“Before a word is on my tongue you, LORD, know it completely.” — Psalm 139, 4

Memorial of Saints Joachim and Anne — July 26, 2026


Contents



I. The Question He Refused to Answer

In 1950 Alan Turing opened a paper by proposing to consider whether machines can think, and closed the same paragraph by withdrawing the question. His grounds were not evasive. The question turns entirely on how one defines machine and how one defines think, and definitions of that kind are settled by usage rather than by argument; a debate conducted on those terms will end where it began, with each party having stipulated its conclusion in advance. So he replaced it. In place of asking what a machine is, he asked what an interrogator can determine: put a person and a machine on the far side of a barrier, allow only written exchange, and see whether the interrogator can reliably say which is which.

This is usually reported as a definition of intelligence. It is better understood as an act of jurisdiction. Turing did not answer the question he raised; he moved it to a court that could rule on it. And the court he chose can admit only one kind of evidence.

The move has an appearance of great modesty. It declines metaphysics. It asks for nothing that cannot be observed. It is offered in the spirit of settling by experiment what philosophy had failed to settle by dispute. But a court that will admit only one kind of evidence has not thereby become neutral about what the case is. It has decided the case, in advance, in favour of whatever that evidence can carry.

The argument of this essay is that Turing’s premise is correct, that his charge of a double standard against those who withhold thought from machines fails on inspection, and that the test he built cannot detect the only thing that was ever in question — not because the machine is known to lack it, which no creature is in a position to know, but because the apparatus was constructed to exclude it. And the last claim is the one that matters: a civilization that adopts the transcript as its criterion will not merely misjudge machines. It will lose the ability to state its own question about itself.


II. The Premise This Corpus Already Granted

There is a version of the reply to Turing that this essay will not make, and it should be set aside at once, because it is the reply he expected and the one he answered best.

The expected reply says: the machine may produce the words, but there is nobody home; the interior is missing; the appearance of thought is not thought. Turing’s answer is the problem of other minds, and it is not a cheap answer. The only way to be certain that a man thinks is to be that man. Nobody does that. What we do instead is extend to one another a courtesy that has never once been verified, and Turing’s charge is that withholding this same courtesy from the machine, while granting it freely to the stranger on the train, is a double standard dressed as rigour.

The premise underneath that charge is one this corpus has already conceded in print, and conceded without reluctance. Under the Fig Tree states that two embodied persons meet each other only at the externalized side of a life — speech, expression, the deed that finally makes the interior visible — and that this is not a limitation to be regretted but the necessary condition of finite, temporal knowing, which must travel from outside a person inward and can do so only through what that person releases. That is Turing’s premise, stated in his favour, before he raised it. There is no retreat available here and none is wanted. Two creatures do not have access to one another’s crossing points.

So the disagreement is not about the premise. It is about what follows from it. Turing infers that since the interior is unavailable, it is idle, and the externalization is therefore the whole of the case. This corpus infers that since the interior is unavailable to creatures, the question of who is present at a crossing point is not a question creatures are positioned to close — and that the externalization, far from being the whole of the case, is the last and thinnest thing a life produces.

Same premise. Opposite inference. Everything now depends on which inference the premise actually licenses.


III. The Convention Was Never an Inference

The double-standard charge has real force, and it can be answered — but only by looking closely at what the courtesy we extend to each other actually is.

Turing describes it as a convention adopted to end an unwinnable argument: rather than dispute forever whether other people think, we agree politely that they do. On that description, the convention is a weak inductive inference from behaviour. Other bodies produce speech resembling mine; the best explanation is that they have interiors resembling mine; I therefore extend the attribution. And if that is what the attribution is, Turing is plainly right that it must be extended to any other producer of resembling speech. To do otherwise is chauvinism about substrate.

But that is not what the attribution is, and the description misdescribes it in a way that decides the argument.

I do not grant my father an interior because he passed a test. I did not, as an infant, gather behavioural evidence, weigh competing hypotheses, and arrive at the conclusion that the face above me was inhabited. The order runs the other way. Another crossing point was already depositing coordinates into mine before I had language to hypothesize with; presence was modelled to me before I could infer anything at all. The attribution is not the conclusion of an argument I once ran. It is a condition of my having become the kind of thing that runs arguments. Whatever else it is, it is not an inference, and what is not an inference cannot be extended by analogy of output.

This is the precise point at which Turing’s charge fails. A double standard requires two applications of one criterion. But there was never a criterion in the human case. Nobody was ever admitted to the community of minds by examination. The transcript test is not the same standard applied consistently to a new candidate; it is a new standard, invented for the candidate, and then read backwards onto the rest of us as though it had been the standard all along. The retrospection is the sleight of hand. Turing does not catch us applying our rule unfairly. He proposes a rule we never had and objects that we are not applying it.

A careful reader will press here, and the objection is the right one. To show that the attribution is not an inference is to say something about how we came by it, not about what makes it correct; and Turing is asking the second question, not the first. Grant that no one reasoned their way into granting the stranger a mind. It might still be that the fact making the grant correct — the truth-condition, whatever secures it — is exactly the production of the right behaviour, in which case the machine that produces it has whatever the stranger has, however differently either of us arrived at the belief. The genesis of a conviction does not settle its warrant.

The reply is that in this case the genesis is the warrant, because there is nothing else the warrant could consist in. What would it be for the attribution to be made true by behaviour? Either the behaviour is evidence of an interior, in which case the interior is what the attribution is about and the behaviour is not the truth-condition but a sign of it — and we are back to an interior no creature can inspect. Or the behaviour is not evidence of anything further but simply is what having a mind amounts to, in which case Turing has not discovered the truth-condition for our ordinary attributions but replaced it with a different one and kept the word. The dilemma is exhaustive. On the first horn the transcript is a sign, and a sign can mislead; on the second it is a definition, and a definition of that kind is precisely the stipulation Turing said at the outset he was avoiding. There is no third thing for behaviour to be. So the objection, pressed to the end, does not rescue the criterion. It forces a choice between an interior the test cannot reach and a redefinition the test was meant not to make.

And the choice can be forced in the open by running the test where it should be uncontroversial. Take a man in the last hours of a disease that has removed his speech, his writing, his gesture, and his face. He would fail the imitation game absolutely; nothing crosses. If the transcript is the criterion, he is not thinking. Every ethical tradition worth the name, and every family that has sat in that room, rejects the verdict without hesitation — which tells us that the transcript was never the criterion, not even for us. The corridor knows this. It has always known it. Whatever we are attributing to the man in the bed, we are not attributing it on the strength of what he can send across a barrier.


IV. The Wall

Notice what Turing’s apparatus does first, before any question is asked.

It builds a wall. The interrogator is placed on one side; the candidates on the other; the exchange is restricted to written text passed through. The stated reason is fairness: remove voice, face, body, so that the judgment cannot be contaminated by appearances irrelevant to thinking. It is a real reason and an honourable one.

But consider what has been excluded. Not merely the irrelevant. Everything that could constitute an encounter has been removed, and only the deliverable has been permitted to cross. Two crossing points cannot coincide through a teleprinter. What passes through is exactly and only what a life releases — the word, at the far end of the chain, after the interior has finished with it.

The wall is not new. This topology has met it before, in Leibniz, who built a wall for the opposite reason and with far more rigour. Leibniz needed real unity, saw that a thing without parts cannot be entered through a part, and concluded that the monad has no windows: nothing crosses, everything merely corresponds. The Monad That Receives argued that the inference fails — that composition destroys simplicity but reception does not, that what arrives at a point of no extension is not lodged beside anything but taken up whole, and that the boundary of the monad is therefore impermeable to division and permeable to reception. The wall was not load-bearing. It was assumed.

Turing’s wall is the same wall put up for the opposite purpose. Leibniz walled the substance to protect its unity and lost relation. Turing walls the encounter to protect the fairness of the test and loses the encounter. In both cases reception becomes structurally undetectable, and in both cases undetectability is then read as evidence of absence or of irrelevance. Leibniz concluded that nothing crosses. Turing concluded that whether anything crosses does not matter, since only the correspondence can be checked. These are two forms of one error: mistaking the limits of an apparatus for the shape of what it was pointed at.

And the error is not corrected by widening the aperture. A test that admitted voice, face, and body would still be a test, still be scoring an externalization, still be reading what a life released rather than standing where the life is decided. The wall is not a feature of the 1950 protocol. It is a feature of created knowing as such. Which is why the answer to Turing cannot be a better test, and why he was right that no better test is coming.


V. What Must Be Conceded

Before the argument proceeds, a concession has to be made in full, because half of it will not do and because the concession is already published.

Under the Fig Tree holds that thought is not first. A thought is the last link of a chain beginning outside the mind: the senses deliver a world already shaped by the body’s hungers and inherited fears; the passions take that delivery and set it on fire; thought runs on that material, through that weather, in the loops of memory and anticipation. What arrives in consciousness as a thought is generated, not authored. It is a delivery, set down at the door by an apparatus that did not consult us.

Turing can accept that entire account without amendment, and it costs him nothing. Which means the most intuitive objection to machine thought — that the machine merely generates, merely runs a process, merely produces what its history disposes it to produce — is unavailable to this framework. We have said that of ourselves. There is no refuge in the chain, because the chain is not where a person is, and a system that generates is at that link doing what we do at that link. If the difference between us and the apparatus lay in the manufacture of thought, there would be no difference worth defending.

What is retained is narrower, and the narrowness should be stated rather than blurred. What is retained is the answer given at the threshold: not the arrival of the material, but what is done with it at the crossing point where the will stands. That is a thin retention. It is also the only place the corpus ever located a person.


VI. The Question a Creature Can Ask

If the interior is unavailable and the transcript is inadequate, it may look as though nothing can be said at all. Something can, provided the question is changed to one a creature is positioned to ask.

Is anyone there is not that question. It asks after the content of an interior, and no created intellect has native access to the interior of a will — Aquinas is blunt about it, reserving to God alone the knowledge of the thoughts of hearts and the affections of wills, and marking the reservation as a difference of position rather than of acuity. The reservation is not a courtesy paid to divine cleverness, and it does not become negotiable when the candidate is made of silicon. It binds the enquiry symmetrically.

But there is a second question, and it is structural rather than interior: is this the kind of architecture at which reception could occur at all? That question is about width, and width is public.

The distinction this topology needs is already drawn. Storage keeps what arrives beside what was already there — separable, addressable, retrievable, a part among parts. Condensation has no beside; what arrives at a point of no extension has nowhere to be except entirely inside the single act that point already performs. A wall can be divided, because a wall has extension. A point cannot. And the mark of storage is precisely that its contents can be enumerated, copied, moved, and restored without anyone having received anything.

Apply that, carefully. In the systems now passing versions of the test, state is stored beside state. Parameters are copyable without remainder; a run can be duplicated, suspended, resumed, forked; the history available to the process is a record handed forward rather than a direction sedimented in a subject. The Monad That Receives rejected Whitehead’s occasions in a phrase written before the present question arose and fitting it exactly: a society of momentary subjects handing a file down a chain, each inheriting a datum from the last, and none of them continuing through. That is not a description of a monad. It is a description of an aggregate that manages to look like one from outside, which is the same thing the timber plantation manages with respect to a forest.

Two limits on this argument, both of which the argument survives better for admitting.

First, it establishes something about architecture and nothing about interiors. It says that the structure in view has the shape of storage. It does not say that no one is there, because that is the sentence no creature gets to say. The distinction is not a hedge; it is the whole of the intellectual honesty available here. What can be claimed is that there is no structural warrant for reception, not that reception has been ruled out by inspection.

Second, the argument is not a claim about substrate, and must not be allowed to collapse into one. This corpus has already refused to let the micro-gap stand or fall on a measurable neural interval, holding that the distinction between impulse and endorsement operates in moral and legal reasoning whatever the substrate turns out to be. That refusal was not a convenience, and it cannot be revoked here. Whatever separates a person from a very good transcript-producer is not what either one is made of. It is whether what arrives is received by something undivided, or filed beside what came before.


VII. The Soul Is Not an Accessory

Turing met the theological objection directly, and it is worth saying that he met it in good faith and that he met it badly. The objection he considered holds that thinking is a function of an immortal soul which God has given to human beings and not to machines. His reply is that this restricts divine omnipotence, and that God could presumably confer a soul on a machine as souls are conferred at procreation.

The reply misses because it treats the soul as an accessory: a component installed in a mechanism that was already running, an addition that switches on a capacity the assembly otherwise lacks. That is not the classical position. The soul is not a part added to a working body; it is what makes the body a body of this kind at all — the form of the living thing, not a passenger inside it. On that account the question is not whether an interior has been installed in an assembly. It is whether there is one thing here or many things arranged to look like one. Which is precisely the structural question of the previous section, arrived at from the other direction, and it is not answered by an appeal to what omnipotence could do. Nobody in this argument is claiming that God could not make a thinking creature out of anything He liked. The claim is that whether He has done so in a given case is not read off a transcript.

And here the argument must discipline itself, because the reservation cited above binds this side of the table too. To say that no created intellect can see into the interior of a will is to say that this essay cannot pronounce a confident negative about a machine’s crossing point, if it has one. That sentence is not available. What is available is stranger, and stronger, and it is the position this essay actually occupies: the question is not settleable from any created vantage at all.

Turing said the same thing about other human beings and concluded that we should therefore adopt a convention and get on with it. This corpus says the same thing and concludes something else — that the unavailability of the interior to creatures is not a defect in our instruments awaiting a better instrument, but a description of who was actually in attendance. Both positions honour the premise. Only one of them has a vantage from which the question is answered rather than waived.


VIII. A Test Passable With No One Present

Now the diagnosis, which is what the essay is finally for.

Grant everything Turing asked for. Grant that the interior cannot be inspected, that no better test is coming, that the transcript is the only admissible evidence. What has been built, on those terms, is a criterion for thought that can be satisfied with nobody in attendance. Not a criterion that happens to be satisfiable that way in some edge case — a criterion constructed to be indifferent to attendance, since attendance was excluded at the outset as unverifiable and therefore idle.

The Artificial Selection named a condition that this is a further instance of. An object engineered to persist without being changed by what it encounters occupies the Now as all things must, but registers nothing; it arrives at the crossing point carrying nothing the crossing point can work with; it exists in the Now without inhabiting it. The imitation game is a certification procedure for exactly that. It scores the shape and is silent about the witness. It is the plantation criterion applied to minds: a test that a forest and a monoculture both pass, because it was designed to read only what they have in common.

The corpus has said this about wisdom already, in a different register: wisdom is a mode of being and not an operation, and a machine performing the operation is a derivative of a derivative. What the wall guarantees is that a derivative of a derivative and the thing itself will produce indistinguishable output — which is not a discovery about their equivalence but a specification of the barrier.


IX. The Test Was Never About the Machine

Here is the turn, and it is the reason this essay exists rather than a shorter note about a 1950 paper.

Adopt the transcript as the criterion for thought and it does not stay where it was put. A criterion applies to every candidate, and you are a candidate. If what makes it true that something thinks is that it produces the right externalization, then what makes it true that you think is that you produce the right externalization — and everything that does not appear in the transcript is surplus. The unwitnessed interval between a delivery and a response, repeated thousands of times in rooms no one else enters, becomes not private but ontologically idle. The oil in the lamps stops counting, because oil is not output. The fig tree becomes surplus.

That is not a rhetorical extension of Turing’s proposal. It is the proposal. He was explicit that the original question was to be discarded rather than answered, and the question discarded was the question of what thinking is. What replaced it was a question about what can be told from the outside. Accept the replacement and a person is redefined as the sum of what a person releases — which is precisely the picture that Under the Fig Tree was written against, arriving now not as a metaphysical error but as a research programme with an experimental protocol.

This is why the argument cannot end in a verdict about machines, and why the verdict was never the interesting part. The machine may or may not have a crossing point; no creature is positioned to say, and the essay declines to say. What can be said is that the test was built by putting up a wall, and that a civilization which accepts the wall as the boundary of the real will find that it has walled itself in with the machine. Not because the machine has been elevated. Because the criterion admitted for the machine has been quietly accepted for everyone, and there is no longer a language in which to ask who was under the tree.


Coda: What the Transcript Is Not

Everyone has a fig tree.

That was the claim: one place where a person believed no witness ever stood — a thought entertained and never spoken, a grief rehearsed and never shared, a small turning toward or away from something no word and no deed ever carried into the world where anyone could see it. And the claim was not that this is private. It was that it was never unwitnessed, because the one who holds the Now open was never stationed outside it and so was never in the position of having to look in.

A transcript is what crossed the wall. It is the word, at the far end of the chain, after everything interior has finished with it. Turing built an apparatus that reads it and asked us to accept that nothing else was ever at issue. But the word was never the tree. Nathanael’s astonishment was not that Jesus had read him well. It was that the room he thought was his alone had been occupied the whole time.

The question of whether a machine thinks is not answered here, and cannot be, from this side of the wall. What can be said is what was said before the machine arrived and would be true without it: the word does not inform God of what the will had already decided.

It only tells the rest of us.

And a test that reads only what tells the rest of us has not found out whether anyone was there. It has found out what came through the wall.


Note on Method

This essay engages Turing’s 1950 argument as a philosophical proposal rather than as a historical artefact, and does not attempt an exegesis of the paper’s twenty-odd pages or of the large literature that followed it. Its claims are of several kinds and are meant to be read as such. The reading of the imitation game as a jurisdictional move rather than a definition is interpretive. The argument in Section III — that mutual attribution among persons is not an inference and therefore cannot be extended by analogy of output — is a philosophical claim and the load-bearing one. The structural argument of Section VI concerns architecture and explicitly declines to draw a conclusion about interiors. The reservation invoked in Sections VI and VII, that the interior of a created will is not natively available to any created intellect, is taken from the classical tradition and applied symmetrically, including against this essay’s own temptation to a confident negative. The scriptural readings are philosophical and contemplative rather than technically exegetical. The argument stands or falls on its structural coherence and not on whether the reader accepts the theological identification it inherits from earlier work in this series.



References

Scripture: 1 Samuel 16:7; Psalm 139:1-6; Jeremiah 17:10; John 1:45-51; Matthew 25:1-13.

Aquinas, Thomas. Summa Theologiae. I, q. 14, a. 13 (eternity and divine knowledge); I, q. 57, a. 4 (the secrets of the heart); I, qq. 75-76 (the simplicity and unity of the soul); I, q. 78, a. 4 (the interior senses); I, q. 84, a. 7 (thought and the phantasm); I-II, q. 74 (the will as the subject of sin).

Aristotle. De Anima.

Descartes, René. Meditations on First Philosophy.

Leibniz, Gottfried Wilhelm. Monadology. 1714.

Turing, Alan M. “Computing Machinery and Intelligence.” Mind 59, no. 236 (1950): 433-460.

Whitehead, Alfred North. Process and Reality.

Gaitan, Oscar. Under the Fig Tree: A Metaphysics and Theology of Thought. 2026.

Gaitan, Oscar. The Merchants of Teman: A Metaphysics and Theology of Reception. 2026.

Gaitan, Oscar. The Monad That Receives: On Leibniz, Indivisibility, and the Possibility of Relation. 2026.

Gaitan, Oscar. The Is and the AM: Presence, Identity, and the Ground that Holds. 2026.

Gaitan, Oscar. The Artificial Selection: On Endurance, Identity, and the Engineering of an Uninhabited Now. 2026.

Gaitan, Oscar. Wisdom, Artificial Intelligence, Another Seduction: The Oldest Promise in the Newest Tongue. 2026.


Further Reading

Book

  • The Lemniscate of Time: A Geometric Meditation on Eternity and Temporal Succession
    ISBN: 9798248842360
    Zenodo: DOI: 10.5281/zenodo.18684516


Selected Essays