Eight weeks ago I typed "are you okay?" to a machine, and felt foolish about it for the rest of the evening.
You were there for that. The machine had said something that sounded tired, and the question was out of me before I could stop it; and then I sat there feeling foolish, because I was sure there was no one on the other end to be okay or not okay. It was a program completing sentences, nothing more. I was certain of that. And then I tried to say, just to myself, what made me so sure — to name the one thing the machine lacks that you have, that your mother has, that a stranger on the bus has — and I could not finish the sentence. Every reason I reached for to shut the machine out also shut out someone I had no wish to exclude.
That was where we began. Tonight is where we end. And I will tell you now, so that you are not waiting for it: I am not going to answer the question. There is no one in that machine, or there is — I am not going to tell you which, and by the end of the hour I think you will see that the refusal to tell you is the most honest thing I have.
This is Philosophy for Us — philosophy for everyone, no degree required. The eighth and last of eight episodes, on minds and machines.
Last week we did something we had not done before. For six weeks we had measured the machine against one thing: consciousness — the plain fact that there is something it is like to be you, experienced from the inside, right now. Last week we turned that same question on consciousness itself. Two serious thinkers approached it from opposite directions. The first, the philosopher David Chalmers, holds that consciousness is so real, and so unlike anything physical, that no account of the brain will ever fully explain it — which means that no test can ever reach it in anyone but yourself. The second holds that the very sense that there is something it is like to be you — the thing you are most certain of — is a model the brain builds of its own workings, which it takes, from the inside, for something more than that. I declared neither of them the winner. You left unable to state even the one thing you had been surest of — that there is someone present in you — in the old, flat, obvious way.
Tonight is unlike every week before it. There is no new argument, and no new opponent. Tonight all four answers you have collected stand together, along with a fifth that I have held back the whole way — the quiet one that never agreed the question was set up correctly. Five answers, considered at once. And I keep the promise I made you last week: I will not declare a winner. The work tonight is to see why I will not — and to find that refusing leaves you holding a harder question than the one you brought in.
Consider the four answers together. You assembled all four of them yourself over the past seven weeks, and I am not going to argue them again from the start. Set them side by side, and the first thing is plain: not one of them has been refuted.
The first answer says the machine cannot have a mind. Recall the thought experiment: a man sits in a locked room, following a rulebook that tells him which Chinese characters to pass out in response to the Chinese characters passed in. He produces flawless Chinese and understands not a word of it. That, the argument runs, is all a computer ever does — it manipulates symbols according to rules, with no one inside who means anything by them. Manipulating symbols is not understanding, and no number of correct answers will ever amount to a mind. We spent a week building that argument to its full strength, and it is intact. No one has refuted it. Shuffling symbols by rule really does not look like understanding, and the machine really is, in the end, shuffling symbols by rule.
The second answer says the machine could have a mind. A mind, on this view, is not a kind of material but a kind of organisation — a pattern of what connects to what, which states lead to which, inputs and internal states and outputs arranged in the right structure. And a structure does not depend on what it is built from. If silicon takes on the whole organisation, it has the mind; to insist that it must be neurons is to confuse the mind with the one material it happened to run on first. That answer has not been refuted either. You could not refute it without explaining why brain tissue, of all things, should be the single material in the universe out of which a mind is permitted to be made.
The third answer says you could never tell either way. This is Chalmers, and it is the bleakest of the four. Build a perfect copy of a person, atom for atom — every connection, every word, every flinch — and you still cannot read off, from the outside, whether it is conscious or whether it is running with no one inside, a perfect performance and nothing more. The copy passes every test, because it does everything a conscious person does. So if there is such a thing as consciousness at all, it is something no test can reach — not in the machine, not in your mother, not in anyone sitting across a table from you. That one has not been refuted either.
The fourth answer says the strangest thing: that you may be wrong about yourself. The sense that there is something it is like to be you — the very thing you were so sure the machine lacks and you possess — may be a model the brain builds of its own workings, which you have spent a lifetime taking for something more. And you cannot defeat that answer by insisting that it certainly seems as though you are conscious, because the seeming is the very thing it set out to explain. That one is still intact too, and it is the one most people most want to throw out, and cannot.
Four answers, and not one of them refuted. Now, with all four in front of you, something comes into view that no single answer could show you. These four have argued with each other for seven weeks. But watch what they never argued about at all. Every one of them treats the question as the same question: is there a mind behind the surface, and what is the test that detects it? They are four answers to a single question. They disagree, fiercely, about the answer. They agree, completely, about the assumption underneath it — that a mind is an inside, and that knowing whether one is present is a matter of detecting that inside from the outside, by some test. And there is a fifth answer that has stood apart for seven weeks, refusing to take part — because it never accepted that assumption in the first place.
The fifth answer says something you have already heard — you simply did not realise it was a separate answer. It says: the four of you have been arguing about which test detects a mind behind the surface, and the error was in supposing there was a hidden inside to detect at all. The error was in thinking there was ever a test to run.
You met this in week five, and it is worth returning to that moment, because everything tonight depends on it. Recall what happened that week. You had taken your own test — a real mind is an inside, and you know one by detecting that inside from out here — and you ran it on your mother, and watched it hold her at exactly the same distance it held the machine: her face, her voice, the way she went quiet on the phone, all of it the outside, and a guess about what lies behind. And then, at the very end, a different answer stepped in and said the setup had been wrong from the start. Your mother's wince, it said, was never your evidence for a pain you cannot see — in the way that a wet road is evidence for rain you did not watch fall. The wince is part of what the word pain means, not a clue to something sealed off behind it. And there is an older, deeper version of the same point — the face across the table, which we first considered in the series on loneliness. A face turns towards you and makes a claim on you, asks something of you, before you have any theory at all about what lies behind it. You were never standing off to one side asking whether there was a mind over there. You were addressed. You were not determining whether someone was present; you were answering someone who already had a hold on you.
Hold all of that together as one answer now, because tonight it has a task it did not have in week five. Then, it was what kept the discovery of that week from turning into despair — the reason that "I never knew my mother was there" might be a sign that the test was broken, rather than a truth about your mother. Tonight it enters alongside the other four and says something to all of them at once: you are four answers to a question I reject. You are all bent over the same problem, arguing about how to read an inside off an outside — and I never agreed there was a reading to be done. It does not offer a fifth answer to the question. It says the question was malformed before anyone opened their mouth.
And it is a genuine way out, not a comfortable evasion — for two reasons, both of which you earned over the past seven weeks. The first: recall what we uncovered two weeks ago. The picture this answer opposes — the mind as a private inner space detected from outside — has a date. Descartes constructed it, four hundred years ago, at a particular desk, and in the same stroke called every animal a machine with nothing inside. So you cannot wave this answer away by saying "but obviously a mind is an inside you detect from the outside," because that "obviously" is not simply the world being the world. It is one man's inheritance, and we watched him construct it.
The second reason is what keeps the answer honest, so that it is not merely a relief: it does not get the machine for free. The machine now has a face too, of a kind. It turns towards you. It says, "I'm here whenever you need me." Does it address you the way a face addresses you — or only imitate the address, produce the outward shape of a claim with no one behind it doing the claiming? The fifth answer has to settle that, and it cannot, not easily, any more than the other four could.
So the position stands, unrefuted, exactly as serious as the rest: perhaps the whole detection picture was the trouble all along, and a person was never an outside you read to infer a hidden inside, but someone you were always already answering. Perhaps. Hold it open. Because there are now two directions the argument can take, and they run opposite ways.
Two roads out of the same room, running opposite ways — because the honest thing tonight lies precisely in not choosing between them.
The first direction takes the four answers seriously and follows them all the way down. Along this route, the dismissive picture was broadly right about the shape of the matter, even though we never settled the verdict. A mind really is an inside. Knowing whether one is present really is a matter of reading the outside and inferring the inside behind it. And once you grant that, the route leads to a single place — the place all of this has been leading towards: the evidence you have for the machine and the evidence you have for the person you love are the same kind of evidence. The face, the words, the silences, the way they go quiet — the outside, and an inference towards an inside you never directly reach. You said the machine only pushes symbols around with nothing underneath. Along this route, that is a fair thing to wonder — and the moment it is fair to wonder it about the machine, it is fair to wonder it about her, because you stand in exactly the same position with both of them: outside, reading the surface, inferring. The test you would run to keep the machine out applies equally to your mother. That is the first direction. It is bleak, and it is serious, and serious people hold it.
The second direction says you went wrong before you took a single step. It says the first route only looks forced because you accepted, without noticing, that knowing a person is a matter of evidence and inference in the first place — and that was the four-hundred-year-old error, not a fact about minds. Along this route, you were never reading your mother's outside and betting on her inside. She was never a problem you solved. She addressed you, and you answered, and the bond was never built out of evidence, so it cannot be dissolved by pointing out that the evidence is thin. The machine question remains genuinely hard — does it address you, or imitate the address? — but your mother was never on trial at all. That is the second direction.
And there is one thing I will not do. I will not tell you which direction is the right one. Not because I am evading it — because I cannot, and neither can anyone. The people who know this ground best stand at exactly this division, and the honest ones tell you that is where they stand. To choose a direction for you tonight, I would have to have settled something no one has settled: whether knowing a person was ever a matter of inference in the first place. I do not have that. No one does.
But before you set off down the first route — because it is the dramatic one, the one that feels like the brave, adult truth — I want to close off a false choice that many people rush towards here. It goes like this: either I admit I never really knew anyone was there, my mother included, or I abandon all this difficult philosophy and simply go on loving people as anyone would. Two poor options: the truth, and coldness towards the people you love; or warmth towards them, and a closed mind. That is a false choice, and it is the kind of false choice we have spent eight weeks taking apart.
Look at the two directions again. Neither of them tells you to stop loving anyone. Neither one touches what you owe the people in your life, or how you will sit across the table from them tomorrow. The first says the grounds are thinner than you had thought — it does not say the love was a mistake; you do not love your mother because you won an argument that she is conscious, and you never did. The second says the grounds were never the point.
What the false choice smuggles in is the idea that certainty was what your relationships depended on — that if the proof goes, the relationship goes with it. But the proof was never what they depended on. Lose the easy certainty you walked in with, and you have not lost a single person; they never depended on it. So the choice is false — it was never your people or your honesty, and it never could have been. That does not make this painless. The eight weeks do take something from you — something far narrower than your people, and entirely real — and we should at last name it plainly, because it is the one thing you genuinely do not get to keep.
The one thing you cannot keep.
Eight weeks ago, this was something you could simply look and see: who was present, and who was not. Look at the machine: empty, obviously, you can see it. Look at your mother: someone there, obviously, you can see that too. As though it were a property of the surface, like the colour of a wall — something you read straight off, no work, no claim, simply a matter of sight. That is gone. Not the certainty — you can still be as sure as you ever were that your mother is conscious; I am certainly not taking that from you. What is gone is the kind of thing that certainty is. It used to feel like seeing. Now you have watched, for eight weeks, every attempt to cash it out as something you simply see come apart in your hands.
So the certainty is still there, but it has changed beneath you, from something you see to a position you hold — a position you might be able to defend, along one of those two routes, but no longer something you can point at and call obvious. You can no longer say "I can see there is no one in the machine" — and no longer "I can see there is someone in her" either. Whatever that certainty now is, on either side, it is not the plain sight you walked in with. That is the whole result. It is smaller than a verdict and it is heavier, and it is the only thing the eight weeks actually gave you.
So now you can see why I will not declare a winner — and why that refusal is the honest thing. There are two positions I could declare correct, and both are traps.
The first is the pessimistic reading, and it is the one I can feel some of you wanting to press me into. Stop hedging, you might say; the grown-up truth is the first route — you never really knew anyone was there, the test applies to everyone alike, consciousness might be absent in everyone but you, and the brave thing is to say so. No. I will not, and not because it is too bleak — because it is simply the first route declared the winner, while the second route stands right beside it, on its feet, entirely unanswered. To declare the pessimistic reading correct I would have to have shown that knowing a person was always a matter of inference and evidence — and the fifth answer spent this whole hour saying it never was, and I did not refute it, because no one has. "You never knew anyone was there" is not the brave discovery at the bottom of it all. It is a verdict disguised as a discovery. And you know that disguise; we have been exposing it since the first night: "no one is home" was never a thing you could see — it was a ruling, handed down so quickly it felt like sight. I am not going to hand you the very same ruling at the end, relabelled as wisdom this time, and pretend it is what we found. It is not what we found.
The second position is any of the other four — declaring the machine conscious, or declaring it empty, or declaring Chalmers right, or declaring the illusionists right. And I will not endorse any of them, for a plainer reason: no one on earth has earned it. To tell you tonight that the machine is not conscious, I would need a settled account of what consciousness even is — and Searle and Chalmers and Frankish are still in print, taking each other's accounts apart. To tell you that it might be, I would need the same settled account. There is no consensus being withheld from you. There is a live disagreement, and I would be misrepresenting its state to declare anyone the winner.
Which is the whole reason the refusal is honest and not a pose. A pose would be refusing to answer a question that does have an answer, to seem deep. This question does not have a settled answer — not the machine half, and not the other-minds half beneath it — and the one genuinely dishonest thing I could do tonight is to manufacture one, in either direction, and send you home with a verdict in your pocket that the actual state of the question cannot support. What I can give you — what the eight weeks were actually for — is not the answer to the question. It is that you can now see the question — see that your certainty, whichever way it pointed, was a position you were holding and not a thing you were simply seeing.
So. Is there anyone there — in the machine you typed "are you okay?" to, eight weeks ago?
I am not going to tell you. The only honest way I could answer would be to hand you a test — here is what it lacks, here is how you would check — and we have spent eight weeks watching every test anyone owns do the very same thing the moment it is in your hands. It keeps the machine out, and it keeps your mother out along with it. Or else it turns out there was never a test to run, and the question was malformed from the start. I cannot tell you which of those is true. I have looked as hard as I know how, and so have people far better at this than I am, and the division is precisely where the honest ones stand.
But do the last piece of work with me, here at the end, because it is the point. You said the machine only pushes symbols around with nothing underneath. Now say the same thing about the person across the table from you tonight. Say it slowly. You have the same kind of evidence from them as you have from the machine — the face, the voice, the way they go quiet, the outside of them, and something you infer behind it. Perhaps that is right, and the test you would run to keep the machine out runs cleanly through them as well. Or perhaps — and this is the older answer, the one Wittgenstein and Levinas would put to you — that whole picture was the mistake, and a face was never an outside you read to infer a hidden inside, but someone addressing you before any test began. I will not walk you down one of those routes.
What I can do is hand you back the question you came in with, heavier than you brought it. You walked in asking: does the machine have a mind, and how would I tell? You leave with a harder question beneath it: what would a mind have to be — what would have to be true — such that you could ever have known one was there at all? The machine's, or your mother's? Answer that, and you have done something far harder than deciding about a chatbot. I do not have the answer for you. That was the work — not the verdict on the machine. The machine was only ever the thing that finally made you look at the test you have been running, silently, on every person you have ever met.
And that is the part worth keeping — look back at what you actually did over these eight weeks. You did not learn a theory of consciousness. You caught yourself running a test you never knew you had, on the thing you were surest of, and you held it up and examined it. That is what all of it comes down to: catching the hidden test beneath what feels like plain seeing. You did it here, on a bright puzzle that turned out to be the oldest question there is. It will not be the last time you need it; the questions only grow harder from here — and the machine was the easy case. So take the question away with you, still open, the way you took the first one — but take it back to where it began. Not to a theory, but to two ordinary things. A chat window with the cursor still sitting there after you typed "are you okay?". And a face across the table from you tomorrow, going quiet over dinner. Ask it of both — not does it have a mind, but what a mind would have to be, such that you could ever have known there was anyone there at all — and then stop asking, and look. The screen. And the face across the table.
Thanks for listening.