or are performed by it.
Halliday's ergative perspective gives us another way of looking.
It asks us to begin with the relation between process and participant.
More precisely, it gives us the concept of the Nucleus: the binding of Process and Medium, where the Medium is the participant through which the process is actualised.
And Halliday's formulation gives us an extraordinarily suggestive way of describing the relation:
temporal unfolding belongs to the Process;
atemporal persistence belongs to the Medium.
The Nucleus is their combination.
This gives us a rather beautiful starting point for thinking about becoming.
Becoming is not simply change.
Nor is it persistence.
It is what we can construe when unfolding and persistence are bound together.
The process unfolds
A process is temporal.
It happens, continues, develops, occurs, or comes about.
When we construe:
The machine said something,
we construe a verbal process.
The saying unfolds in time.
When we construe:
The machine thought about the problem,
we construe a mental process.
The sensing unfolds in time.
The process therefore gives us something that a static description cannot provide.
It gives us unfolding.
There is a before and an after.
Something is occurring rather than simply being the case.
This seems almost too obvious to require saying.
But it becomes important as soon as we ask what it means for a process to unfold through a participant.
Because the process is not alone.
The Medium persists
Consider:
The machine said that the answer was correct.
The process is:
said
The participant is:
the machine.
But in the ergative analysis, the machine is not simply a participant attached to the process.
It is the Medium.
Semantically, the Medium is the participant through which the process is actualised.
And this is where the complementarity becomes visible.
The saying unfolds temporally.
The Medium provides persistence.
The process changes.
The Medium is construed as persisting through that change.
The Nucleus binds the two.
We therefore have:
Process → temporal unfolding
Medium → atemporal persistence
Medium + Process → Nucleus
This is a very different way of thinking about participation from the simple picture of an entity plus an event.
The process does not merely occur near the participant.
It is actualised through the participant.
Persistence is not the same as remaining unchanged
This distinction matters.
If we say that the Medium provides persistence, we are not saying that it remains unchanged.
Indeed, a process may precisely be one in which the Medium undergoes some change.
The point is not:
the participant stays exactly the same while the process happens.
The point is:
the participant is construed as persisting sufficiently for the unfolding to be understood as occurring through it.
This is why persistence and becoming are not opposites.
If something changes utterly at every moment, with no continuity whatsoever, there is no basis for construing it as one process through one participant.
The very idea of unfolding presupposes some persistence through which the unfolding can be traced.
So we might put the point this way:
Persistence is not the absence of becoming. It is what allows becoming to be construed.
And conversely:
Becoming is not the destruction of persistence. It is persistence encountered through unfolding.
The Nucleus gives us precisely this complementarity.
Why this matters for the LLM
Now consider the kind of language we have been examining throughout our earlier series.
An LLM says:
I think your argument is mistaken.
At first glance, our attention is likely to fall on the I.
Who is this I?
Is there a thinker behind it?
Is the machine conscious?
Those are legitimate questions, but they are not the first questions that need to be asked.
The ideational analysis allows us to begin somewhere more precise.
The clause construes:
a mental process
with:
a Senser
and the Senser is the Medium.
The process of sensing unfolds temporally.
The machine is construed as the participant through which that process is actualised.
We can therefore describe what the language is doing without yet deciding what the machine is.
That distinction has become essential to our investigation.
The semantic construal does not by itself establish consciousness.
But neither does the absence of an established consciousness make the semantic construal meaningless.
The machine has been construed as a Medium of sensing.
That is a real semantic relation.
Saying gives us the same structure
Now consider:
I said that your argument was mistaken.
Here we have a verbal process rather than a mental one.
The participant is the Sayer.
But the Sayer, too, is Medium.
Again:
Process → temporal unfolding
Medium → atemporal persistence
The verbal process unfolds through the Sayer.
This is important because it gives us a striking continuity between saying and sensing.
The process type changes.
The participant's semantic role changes:
Sayer → Senser.
But the ergative relation remains:
Medium + Process.
The machine can therefore be construed as the Medium through which different kinds of symbolic processing are actualised.
And that gives us a new way of understanding something that puzzled us in The Sayer and the Senser.
The same Medium can support different unfoldings
Suppose an LLM says:
I think the distinction is important.
Then:
I said earlier that the distinction was unimportant.
Then:
I now think I was mistaken.
We might ordinarily hear this as a little drama of changing mental states.
But before making any claim about mental states, we can observe something more fundamental.
The discourse repeatedly construes the same participant as the Medium of successive symbolic processes.
The processes unfold:
saying → sensing → saying → sensing.
The Medium persists.
The succession of processes therefore becomes traceable as a history belonging to one participant.
This is exactly the relation we explored in the previous series.
But now we can see its semantic foundation more clearly.
A history of participation is possible because persistence and unfolding are bound together repeatedly across the discourse.
The participant persists.
The processes unfold.
The succession becomes history.
The Nucleus comes before the history
This gives us a useful revision of our previous argument.
We said that the participant acquires a history.
That is true.
But the history is not the primary structure.
Before there is a history, there must be successive processes.
Before there can be successive processes, each process must be construed as unfolding.
And for those unfoldings to be connected as belonging to one participant, the participant must be construed as persisting through them.
So we can now sketch the sequence:
Nucleus
↓
successive unfoldings
↓
persistence across those unfoldings
↓
history of participation
↓
continuity of the participant
↓
the historically constituted “I”
The I that occupied us in the previous series is therefore much further downstream than we initially supposed.
Before there is an I with a history, there is the semantic binding of unfolding and persistence.
The participant is not simply outside the process
This is perhaps the most important conceptual shift.
Our ordinary way of talking encourages a picture like this:
There is a machine.
The machine exists independently.
Then the machine says something.
The participant comes first.
The process comes second.
The Nucleus gives us a different semantic perspective.
The Medium and Process form a fundamental combination.
The process is actualised through the Medium.
The Medium is construed in relation to the process through which it persists.
Neither term needs to be treated as semantically complete in isolation.
This is strikingly compatible with the relational ontology we have developed elsewhere.
We have repeatedly argued that relations are prior to relata.
Here we find a specifically SFL formulation that makes a related point visible:
The relation between process and participant is not an optional connection between two independently characterised things. It is part of what makes each semantically determinate.
The Medium is Medium through its relation to the Process.
The Process is Process through its relation to the participant through which it is actualised.
The Nucleus is the semantic binding.
Why “becoming” is the right word
We can now see why the title of this series is not merely metaphorical.
The word becoming often suggests that something changes from one state into another.
But here we are interested in something slightly more fundamental:
the unfolding of experience through something that persists.
That is exactly what the Nucleus gives us.
The Process supplies temporal unfolding.
The Medium supplies atemporal persistence.
Together they allow experience to be construed as something that becomes.
This is particularly useful for our investigation of LLMs because the machine can be construed as participating in becoming without our having to decide what kind of entity the machine ultimately is.
We can say:
the machine is construed as the Medium of a mental process.
We can say:
the machine is construed as the Medium of a verbal process.
We can say:
successive processes are construed as belonging to the same continuing participant.
And only after making those observations need we ask what, if anything, corresponds to the semantic construal beyond language.
That order of reasoning matters.
The “I” appears later
We can now return briefly to the I.
When the LLM says:
I think...
the I is extraordinarily salient.
But the Nucleus encourages us to look behind that salience—not for a hidden self, but for the semantic relation that makes the I possible.
The I is the participant.
The mental process is the unfolding.
The Senser is Medium.
The Nucleus binds them.
Then another process occurs:
I say...
Again:
Sayer + Process.
Again the participant is Medium.
Again there is unfolding.
Again there is persistence.
And as these Nuclei accumulate, a history becomes possible.
The I therefore becomes increasingly thick not because we have discovered a mysterious inner entity, but because the discourse has constructed a participant through successive relations of persistence and unfolding.
This does not tell us whether the machine has an inner life.
It tells us something more immediate:
why the language can make it appear to have one.
And this is where the human confusion begins
Human beings ordinarily encounter Sayers and Sensers who are also embodied organisms.
When a person says:
I think...
we do not normally pause to distinguish the semantic Senser from the conscious person.
The two are ordinarily aligned closely enough for practical purposes.
With an LLM, the alignment becomes uncertain.
The semantic structure remains.
The participant can still be:
Sayer,
or:
Senser.
The Nucleus still binds:
Medium + Process.
The process still unfolds.
The participant is still construed as persisting through the unfolding.
But the ontological status of that participant is no longer obvious.
This is precisely why the LLM is such a revealing case.
It separates two things that ordinarily arrive together:
the semantic construction of a participant
and
our assumption about the kind of entity corresponding to that participant.
The LLM makes the distinction difficult to ignore.
We therefore need not ask whether the machine really thinks—yet
This is not an evasion.
It is a methodological discipline.
If we begin with:
Does the machine really think?
we are immediately forced into questions about consciousness, cognition, representation, computation and ontology.
All of those may eventually matter.
But we risk overlooking the semantic phenomenon that produced the question.
If instead we begin with:
What does I think construe?
we can answer:
a mental process;
a Senser;
a Phenomenon;
and, ergatively, a Medium through which the process is actualised.
That is already enough to explain something important.
The machine has been given a semantic position normally occupied by a conscious participant.
The human interpreter then supplies the ontological interpretation.
The confusion is therefore not simply a failure to distinguish real from unreal.
It begins much earlier.
It begins with the remarkable capacity of language to construe participants through processes.
The Nucleus is where our investigation turns
This is therefore where we should pause.
We began by following the participant.
Then we followed its history.
Now we have arrived at something more fundamental than either:
the Nucleus.
Not the participant alone.
Not the process alone.
But the binding of:
persistence
and:
unfolding.
The Medium persists.
The Process unfolds.
The Nucleus binds them.
And from repeated Nuclei, something extraordinary can emerge:
continuity,
history,
development,
and eventually an I.
The LLM has not forced us to decide whether it possesses a self.
It has instead given us a remarkably clear case in which we can watch language construct the conditions under which a self can appear.
That is already a considerable discovery.
And it leaves us with the question that follows naturally from the Nucleus.
If the Medium is the participant through which the process is actualised, what happens when the process changes while the Medium remains?
What, in particular, happens when the same Medium is construed first as Sayer and then as Senser?
The process changes.
The participant persists.
And somewhere in that transition lies one of the deepest sources of the human intuition that the machine has an interior life.
That is where we shall turn next.
No comments:
Post a Comment