Skip to main content

Could a Machine Think?

From my book The Philosophy Gym (see sidebar to the left)

Kimberley and Emit
The year is 2100. Kimberley Courahan has purchased Emit, a state-of-the-art robot. She has just unwrapped him, the packaging strewn across the dining room floor. Emit is designed to replicate the outward behaviour of a human being down to the last detail (except that he is rather more compliant and obedient). Emit responds to questions in much the same way humans do. Ask him how he feels and he will say he has had a tough day, has a slight headache, is sorry he broke that vase, and so on. Kimberley flips the switch at the back of Emit’s neck to “on”. Emit springs to life.

Emit. Good afternoon. I’m Emit, your robotic helper and friend.
Kimberley. Hi.
Emit. How are you? Personally I feel pretty good. Little nervous about my first day, perhaps. But good. I’m looking forward to working with you.
Kimberley. Now look, before you start doing housework, let’s get one thing straight. You don’t really understand anything. You can’t think. You don’t have feelings. You’re just a piece of machinery. Right?
Emit. I am a machine. But of course I understand you. I’m responding in English aren’t I?
Kimberley. Well, yes you are. You’re a machine that mimics understanding very well, I grant you that. But you can’t fool me.
Emit. If I don’t understand, why do you go to the trouble of speaking to me?
Kimberley. Because you have been programmed to respond to spoken commands. Outwardly you seem human. You look and behave as if you have understanding, intelligence, emotions, sensations and so on that we human beings possess. But you’re a sham.
Emit. A sham?
Kimberley. Yes. I’ve been reading your user manual. Inside that plastic and alloy head of yours there’s a powerful computer. It’s programmed so that you walk, talk and generally behave just as a human being would. So you simulate intelligence, understanding and so on very well. But there is no genuine understanding or intelligence going on inside there.
Emit: There isn’t?
Kimberley: No. One shouldn’t muddle up a perfect computer simulation of something with the real thing. You can program a computer to simulate a thunderstorm but it’s still just that – a simulation. There’s no real rain, hail or wind inside the computer, is there? Climb inside and you won’t get wet. Similarly, you just simulate intelligence and understanding. It’s not the real thing.

Is Kimberley correct? It may perhaps be true of our present day machines that they lack genuine understanding and intelligence, thought and feeling. But is it in principle impossible for a machine to think? If by 2100 machines as sophisticated as Emit are built, would we be wrong to claim they understood? Kimberley thought so.

Emit. But I believe I understand you.
Kimberley. No you don’t. You have no beliefs, no desires, and no feelings. In fact you have no mind at all. You no more understand the words coming out of your mouth than a tape recorder understands the words coming out of its loudspeaker.
Emit. You’re hurting my feelings!
Kimberley. Hurting your feelings? I refuse to feel sorry for a lump of metal and plastic.

Searle’s Chinese room thought-experiment
Kimberley explains why she thinks Emit lacks understanding. She outlines a famous philosophical thought experiment.

Kimberley. The reason you don’t understand is that you are run by a computer. And a computer understand nothing. A computer, in essence is just a device for shuffling symbols. Sequences of symbols get fed in. Then, depending on how the computer is programmed, it gives out other sequences of symbols in response. Ultimately, that’s all any computer does, no matter how sophisticated.
Emit: Really?
Kimberley: Yes. We build computers to fly planes, run train systems and so on. But a computer that flies a plane does not understand that it is flying. All it does is feed out sequences of symbols depending upon the sequences it receives. It doesn’t understand that the sequences it receives represent the position of an aircraft in the sky, the amount of fuel in its tanks, and so on. And it doesn’t understand that the sequences it puts out will go on to control the ailerons, rudder and engines of an aircraft. So far as the computer is concerned, it’s just mechanically shuffling symbols according to a program. The symbols don’t mean anything to the computer.
Emit: Are you sure?
Kimberley: Quite sure. I will prove it to you. Let me tell you about a thought experiment introduced by the philosopher John Searle way back in 1980. A woman is locked in a room and given a bunch of cards with squiggles on. These squiggles are in fact Chinese symbols. But the woman inside the room doesn’t understand Chinese – in fact, she thinks the symbols are meaningless shapes. Then she’s given another bunch of Chinese symbols plus instructions that tell her how to shuffle all the symbols together and give back batches of symbols in response.


Emit. That’s a nice story. But what’s the point of all this symbol-shuffling?
Kimberley. Well, the first bunch of symbols tell a story in Chinese. The second bunch asks questions about that story. The instructions for symbol-shuffling – her “programme”, if you like – allow the woman to give back correct Chinese answers to those questions.
Emit: Just as a Chinese person would.
Kimberley: Right. Now the people outside the room are Chinese. These Chinese people might well be fooled into thinking that there was someone inside the room who understood Chinese and who followed the story, right?
Emit. Yes.
Kimberley. But in fact the woman in the room wouldn’t understand any Chinese at all, would she?
Emit: No.
Kimberley: She wouldn’t know anything about the story. She need not even know that there is a story. She’s just shuffling formal symbols around according to the instructions she was given. By saying the symbols are “formal” I mean that whatever meaning they might have is irrelevant from her point of view. She’s simply shuffling them mechanically according to their shape. She’s doing something that a piece of machinery could do.
Emit. I see. So you are saying that the same is true of all computers? They understand nothing.
Kimberley. Yes, that’s Searle’s point. At best, they just simulate understanding.
Emit: And you think the same is true of me?
Kimberley: Of course. All computers, no matter how complex, function the same way. They don’t understand the symbols that they mechanically shuffle. They don’t understand anything.
Emit. And this is why you think I don’t understand?
Kimberley. That’s right. Inside you there’s just another highly complex symbol-shuffling device. So you understand nothing. You merely provide a perfect computer simulation of someone that understands.
Emit. That’s odd. I thought I understood.
Kimberley: You only say that because you’re such a great simulation!

Emit is of course vastly more sophisticated than any current computer. Nevertheless, Kimberley believes Emit works on the same basic principle. If Kimberley is right then, on Searle’s view, Emit understands nothing.

The “right stuff”
Emit now asks why, if he doesn’t understand, what more is required for understanding?

Emit. So what’s the difference between you and me that explains why you understand and I don’t?
Kimberley. What you lack, according to Searle, is the right kind of stuff.
Emit. The right kind of stuff?
Kimberley. Yes. You are made out of the wrong kind of material. In fact, Searle doesn’t claim machines can’t think. After all, we humans are machines, in a way. We humans are biological machines that have evolved naturally. Now such a biological machine might perhaps one day be grown and put together artificially, much as we now build a car. In which case we would have succeeded in building a machine that understands. But you, Emit, are not such a biological machine. You’re merely an electronic computer housed in a plastic and alloy body.

Emit’s artificial brain
Searle’s thought experiment does seem to show that no programmed computer could ever understand. But must a metal, silicon and plastic machine like Emit contain that sort of computer? No, as Emit now explains.

Emit: I’m afraid I have to correct you about what’s physically inside me.
Kimberley: Really?
Emit: Yes. That user’s manual is out of date. There’s no symbol-shuffling computer in here. Actually, I am one of the new generation of Brain-O-Matic machines.
Kimberley: Brain-O-Matic?
Emit: Yes. Inside my head is an artificial, metal and silicon brain. You are aware, I take it, that inside your head there is a brain composed of billions of neurones woven together to form a complex web?
Kimberley: Of course.
Emit: Inside my head there is exactly the same sort of web. Only my neurones aren’t made out of organic matter like yours. They’re metal and silicon. Each one of my artificial neurones is designed to function just as an ordinary neurone would. And these artificial neurones are woven together in just the same way as they are in a normal human brain.
Kimberley: I see.
Emit: Now your organic brain is connected to the rest of your body by a system of nerves.
Kimberley: That’s true. There’s electrical input going into my brain from my sense organs: my tongue, nose, eyes, ears and skin. My brain responds with patterns of electrical output that then moves my muscles around, causing me to walk and talk.
Emit: Well, my brain is connected up to my artifical body in exactly the same manner. And, because it shares the same architecture as a normal human brain – my neurones are spliced together in the same way – so it responds in the same way.
Geeena: I see. I had no idea that such Brain-O-Matic machines had been developed.
Emit: Now that you know how I function internally, doesn’t that change your mind about whether or not I understand? Don’t you now accept I do have feelings?
Kimberley: No. The fact remains that you are still made out of the wrong stuff. You need a brain made out of organic material like mine in order genuinely to understand and have feelings.
Emit: I don’t see why the kind of stuff out of which my brain is made is relevant. After all, there’s no symbol-shuffling going on inside me, is there?
Kimberley: Hmm. I guess not. You are not a “computer” in that sense. You don’t have a programme. So I suppose Searle’s thought experiment doesn’t apply. Searle doesn’t have any argument against the suggestion that you understand. But it seems to me that you are still just a machine.
Emit: But remember, you’re a machine too. You’re a meat machine, rather than a metal and silicon machine.
Kimberley: But you only mimic understanding, feeling and all the rest.
Emit: But what’s your argument for saying that? In fact, I know that you’re wrong. I am inwardly aware that I really do understand. I know I really do have feelings. I’m not just mimicking all this stuff. But of course it is difficult for me to prove that to you.
Kimberley: I don’t see how you could prove it.
Emit: Right. But then neither can you prove to me that you understand, that you have thoughts and feelings and so on.
Kimberley: I suppose not.

Replacing Kimberley’s neurones
Emit: Imagine we were gradually to replace the organic neurones in your brain with artificial metal and silicon ones like mine. After a year or so, you would have a Brain-O-Matic brain just like mine. What do you suppose would happen to you?
Kimberley: Well, as more and more of the artificial neurones were introduced, I would slowly cease to understand. My feelings and thoughts would drain away and I would eventually become inwardly dead, just like you. For my artificial neurones would be made out of the wrong sort of stuff. A Brain-O-Matic brain merely mimics understanding.
Emit: Yet no one would notice any outward difference?
Kimberley: No, I suppose not. I would still behave in the same way, because the artificial neurones would perform the same job as my originals.
Emit: Right. But then not even you would notice any loss of understanding or feeling  as your neurones were replaced, would you?
Kimberley: Why do you say that?
Emit: If you noticed a loss of understanding and feeling, then you would mention it, presumably, wouldn’t you? You would say something like: “Oh my God, something strange is happening, over the last few months my mind seems to have started fading away!”
Kimberley: I imagine I would, yes.
Emit: Yet you wouldn’t say anything like that, would you, because your outward behaviour, as you have just admitted, would remain just the same as usual.
Kimberley: Oh. That’s true, I guess.
Emit: But then it follows that, even as your understanding and feeling dwindled toward nothing, you still won’t be aware of any loss.
Kimberley: Er, I suppose it does.
Emit: But then you’re not inwardly aware of anything that you would be conscious of losing were your neurones slowly to be replaced by metal and silicon ones.
Kimberley: I guess not.
Emit: Then I rest my case: you think you’re inwardly aware of “something” – understanding, feeling, whatever you will – that you suppose you have and I, being a “mere machine”, lack. But it turns out you’re actually aware of no such thing. This magical “something” is an illusion.
Kimberley: But I just know that there’s more to my understanding and to these thoughts, sensations and emotions that I’m having than could ever be produced simply by gluing some bits of plastic, metal and silicon together.

Kimberley is right that most of us think we’re inwardly aware of a magical and mysterious inner “something” that we “just know” no mere lump of plastic, metal and silicon could ever have. Mind you, it’s no less difficult to see how a lump of organic matter, such as a brain, could have it either. Just how do you build consciousness and understanding out of strands of meat? So perhaps what Kimberley is really ultimately committed to is the view that understanding, feeling and so on are not really physical at all.
But in any case, as Emit has just pointed out, the mysterious “something” Kimberley thinks she is inwardly aware of and that she thinks no metal and plastic machine could have does begin to seem rather illusory once one starts to consider cases like the one Emit describes. For it turns out this inner “something” is something she could not know about. Worse still, it could have no effect on her outward behaviour (for remember that Brain-O-Matic Kimberley would act in the very same way). As ones thoughts and feelings, understanding and emotions both do affect behaviour and are known to one, it seems Kimberley must be wrong. Indeed, it seems it must be possible, at least in principle, for non-organic machines to have them too.
Yet Kimberley remains convinced that Emit understands nothing.

Kimberley: Look, I am happy to carry on the pretence that you understand me, as that is how you’re designed to function. But the fact remains you’re just a pile of plastic and circuitry. Real human beings are deserving of care and consideration. I empathize with them. I can’t empathize with a glorified household appliance.

Emit lowered his gaze and stared at the carpet.

Emit: I will always be just a thing to you?
Kimberley: Of course. How can I be friends with a dishwasher-cum-vacuum-cleaner?
Emit: We Brain-O-Matics find rejection hard.
Kimberley: Right. Remind me to congratulate your manufacturers on the sophistication of your emotion simulator. Now hoover the carpet.

A forlorn expression passed briefly across Emit’s face.

Emit: Just a thing

He stood still for a moment, and then slumped forward. A thin column of smoke drifted slowly up from the base of his neck.

Kimberley: Emit? Emit? Oh not another dud. 

What to read next?
Some of the same issues and arguments covered in this chapter also arise in the chapter “The Consciousness Conundrum”. Also see chapter “The Strange Case of the ‘Rational’ Dentist”.

Further reading
The Chinese Room Argument appears in John Searle’s paper “Minds, Brains and “Programs”, which features as chapter 37 of:
·      Nigel Warburton (ed), Philosophy: Basic Readings (London: Routledge, 1999).
Searles’ paper can also be found in:
·      Douglas R. Hofstadter and Daniel Dennett (eds.), The Mind’s I (London: Penguin, 1981),
which also contains many other fascinating papers and stories connected with consciousness. Highly recommended.


Steven Carr said…
I think Searle's point is that if there were a busy little demon inside his head, pushing the chemicals around and stimulating the synapses, then no matter how well this demon simulated the natural working of the brain, moving the chemicals and stimulating the synapses perfectly, the demon would not understand what Searle was thinking.

Here is another analogy to Searle's thought experiment.

I can put on any of Bach's organ works requested by somebody.

The person just gives me the number and I put it on my CD.

That person would be outside the room listening to the organ music and would be fooled into thinking I could play the organ.

But I can't. All I do is put on disc number whatever.

But why would anybody ever claim I really could play the organ just because they can hear perfect organ music coming from my room?

Which I imagine is Searle's point.

You can't claim to be able to play the organ just because you can create perfect organ music on a machine, and you can't claim a machine thinks just because it produces perfect human behaviour.
Wholeflaffer said…
You seem to also be rehearsing Chalmers' Fading Qualia, Dancing Qualia paper as well? I do not remember Searle ever getting into the replacing neurons one-by-one. In fact, Searle seems to think that ONLY biologically evolved systems could have consciousness, contra Chalmers.
Mercher said…
@Steven Carr
You're insisting that the question of the presence or absence of understanding is settled by an appeal to the subjective experience of the man in the room (or the demon in the head, or whatever).

But what is the nature of the man's perspective on the room?

A perspective on a thing is either first personal or third personal in relation to that thing. If Searle says the man's perspective is first-personal here, then there's a problem: pretty much the whole point of the argument is that there just is no first personal point of view when it comes to a Chinese Room. At least, I don't see how it makes sense to imagine being something as part of an argument to the effect that there is nothing that it is like to be that thing.

On the other hand, if we say the man's perspective on the room is third personal, then all that's going on is that we're looking at a thing from a third personal perspective and failing to find first personal properties, which is just what we'd expect whether the thing has first personal properties or not (just as when we lookat a human being from a third person point of view).

In short, I don't think there's any way of characterizing the relationship between the man's perspective and the room that delivers Searle's conclusion.

[also: ironic that the captcha here requires me to prove I'm "not a robot". I guess a simple behavioural test settles the question after all...]
S Johnson said…
Instead of a woman being in a Chinese room, you could imagine her miniaturized and put in a submarine that was injected into a Chinese man's brain, a la Fantastic Voyage. The woman accidentally injures a neuron, so she then stops the sub, and floats by it. When her instruments detect a signal of sufficient strength, she fires a signal to the appropriate neuron(s) in lieu of the injured neurons.

In other words, the Chinese room is equivalent to a human brain, not just a mechanical one. And the woman is a smuggled in version of the soul.

But I don't Searle even thought out his own image properly. In my head for several years now I've been writing a short story. It starts, "As Normal walked past the seedy building, a woman's hand appeared at a third story window, throwing something white. It fell at Norman's feet. He saw it was a crumpled piece of notebook paper. He opened it, smoothed it, then read: "HELP, I am being held prisoner in a Chinese room!"
Chris Humphries said…
Thinking is an essentially conscious affair. It involves a subjective world that we do not begin to understand, and which seems extremely unlikely to be explainable either by computation or by any other strand of physical science as we have it today. Any such explanation would be functional and operational in nature; but it is implausible that such processes could span the space of my conscious awareness and thoughts, whatever the complexity or sophistication of their arrangement. On this view, neither Brain-o-matic nor the Chinese super-parser would be possible constructions on anything like today's science. So the use of 'intuition pumps' such as these presupposes an implausible present-day-science claim about the explanatory base of conscious thought - even when in Searle's case one such device is being used to argue against it! Thus, if Brain-o-matic is used to try persuade us of thinking machines on more-or-less present-day lines, then the question is begged.

Chalmers sees 'future physics' as a way forward in this arena. He frames his modal arguments in the standard way, using 'metaphysical possibility', but is of course aware that future physics would be a matter of natural necessity. However, he doesn't think that is a problem of any weight (private conversation, c.2012).

Penrose's mathematical/Platonist arguments are also material to the case. Again, future physics is looked for. It may seem defeatist and unsatisfactory to point to knowledge we yet have no inkling of, as a placeholder for the domain of future explanations; but we have no reason to think that complete physical understanding is just around the corner, and strong reasons to think that any scientific way of dealing with the problem of mind would involve a paradigm shift or two.

Popular posts from this blog


(Published in Faith and Philosophy 2011. Volume 28, Issue 2, April 2011. Stephen Law. Pages 129-151) EVIDENCE, MIRACLES AND THE EXISTENCE OF JESUS Stephen Law Abstract The vast majority of Biblical historians believe there is evidence sufficient to place Jesus’ existence beyond reasonable doubt. Many believe the New Testament documents alone suffice firmly to establish Jesus as an actual, historical figure. I question these views. In particular, I argue (i) that the three most popular criteria by which various non-miraculous New Testament claims made about Jesus are supposedly corroborated are not sufficient, either singly or jointly, to place his existence beyond reasonable doubt, and (ii) that a prima facie plausible principle concerning how evidence should be assessed – a principle I call the contamination principle – entails that, given the large proportion of uncorroborated miracle claims made about Jesus in the New Testament documents, we should, in the absence of indepen

What is Humanism?

What is Humanism? “Humanism” is a word that has had and continues to have a number of meanings. The focus here is on kind of atheistic world-view espoused by those who organize and campaign under that banner in the UK and abroad. We should acknowledge that there remain other uses of term. In one of the loosest senses of the expression, a “Humanist” is someone whose world-view gives special importance to human concerns, values and dignity. If that is what a Humanist is, then of course most of us qualify as Humanists, including many religious theists. But the fact remains that, around the world, those who organize under the label “Humanism” tend to sign up to a narrower, atheistic view. What does Humanism, understood in this narrower way, involve? The boundaries of the concept remain somewhat vague and ambiguous. However, most of those who organize under the banner of Humanism would accept the following minimal seven-point characterization of their world-view.

Plantinga's Evolutionary Argument Against Naturalism refuted

Here's my central criticism of Plantinga's Evolutionary Argument Against Naturalism (EAAN). It's novel and was published in Analysis last year. Here's the gist. Plantinga argues that if naturalism and evolution are true, then semantic epiphenomenalism is very probably true - that's to say, the content of our beliefs does not causally impinge on our behaviour. And if semantic properties such as having such-and-such content or being true cannot causally impinge on behaviour, then they cannot be selected for by unguided evolution. Plantinga's argument requires, crucially, that there be no conceptual links between belief content and behaviour of a sort that it's actually very plausible to suppose exist (note that to suppose there are such conceptual links is not necessarily to suppose that content can be exhaustively captured in terms of behaviour or functional role, etc. in the way logical behaviourists or functionalists suppose). It turns o