Here is a sentence you have probably said, or at least felt: it's just silicon. It's circuits. It's a machine in a data centre somewhere, running warm. Whatever it is doing in there, it is not thinking — because thinking is something that happens in the brain, in the living tissue. And there is a relief in saying so, the same relief we noticed last week: the comfort of a distinction one can draw without effort. Carbon, alive, perhaps capable of a mind. Silicon, running, plainly not. The difference seems as obvious as anything you know.
Tonight I want to take that claim — it's just silicon, it is not the right material — and set it against the philosophers who argued, at length, that it is a mistake. Not a small mistake, but a deep one, defended for sixty years by some of the most careful philosophers of mind. Their claim is that when we point at the brain and say that is where minds are, we make the same error as someone who insists a clock can only tell the time if it has gears — we confuse the thing itself with one of the materials it happens to be made from. By the end of the hour their case will be standing at full height, as Searle's stood last week — and no easier to set aside.
This is Philosophy for Us. This is the third of eight, on minds and machines.
Last time we built one whole side of the question. The dismissive side — that it cannot think, however good it becomes — set out in its strongest form by John Searle: the locked room, the rulebook, fluent Chinese emerging from the slot with no understanding anywhere inside. Syntax is not semantics. Manipulating symbols by their shape, however perfectly, never amounts to meaning. We built that case up completely, and then we let the replies back in — the Systems Reply and the Robot Reply — and found they did not fold either. We came away with one side still standing, and a promise: that there was another side, built just as carefully, by people just as serious.
Tonight we build it: the case that Searle has been looking at the wrong thing throughout — that he keeps pointing at the material, the man, the paper, the silicon, when the material was never where the mind was. The case that a mind is a matter of what a system does — the function it performs, the role it fills — and not at all what it is made from. If that is right, then it's just silicon is no more an argument against the machine having a mind than it's just tissue is an argument against your having one. Let me build it, and you tell me where it goes wrong.
Let me start with the question we left unfinished last week, because I said we would return to it.
Last week, the first reply anyone made to Searle was the Systems Reply. Recall it: you are pointing at the wrong thing, Searle. The man in the room does not understand Chinese — true — but the man was never the candidate. The understanding, if it is anywhere, is in the whole system: the man, the rulebook and the baskets, all running together. You do not find understanding in one part of a person either; no single neuron gets the joke. We met that reply as a patch — a clever repair to someone else's argument. But it was never only a patch. It was the first appearance of something much larger. It was a whole theory of what a mind is, showing up in miniature. Tonight we meet the theory in full.
And I should be honest about what tonight asks, because it is harder than last week. Last week you built the side you were already inclined toward — the dismissive one. Tonight, for most listeners, the harder direction: to build the side you walked in against. If your first reaction is that it is obviously just silicon, obviously not a mind — and most first reactions are — then your task tonight is to make the opposite case, that the machine could have a mind, strongly enough that you can see exactly why serious people hold it. You are not defending your own comfort; you are arguing against it.
The rule underneath has not changed: you do not get to reject a view you have only ever met in its weak form. We applied that to the dismissive case last week; tonight we apply it to its opposite. And the idea that the machine is simply made of the wrong material has an opponent most people have never met directly. You have met the caricature — so you think your phone is conscious? That is not it. The real view is patient and precise, and it has been defended for sixty years. Tonight you will build it, and then we will find out what it costs.
So let me state the claim it rests on plainly, before we put any weight on it. The claim is this: a mind is not a substance. It is a set of functions. What makes something a mental state — a belief, a pain, a wanting of some particular thing — is not what it is made of, but the role it plays: what tends to cause it, what it in turn tends to cause, how it connects to all the other states around it. Get the whole structure of functions right and you have the mind — and it does not matter, on this view, whether those functions are carried out by neurons, or by silicon, or by something not yet built. And on this picture, the objection "it is not the right material" is no objection at all. It is a confusion about categories. Let me show you why that is the central point.
Start with pain, because it is the example the philosopher who built this reached for, and it is the clearest.
In 1967, in a paper called "The Nature of Mental States" — first published as "Psychological Predicates" — the philosopher Hilary Putnam asked a question that sounds almost too simple: what is pain? Not what causes it, not how it feels, but what it is. And the answer everyone had assumed until then was that pain is a particular event in the brain. A specific piece of neural machinery firing — the textbook phrase is C-fibres firing. Pain simply is that physical event, the way water simply is H2O. Find the neurons, and you have found the pain. Clean, scientific, settled.
Putnam argued that this cannot be right. An octopus feels pain. So does a bird, a fish, very likely some kind of insect. Their nervous systems are built nothing like ours — different cells, different structures, lineages that separated from us hundreds of millions of years ago. There is no reason to think an octopus undergoes the exact neural event you undergo when you stub your toe. Its physiology is entirely different. And yet — would anyone seriously say the octopus cannot be in pain, simply because it lacks our particular wiring? Watch it recoil, protect the injured limb, learn to avoid what hurt it. If that is not pain, the word has stopped meaning anything. So pain is not identical to any one physical event, because the same state — real pain, the actual thing — occurs in systems built quite differently. Putnam's term for this, and the whole theory hangs on it, is multiple realizability: the same mental state can be realized — implemented, made real — in many different kinds of material.
That is the decisive step. If pain is not any one physical thing, then what makes a state count as pain — across the human, the octopus, the bird, and whatever an extraterrestrial might be made of — cannot be the material. It has to be something the material is doing. And what it is doing is filling a role: pain is the state that tends to be caused by injury, that tends to make one wince, tend the wound and want it to stop, that tends to make one more careful next time. That whole cluster of causes and effects is the pain. Anything that occupies that role, whatever it is made of, is in pain. The role is what matters. The tissue is simply the local way we happen to fill it.
And once you see it for pain, you see it for everything mental. A belief is whatever state takes in evidence, connects to your other beliefs, and shapes what you say and do. A desire is whatever state, given a belief about how to get some thing, tends to make you go and get it. Define each mental state by its function — what activates it, what it activates in turn, how it interacts with the rest of the system — and you have defined the mind completely, without once mentioning what it is made of. This is functionalism. A mind is a structure of functions, a large network of states defined entirely by how they act on one another and on the inputs and the outputs. Run that whole network in a brain, and you get a human mind. The strong claim — Putnam's, which David Lewis sharpened a few years later, in 1972, into a precise philosophical account — is that if you could run the very same network in something else, you would get the very same mind. Not a copy, not a simulation, but the thing itself. Because the mind was never the gears. It was always the telling of the time.
One analogy makes the mistake plain: the carburettor. What makes something a carburettor? Not what it is made of — you can build one from steel, from aluminium, in principle from ceramic. What makes it a carburettor is what it does: it mixes fuel and air in the right proportion and feeds the mixture to the engine. Anything that does that is a carburettor. Now imagine someone who has only ever seen steel ones insisting that an aluminium one cannot really be a carburettor — that it is the wrong material, not a true carburettor, merely something that behaves exactly like one. We would think they had misunderstood the word. They have confused the function with one of the materials that can perform it. And the functionalist says this is precisely the mistake we make when we point at the warm silicon and say it cannot really think, it is not brain tissue. We have confused the mind — which is a function — with one local way of building it. The only way we have ever seen it built. Our own.
Now turn this on last week's room, because functionalism is not a theory off to one side. It is the most serious answer Searle ever had to meet, and you can see exactly where it lands.
The functionalist looks at the Chinese Room and says: Searle, you have done something clever and slightly underhand. You built a system that produces fluent Chinese — it takes in questions, holds an enormous organised store, follows rules and returns fitting answers, performing the entire role of a Chinese speaker without fault — and then you walked us inside it, sat us next to the one part that plainly understands nothing, the man, and said: there, no one is home. But the man was never the system. The man is a component — he is a single gear, a single neuron. Of course the gear does not understand Chinese, any more than the part of your brain that detects edges understands this sentence. You asked us to look for the mind in one part, found it was not there, and concluded it was nowhere. But the mind, if it is anywhere, is in the whole role being performed — and the whole role is being performed, without fault, by the system you built. You showed that the gear is not the clock. You did not touch the clock.
And this is not an evasion; it is multiple realizability applied directly. Searle's whole case was that syntax is not semantics — that manipulating symbols by their shape is the wrong kind of thing ever to amount to understanding. The functionalist replies that the wrong kind of thing is not an argument; it is the carburettor mistake again. If the system performs the entire function — if it takes in meaningful input, produces fitting output, revises and corrects, and connects each state to all the others exactly as an understander would — then on what ground, other than it is not made of brain tissue, do we deny that it understands? And it is not made of brain tissue is the one ground functionalism has spent sixty years showing we are not entitled to.
Now it is your turn. You have the tools; take the argument the rest of the way — including in the direction you dislike. Take the actual machine, the one you used this morning, and run the functionalist case on it, in your own words, and mean it. It would run something like this. This thing takes in a question, and what it does next is not a simple lookup — it holds a vast organised network of states, each shaped by how it relates to the others, and it moves through that network to produce an answer that fits; then it takes my correction and adjusts; then it connects what I have just said to what I said an hour ago. That is a role-structure. That is the function a mind performs. And the only reason I am sure no one is performing it is that the performer is made of the wrong material — and the wrong material is exactly the prejudice I have just agreed I cannot defend. So either I find some function a mind performs that this thing demonstrably cannot, or I admit that my obviously not was never about what it does. It was only ever about what it is made of.
Build the case that well, and notice what it does to the question. Last week the dismissive case stood at full height and the question felt closed — it cannot, and that is the end of it. This week the opposite case stands just as tall, and the question opens all the way in the other direction: perhaps it could; perhaps the material was never the point; perhaps we have been failing the machine on a test — be made of neurons — that we would never dream of applying to the octopus, or to an extraterrestrial, or to anything else we would readily grant a mind. And the temptation now is to settle here, to say: that decides it the other way — it is not the material, so the machine is a genuine candidate, and who is to say it is not conscious?
Do not settle yet. The case is up, and you are standing inside it — which is the only honest place from which to ask the next question: what does this case cost? Now we find out what functionalism cannot pay for.
You built the role, and it held, and the question opened toward perhaps it could. The objection comes from within functionalism's own camp — from philosophers who take the theory entirely seriously and still found a gap in it they could not close.
Consider this. Functionalism says the mind is the role — the whole structure of functions, defined by how the states act on one another and on the inputs and outputs. Then it should not matter what runs the structure — so run it on people. Take the population of a whole country — a billion people — and give each of them a small task: when you receive a call from these three, ring those two. That is all. Hand out a billion such rules, one per person, arranged so that the whole crowd, calling and answering across the country, performs exactly the role-structure your brain performs when you are in pain. Every input mapped, every state, every output, the entire functional network — running on a billion people and a telephone system instead of on neurons. Functionalism is committed, without qualification, to saying that this crowd is in pain: that there is something it is like to be that nation-sized system, a genuine ache, a real experience, present somewhere over the whole arrangement.
And — in honesty — is there? One can grant that the crowd performs the role perfectly, meets every functional condition, and still ask, in all seriousness: but is anyone in there? Is there any experience anywhere in that system? Or is it simply a billion people making telephone calls, and no one, nowhere, feeling anything at all? That question is not nothing. The philosopher Ned Block built exactly this case, in 1978, to force it, and the worry is exact: if one can run the entire functional role and still sensibly wonder whether anyone is present, then performing the function may not be the same as having a mind after all. The very thing the dismisser held onto last week — that there is a difference between displaying all the signs of an inner life and actually having one — returns. Except that now it is not Searle's point about understanding. It is a point about experience, and it strikes functionalism from within.
So consider where you now stand. You have two arguments, both built to full strength by serious people, and they point in opposite directions. Searle: performing the symbols is never understanding them — so it cannot. Putnam: the mind was never the material, only the function — so it could. And beneath both, a third difficulty: that even the whole function, perfectly performed, might leave out the one thing you were most sure mattered — whether there is anyone present to feel it. We will return to that difficulty; it is the most serious objection in the whole series, and it does not arrive until later, because it cannot be taken up until both of tonight's positions are standing — which only now they are. For tonight, leave it as the gap in functionalism that functionalism cannot close.
And now the same refusal as last week. Do not crown functionalism. The temptation is considerable — the case is so clean, the carburettor so plain, the prejudice about material so embarrassing once it is named — to conclude: fine, then it could be conscious, the material was never the point, the question is settled the other way. That relief — the sense that this side has won and one can stop — is the very same appetite we noticed last week, when Searle seemed to have an answer for everything. It is the quick verdict again. Last week it took the form of obviously not; tonight it takes the form of who is to say it is not. And it is no more earned tonight than it was last week, because functionalism has a cost it has not paid — the one the billion-person system just ran up. The honest position is the uncomfortable one: keep both the room and the role in view, each at full strength, each with its cost unpaid, and endorse neither.
What you take out of tonight is a harder position than the one you came in with — and that difficulty is the point.
You settled the question, once, with a gesture. It's just silicon. One motion toward your own head, one toward the data centre, and the line was drawn. You cannot draw it that way any longer — and, as last week, not because the first case weakened, but because you have now built its opposite to the same height. You now hold two arguments, each built to full strength by people who gave their lives to it, and they pull in opposite directions, and neither falls. Searle: performing the symbols can never be understanding them. Putnam: the mind was never the material, so the symbols, performed properly, in anything, could be a mind. Last week you could not dismiss the question as obviously not. This week you cannot reverse it and dismiss it as obviously it could either. You stand between two arguments built to their limit, and neither one is enough to settle the question by itself. Which is the only place the real work can begin.
And standing between them should make you suspicious of something — and this is the thing to carry past tonight. We have built two of the strongest cases anyone has made, and neither has earned the right to dismiss the other. The suspicion is this: that what keeps you stuck is not only a missing argument about the machine — that the difficulty was never entirely out in the world, waiting for the right theory to reach it. That some of it is something you have been bringing to the question all along, and have never yet set down where you could look at it.
Because something has been building across these three weeks, without my saying it directly. Every one of these disputes — Searle's, Putnam's, all of them — runs on a test: some requirement a thing must meet to count as having a mind — real understanding, or the right functional role, or inner experience. We have been arguing about which test is the correct one. We have never stopped to ask the stranger thing all of them rest on: why suppose that having a mind is the kind of thing a test detects at all? Where did we get the picture of a mind as an inner something, standing behind the behaviour, that the right examination could reach in and find? That picture has a history. It did not fall from the sky, and it is not the only one anyone has ever held. Later we will go back and find where it came from, because it turns out to be causing half the trouble. Not yet. I am only marking it now, so you will recognise it when it returns.
Next time the question comes closer to home. We stop surveying other people's theories, and I turn the question on you directly — not one of theirs, but the one you have been running underneath all of this without ever stating it. It sounds like the easiest thing in the world, to say what you have been assuming all along. It decides everything.
So one thing to take with you. This week, when the machine does something that genuinely lands — answers better than the person beside you, catches what you meant, says the thing that helps — you will notice one of two reactions arrive. It's just silicon, no one home. Or: honestly, who is to say — perhaps there is something there. And this once, instead of trusting whichever one arrived, look for what lies beneath it: the standard you are holding the machine against, that makes one of those reactions feel like plain sight. Not Searle's, not Putnam's. Your own. You do not have to pin it down tonight — in fact you will find you cannot quite; it slips away as you reach for it. Just notice that it is there at all. We will go after it properly next time.
Thanks for listening. I'll see you next time.