Searle: Philosophy of Mind, lecture 4

all right now we're locked in the first half of the course which is where I'm going to teach you mainstream philosophy of mind I'm going to teach you what other philosophers think is the important stuff in the philosophy of mind and I think a lot of it is important but I think the really important stuff comes in the second half of the course when I try to tell you how the mind actually works and how it relates to the brain in some detail but that comes later now we're still locked in the debates that follow essentially the rejection of cartesianism if you reject cartesian dualism as most people do i then what is the what is a rational position to adopt now the contemporary orthodoxy is some form of materialism a materialism is sort of the religion of our time in in the in the sciences of the mind that is to say in the philosophy of mind in a cognitive science in psychology and the other social sciences some type of materialism is just generally accepted and the idea is somehow to make it consistent with what we otherwise know and that the last lecture I went through the different forms of materialism that were influential in the last half of the 20th century and many of which continue to be influential today I we went through behaviorism both analytic behaviorism and methodological behaviorism I think behaviorism is now pretty much discredited there are self-confessed behavior still running around I guess Dan Dennett counts as a behaviorist in fact he says he is so that's a fair enough characterization but there aren't an awful lot of behaviorist left and I think even in psychology there very few professors who would get up in front of a class and say psychology is a science of human behavior they used to say that for a long time because they were embarrassed to say science is the psychology is a science of the mind which what I think it is because that sounded my god you're back in bed with Descartes you know the mind as if it were some kind of a separate entity well I think psychology is that science of the mind and cognitive science is a type of psychology so I'm going to be talking about the foundations of cognitive science when we get to it ok so the idea then was says some sort of materialist conception of the mind it isn't the behavior but it's the brain which is the basis of the mind the essence of the mind is the brain and then the idea is to get some form of what's called identity theory that the mind just is identical with the brain that mental states just are identical with brain States and favorite examples were always embarrassingly bad neurobiology the favorite example was pain is see fiber as stimulation I'm embarrassed to tell you that because it's it's worse than bad philosophy it's bad brain science see fibers are type of axon I and there they are different from il-2 Delta a fibers which carry a different type of signal one of them is myelinated and the others not myelinated and the myelinated axons carry the signal faster than the unmyelinated ones and one of them is more common with aching burning sensations and the other go with the other type of axon tends to go with this sharp pricking sensation where you can localize where the pain is exactly but I its unintelligent to identify pain with a type of axon which is an anatomical feature of neurons however the basic impulse remains even after the bad neural physiology is removed and the basic impulse is to say well when you're in pain you're at a certain type of brain state and that brain state is necessary and sufficient for pain now there's a mean there's an immediate example and that is it looks like yeah but if you're going to say that that is an identity that you can discover the way you discover that water is identical with h2o then there's got to be two different sets of properties on each side of the identity statement the water properties in the h2o properties or the evening star properties in the morning star properties whatever is your identity statement so if you're going to say pain this feeling is identical with this state of the brain I then it looks like you got pain properties and brain properties and that looks like property dualism now when I used to argue with these guys they always would say well we're going to get rid of those mental properties it's going to turn out that they are really something else and I mentioned some of the strategies they had for doing that the idea was well we can talk about them in a topic neutral vocabulary we don't have to use a Cartesian vocabulary that seems to me beside the point they're still there I even if you don't mention them I mean you can't yeah I get I away from them by refusing to talk about it you know you don't deny the existence of airplanes just by saying what we're not going to mention airplanes we're going to talk about the properties owned by United Airlines and American Airlines and so on airplanes are still there they're the irreducible mental phenomena are still there so the aim is to get a kind of reductionism is to show well really how the mental states can be reduced to something else so the difficulties of type type identity and I won't repeat all those difficulties again but one of the most powerful difficulties was type type identity looks like neuronal chauvinism you don't want to say in order to have a pain a system has to have a certain type of axons because who knows what kind of pains exist among animals that don't have that type of axon I have a friend who does research on termites and termites have a much simpler nervous system they've only got about a hundred thousand neurons so they're easier to study now I figure I lose a hundred thousand neurons on a big weekend so I don't know what kind of mental life the damn termites have but no one wants to say well the termite can't suffer a pain because it hasn't got my type of axon so we got to avoid neuronal chauvinism and that means we've got to have token identity now I want everybody to understand this type type identity says every mental state of a certain type is identical with a neurobiological state of a certain type they'd like to put this in a sloppy way by saying mental state types are identical with brain state types that's wrong it's not the right way to put it rather it's every token of a certain type is identical with a token of a certain a mental state type so we uh we get type identity identifies the types of mental with types of the neurobiological but now the token identity is a weaker thesis what it says is look for every mental state there's some damn physical token or other that it's identical with but we don't say that it always has to be the same it has to be a token of the same type of physical states because we don't want to exclude the possibility that you can build machines that are conscious we don't want to exclude the possibility that alligators are conscious but have a different kind of neurobiology from the kind that we have so token identity looks like it's the way to go and I think most identity theorists today are token identity theorist I don't know anybody well I shouldn't say that because then there's always somebody who will point out to me that there are some so there must be some running around the woods but the basic form of the identity theory today is token identity every mental state you have is identical with some damn token or other of a neurobiological state in your brain so if I believe Denver's the capital of Colorado and you also believe that then there'll be something going on in my brain which is identical with that belief and something going on in your brain but the neurobiology of what you have needn't be the same as the neurobiology of what I have okay everybody's got that that seems reasonable I'm sympathetic to these guys but then you got a question what fact makes them both the same mental state now notice that's a serious question for them because we both have exactly the same belief but there's nothing there to the belief except the neurobiology but the neurobiology is different so what makes it the same see the Cartesian says what's the same conscious thought I'm busy thinking denver's the capital of Colorado and you're having the same conscious thought I and that and the type identity theory says well there's the same neurobiological structure but the token identity theorist can't say that because it was that token identity theorist wants to get rid of irreducible mental phenomena and and wants to deny that there is a type identity between the mental states and the physical states so what is the token identity theorists going to say and answer the question what fact about different physical tokens makes them tokens of exactly the same type of mental state and that answer has to be given without mentioning anything mental now comes a major breakthrough and this is what I think most philosophers believe today and it's called functionalism and functionalism says well what makes them tokens of the same type is they perform the same function and what is a function a function is a set of causal relations and I showed you briefly how they did that last time you say what causes the belief let's say the belief that it's raining well the perception that P that it's raining causes I wrote an arrow last time but let me write out the word causes the belief that P and then you also want to say the belief that P together with a desire that queue that say the desire to stay dry causes an action causes action a okay now then we can say anything that stands in those causal relations is the belief that it's raining so you knock out the reference to a specific belief and you just put in a variable there is some X such that X sorry I had the wrong perception here the perception that P I you just say there's some X such that the perception of the P causes X and the X together with a desire that Q causes action a and of course this is just a sketch we'd have to have maybe thousands of such causal relationships answering the question what do you do under certain counterfactual situations but the idea is anything that stands in the right causal relations it's caused by the perception that it's raining and together with a desire to stay dry it causes you to carry an umbrella anything is the belief that it's raining it can be anything at all a provided it stands in the right causal relations so one of the objections to behaviorism namely it can't explain the fact that the mental states function causally that's removed immediately mental states are defined by their causal relations and if you say yeah but you still got the desire that Q we get rid of that - that's going to be analyzed in terms of its causal relations that's just a why so there's some Y and there's some X such that the perception of P causes X and X together with y i causes a you say there's some X and some Y such that the perception that P causes X and X together with Y causes action a and so on you analyze the desires the same way you analyze the beliefs and if you're really going to get hard-nosed about it you do the same with the perception the perception is itself a mental state it has stands it's caused by certain types of neuronal stimulations of your of your perceptual system the photons strike the retina they strike the photoreceptor cells so the idea is the mind consists in a set of causal relations and those causal relations can be specified in a way that's neutral about the physical sub structure now that's a great idea because that allows for the next great revolution their next great breakthrough we want to know what these how these causal mechanisms work how does it actually work in real life and the answer is they are computational states of a computer system what is a belief really it's a certain type of computational state in your belief program and when the belief program is implemented in your brain but it could be implemented in any physical system any physical system that is rich enough and stable enough to carry the hardware to carry the software that is to carry the program for having beliefs and acting on beliefs any physical system can have a mind and that leads to the idea well really what are we really Descartes thought were really Cartesian souls the behaviorist thought were really systems of behavior I the type type identity thought were really neurobiological structures and the next great breakthrough was to say we're really computers or rather we are computer programs that happen to be implemented in this icky wetware in the brain now I believe means there are guys out there who still believe this as a guy named Ray Kurzweil who says the next great step in human evolution is we're going to get out of this messy crap we carry around in our skulls and get into some decent hardware we're going to be we're going to keep week we're going to be downloaded or maybe it's uploaded I never know which direction we're going to be uploaded onto a decent piece of hardware so we can get out of this junky thing we carry around is a key thing in the skull and we're now going to be he says immortal because we will be in decent hardware we'll be on hard disks now he says you got problems nobody wants to spend their life as a desktop aha but he says we're going to have a different kind of body instead of going around in this icky body that we have we're going to have a different sort of physical structure all together now ray is a bit hesitant to tell you what the body is going to look like but I remember a movie I once saw where the hero was told consider plastics consider plastic so we're going to have plastic bodies and now here comes another magic word by nanotechnology a plastics by nanotechnology with a decent hardware are going to enable us to achieve immortality but suppose running around your plastic body you're run over by a truck the whole thing you're crushed no worry because you've got a backup disk the backup disk can just be plugged into a new body and you're off and running literally now ray admits you've got to make frequent backups because if you don't make frequent backups you'd have to repeat the same mistakes over and over again I guess you'd have to get divorced again or you'd have to break up with your boyfriend again or whatever you'd have to do this now I'm not inventing this I actually reviewed this guy in the New York Review of Books nice guy and he's very smart and you can look at his books on the subject he keeps sending me many copies obviously because he wants me to give him away to my students but in any case you can I check him I am NOT parodying raise dues at all his name is Ray Kurzweil he is an extreme version of the computational theory of the mind and you might say well it's a kind of reductio ad absurdum of the computational theory of the mind however I'm criticized as stuff yet I'm just telling you what it is so but if you're interested in this I did review his book in the New York Review of Books and you can probably find it online or if you can't oh I'll put my review I can if somebody remind me I'll put it on on bSpace okay so now we're up to the point where the functionalist says we want to know the nature of the mechanism that implements this set of causal relations and at long last we have discovered it it's a computer and the brain we just happen to be implemented in the wetware of the brain but any system at all that is capable of implementing the program could have a mind in exactly the same sense that you and I have minds because that's what a mind is it's an implemented computer program I debated a guy once on British television and he said we are creating Minds well think about he sits there any programs in Lisp all day and at the end of the day he's got a mind on the program he said he is created a mind well I'm a bit skeptical about that my wife was brought up at upper Catholic she's an atheist now but she heard that and she said that's a sin that's that's a sin to say that but in any case whatever else you can say about these guys it's kind of interesting I know I had a lot of fun debating them I it's one of the few things well I would say one of the few things but it is a type it's an issue where I was pretty damn well convinced that I was right and these guys will run but I haven't told you what the issue yet so let me make the case for them I will not have done my job if I don't make this sound possible so I'm now going to tell you the intellectual underpinnings and all the stuff I'm going to tell you in the next few minutes is stuff you should know as educated citizens of the 21st century it's part of the intellectual culture you live in so I'm now going to give you the background for the computational theory of the mind and a lot of it is essential to the theory of computation it has to do with what a computer is and how the computer works and why it is such a powerful type of mechanism so I'm going to introduce a whole lot of crucial notions ok any questions before we do that yeah yes what it you see here was that that is the objection that is made to the identity theorist that says well really it is a type of property dualism because if you're going to say there's an identity statement that you can discover in the same way you discover that lightning is an electrical discharge water is h2o the evening star is the morning star those are the favorite examples of identity statements in each of those cases you have two different sets of properties in order to say that a is identical with B you got to have some a properties over here and some B properties over here and it looks like you're going to have to say well there the mental properties the pain it hurts and then there are the physical properties it's a C fiber stimulation and if you got two different sets of properties that's property dualism now what those guys invariably set well with one exception what they said was we're going to get rid of the mental properties we're going to show that those really are physical properties too so in the end it was a it was like earlier versions of materialism one exception was a guy in Minnesota I only discovered his work after he was dead unfortunately his name is Grover Maxwell and Maxwell says who knows we might discover that all of these glorious psychological features we have we're appreciating the spring morning or rejoicing in romantic poetry that is that's those irreducible mental states irreducible mental properties that are identical physical properties and that's more like my view I'm going to say something not so different from that but the early the early defenders of token of both type and token identity theorist said there aren't any irreducible Cartesian properties and if you believe there are you're in bed with Descartes or is David Armstrong said to me you believe in spooks if you don't think you can get rid of these irreducible mental phenomena then some or other there's some sort of superstition yes okay here is there are two sets of problems one is a behaviors problem if you had a machine that behaved exactly like I do I would that machine have the same mental states that I have and the answer to that by given by that inhave Urist is yes because that's all there is to having the mental states is behaving in a certain way and this was also characterized the computer functionalist so we're going to get to now because what they said is look if the mesh if the machine can do exactly what a human can do it can write poetry and it can speak Chinese and do all the things a human can do then you have to say that it has the same psychological States as the human because otherwise you're there's some mystery of some mystification some irrationality in your view now the question you ask those are tougher question and that is are they token identical and the answer these guys would give no you could I think they'd have to say their ID their identity like identical twins if you you might make a system that was identical with me in its behavior when its capacities then it would have the same mental capacity as I do but it wouldn't actually be me it would be an identical twin to me now Ray has a problem and that is well what thing what makes it me if the machine get that machine got run over by a truck and we put in a new disk that had all the same programs on it and he hasn't answered that but I forget what it is look him up he's he has fun defending his views or not they don't seem very plausible to me and I'm glad I don't have to defend them but look up Kurzweil on Google and see what he says about that question I know he has an answer to that I'd just forget what it is okay here we go we're going to go to work now I'm going to give sort of set beasts lecture on several absolutely fundamental notions that aren't just important for philosophy but they're important for your general education all right the first you're probably familiar with already and that's the distinction between syntax and semantics syntax has to do with form semantics has to do with meaning now the point however is that in any linguistic eye realization the meaning always has to be carried by some form it has to be carried by sentence or sequence of symbols there has to be some syntactical entity which is the bearer of the meaning syntax is form semantics is meaning and you know that you need a distinction because the same form I the same meaning can be carried by different forms we say in English it's raining but in French you say you blew in German you say s hike net and so on in other languages in each case you have the same semantics it's raining but you have a different syntax the syntactical form in English is it's raining I in in French and German it's I implore an S Wagner it's an Eliot's and it's an interesting feature of English that it has this weird tense the present continuous it is raining we can't say it rains it rains is I I is ungrammatical you have to add something rains too often in England or something like that you can do that but just it rains the dramatic present is odd in English but that's that's a foot the button I want you to get across now it is you must make a distinction because the same semantics can be carried by different syntax and of course you all know the same syntax can carry different semantics I went to the bank can be ambiguous between I went to the side of the river and I went to the finance house so you've got a distinction between syntax and semantics I cannot tell you how important this is I but I'll just give you a hint and the hint is that though we think logical relations are about semantics in fact you can get syntax to do almost all of the work in ordinary English if you hear the sentence if it rains tonight the ground will be wet tomorrow it will rain tonight you hear that as implying the ground will be red wet tomorrow and you're all familiar with that form of inference you have rain if rain then wet and you have the extra premise rain and so you can draw the conclusion wet now all of that is I've given it to you is is semantics encoded in a syntax but a very great I think ER Aristotle discovered well you can forget about the specific examples here and just use the if P then Q or they all and some so if you have if P then Q and P then you can derive Q and that's got a name that's called the law modus ponens and it's said the derivation are said to be formal because it works in virtue of the form of the if P then Q and P structure it works in virtue of the syntactical form and you don't have to worry about the meanings of P and Q however in Aristotle you still have to worry about the meanings of if and then and all and Sun the next step is to say you can get rid of all of the ordinary English sentences here and just put in symbols symbols that are otherwise meaningless you have P arrow Q and P and you program your computer so that whenever it sees P arrow Q and P arrow Q and P it prints out Q now notice those operations are now purely over syntactical elements we didn't have to worry about the meanings of if and then and and and so on we just had a bunch of syntactical elements and we had a program that operates on the syntactical elements and generates an answer it that is a prodigious intellectual achievement and that the message of it is where logic is concerned you can forget about the semantics and let the syntax do the work if you get the right match between the semantics and the syntax at the beginning when you state your premises and you follow your rules right the rules are purely syntactical now they operate on this intact you'll get the right semantics out at the end you'll get at the end the ground is wet and that's what is implied by the set of premises I said with very few exceptions you can forget about the semantics the syntax will do all the work and why is that important because that's what the computer is going to do it's going to manipulate symbols I said there are exceptions and the famous exception is girdle's theorem and I won't tell you about that because I'm going to tell you so much crap already this morning I don't want to overburden you but good old theorem shows I well there are things that are semantically right but you can't prove there are truths that are there are truths and logical systems there are statements that are true but they're not theorems you can't prove them okay but that's just a footnote the basic idea now the computational theory of the mind is you can operate with the syntax okay now how do you do it well this is this is point number one syntax and semantics the next point is we need to introduce a notion that you should know anyway and that's the notion of an algorithm an algorithm is any set of procedures that is guaranteed to give you the right answer if you follow the procedures correctly the procedures must be finite it cannot an algorithm cannot have an infinite set of steps but an algorithm if carried out correctly guarantees you the correct answer to a question so for elementary arithmetic there is an algorithm if you follow the steps in the algorithm you will get the right answer to arithmetic equestions unfortunately not all of life's problems have algorithmic solutions falling in love with the right person no one has yet discovered the algorithm for that but stay with elementary arithmetic addition and subtraction yes there are algorithms for doing it right and if you follow the steps you'll get the right answers now why is that so important on suppose we talking about the mind here and itay about syntax semantics and algorithm well the next great invention is a next great notion is the notion of a Turing machine now a Turing machine is an idea invented by Great British logician and philosopher named Alan Turing tragic life incidentally law would force the suicide because of his homosexuality I'll let you track it down on Google it's a ridiculous life that I never met the guy but I knew people who knew him but in any case Alan Turing was a great genius and among his many inventions was the idea of a Turing machine the Turing machine is a machine that operates with only two types of symbol it uses a binary symbolism I and these are usually thought of as zeros and ones but any two symbols would do they could be Chinese symbols or different states of voltage in the computer a Turing machine is a machine that operates with two different kinds of symbols with a binary symbol system I and a program the Turing machine has to have a set of rules for manipulating the symbols however the Turing machine is stunning in its simplicity the rule the only four types of rules the rules there's a tape that moves through the Turing machine and then there's a head that scans the tape and that and the instructions to the head are move one square to the right move one square to the left erases zero and printer one erasing one and print a zero that's it that's all you need for the Turing machine now why am I telling all this because the ordinary computers that you own are in effect Turing machines now I have to qualify that immediately the notion of a Turing machine is not a notion of a machine in an ordinary sense you cannot go into a store and say I want to buy touring machine because a Turing machine is a purely abstract mathematical idea I it has an infinite amount of tape it has an infinite memory and no machine has that computers that I buy in stores are subject to all kinds of problems I poured inadvertently a glass of beer on a brand new laptop I once had and I had a hell of a time explaining it why I still needed them to honor the warranty they did they said well if you have problems it's likely to occur in the first week that's when I appoint the bills good beer - ok but the point is a real Turing machine you can't pour beer on it you don't have to worry that it's going to rust or you're going to drop it on the floor touring machine is a purely abstract notion but it is a notion of a machine that operates by manipulating symbols by manipulating symbols that are of two kinds zeros and ones now again and they and it just goes back and forth according to the program with these zeros and ones now again why is that so important well it is important because it turns out that any algorithm at all anything that has an algorithmic solution can be implemented on a Turing machine this the binary the system of binary symbols is sufficient by itself to implement any algorithm at all so for any algorithm there's some Turing machine that will implement that algorithm now why is that important well what it says in sort of mathematical jargon is that any computable function any function that you can compute in mathematics is touring computable that any I a Turing machine system will be able to carry out any algorithm and that also has a name it's called church's thesis after Alonzo Church and is sometimes called a church-turing thesis it's a thesis and not a theorem because the notion of an algorithm isn't a precise enough notion to admit of a theorem but I don't know any mathematician who would deny church's thesis that any algorithm at all any computable function you can do with a Turing machine a Turing machine is rich enough it has a rich enough apparatus to carry out any algorithmic program now there's another idea I want to introduce in connection with a tauren and that is the idea of a universal Turing machine most of the computers that you have in your car are specialised they're specialized for things like measuring the rate of gasoline consumption or measuring I on the the the efficiency of the cooling system they're special-purpose computers but touring proved there is a universal Turing machine that can simulate any other Turing machine so now to put this more carefully for any Turing machine at all there is a universal Turing machine that can simulate that Turing machine that can carry out the programs of that Turing machine okay so you got the idea then there are these algorithms and when you write them down they turn out to be computer programs I they can be implemented on Turing machines and indeed there is a Turing machine that is so powerful that it can implement any algorithm at all now remember this is abstract mathematics you're not don't don't go into a store and say you guys didn't sell me a universal Turing machine you just sold me an ordinary Apple I want a universal Turing machine it's an abstract mathematical idea but the important point is for practical purposes the ordinary laptop that you are busy messing around with in class when you ought to be concentrating on the lecture and you're busy doing your email or whatever that is a universal Turing machine because it's there's no limit on the number of programs that it can implement okay and now then we've got the reason that this is also exciting is it turns out anything any computable function is Turing computable any algorithm at all can be implemented on a universal Turing machine now why is that so exciting well here is the idea that sent shivers up and down the spines of testosterone laden intellectuals like yourselves about 30 or 40 years ago is maybe we are all universal turing machines maybe the brain is a universal even i feel shivers when i said maybe the brain is a universal Turing machine now why is that so exciting well we've solved all the philosophical problems of the past 2,000 years people wondered how can the mind affect the body how can we get their connection between the mind and the body we know the answer that it's the relation between a program and the hardware implementation but that's a problem we solve on a practical basis every day thousands of times in this very university it's just not a mystery to us so I mean and you have to go back and read the books of that era they're all out of print now I suppose but there are a whole lot of books that all had more or less the same title they were either are called computers in the mind or mind and computers and then there were variations on that and you can look them up and one of my favorites is by a guy at the Rutgers names in on pollution and I wish I had memorized the sentences a gorgeous sense verses after two thousand years of fruitless philosophical speculation we now know the answer to how mental processes can affect behavior the way they do it is that mental processes are states of a physical symbol system that's a fancy word for computer and they affect the the body the same way the states of any computer system affect the other states other computer system and you should look this stuff up because it was a wonderful era if I if I can escape my if I can get out of my snake here and the guy's name is Zen on politian it's one of those polish names with lots of Y's he'll listen yeah I was in on and first name is n on and look up his book and it's one of these books which either mind and computers or computers in mind one of the one of the permutations and if this was a wonderful optimistic era okay we keep going I've given you a whole bunch of great ideas syntax and semantics algorithm Turing machines see the Turing machine need only operate with syntax you don't have to worry about the interpretation because that's imposed you can decide to program any idea you want provided only that you can specify it as a finite set of steps in a precise enough vocabulary that it is implementable on a Turing machine that it's implementable in a system capable of operating with binary symbols now how the hell does it works so well well I'm always stunned by that but the fact is it works well because your modern computer can do literally millions of operations per second in a second it can carry out millions of these steps I always say that in lectures I'm not sure I'd believe it but I do do it every day with my damn computers but in any case that is the reason it's so effective is that you get a very rapid set of computational processes ok now then we need a test I we need a test for whether or not we are effort to create a mind has actually created a mine and once again touring to the rescue touring invented something called the Turing test and the Turing test Oh leaving out various details says look if the behavior of the computer is such that an expert can't tell the difference between the computers answer to the questions and a human answer to the questions then you have to say that the computer has the same cognitive capacity as the human so if you design a computer program that can answer questions in Chinese as well as a native Chinese speaker then you have to say it would just be superstition to deny this you have to say that the computer program or more strictly the implemented computer program can understand Chinese the Turing test in short is a form of behaviorism it says if it behaves like a duck and walks like a duck and acts like a duck it's a duck and what the Turing test says is if anything that can simulate the behavior of a human being so accurately that an expert can't tell the difference between the behavior of the computer the success of the computer and a human being then you have to say the computer has the cognitive capacities of the human being it has an equivalent cognitive capacity so we got five ideas out here syntax semantics algorithm touring machine churches thesis and the Turing test now we're going to show how it might actually work in the human brain I so let's stop for questions I don't know if this is fun or not but anyway these are important ideas you ought to know this kind of stuff quite apart from philosophy you won't be a fully a citizen of your culture if you don't know what a Turing machine is and what an algorithm is and what the Turing test is this is part of general culture okay like questions about what I said so far yes hang on I got to get closer my hearing isn't all that great anymore so you have to shout loud yeah how does he claim what learning okay what happens in learning is it you reprogram you can alter the program or you can introduce new programs now you have to have a learning program and to enable you to learn so the learning program enables you to learn Chinese but of course once you learn Chinese then you have got the Chinese understanding program so you can have programs that operate on programs now there are complexities to this some people thought well the problem with the brain is that it's if it's a regular digital computer is too slow I mean there's a what's called a hundred step I the brain is so slow that it can only carry about out about a hundred computational if it were digital computer one hundred steps I in a half a second but a hundred steps is nothing anybody who programs the computer knows you can't do anything interesting with a hundred lines of computer programming I have a friend who works for Xerox and he said that a really good Xerox machine has half a million lines of computer programming you know the kind that will print on both sides and then staple it and all that kind of stuff that takes a lot of computer programming so it looks like the brain can't do much if it really is a conventional digital computer AHA but there's another theory to the rescue and that is it's not a linear digital computer it's a special kind of a computer that uses parallel distributed processing that's another magic word PDP well get to that okay nowadays we're still back in the 1970's and 80's with the idea that whatever else that brain is it's a digital computer and some people even said look the fact that neurons either fire or don't fire they're either on-off that proves we're digital you see because fire don't fire that's like zero and one that's evidence that we are in fact digital this guy then that guy you're next yeah well human error can arise because you've got a bug in the program or you may have a fault in the hardware you had too much to drink and that affected the hardware but these are kind of details you know if all the only problem with its theory is well you guys have got to better explain how I'm so dumb at doing long division they wouldn't think that's not our big worry ISO but they do have answers to that again I'm in a position of Voltaire trying to explain the Catholic Church with this I'm trying to make it seem plausible I don't believe that the brain is a digital I mean I just think it's well well I'll get to my objections you're going to hear more of them than you ever wanted to hear okay yes yeah yeah okay the question is how is it that the that the people that people in animals like honeybees can be so smart if they have such a simple computational make the machine is such a simple machine right okay we're going to get to that and there are there there is a key notion it's recursive decomposition I love these fancy words with a lot of syllables we'll get to that it's about you're five minutes away from recursive decomposition yes yeah yeah yeah yeah this was a standard objection at the time I'll repeat it for people in the back it seems to work well for mathematics but how about emotions and how about art how about aesthetic reactions a fair summary of the question and this was a standard objection and there was always an argument against the computational theory of the mind that said the computer can do x and y but it can't do Z and then followed your favorite thing the computer can do arithmetic and logic but it can't fall in love or it can't feel the angst of post-industrial man under late capitalism or whatever you whatever is your your favorite feeling the idea was the computer is pretty good but there's that last little bit that the computer can't do I think that's not a serious objection of these guys because what they did was immediately I try to get computers that could do that last little thing some people said well the computer can't write poetry immediately they produce programs that could write poetry wasn't very good but still Who am I I say that to be I you don't want to think you can defeat these guys with literary criticism and furthermore they say yeah but it can't fall in love oh yeah well we can get a computer that says I love you I says all the things that that you want the computers say you can program and you can even program your computer to say I am suffering the angst of post-industrial computers under late capitalism why not so anything you can program you can get the computer to do and if the touring tested to be taken seriously you might get somebody who was fooled in the thing oh well the computer really is in love there was a guy who invented a program that would do a psycho analysis what's his name Weizenbaum right with Joe why isn't mom invented this it wouldn't do psychoanalysis and it did with simple tricks so it asks you certain questions tell us about your family and as soon as it heard the word mother it would ask certain questions how do you get on with your mother it just did and whenever didn't know what to say it would say tell us more or something lady had a certain gimmicks it was a Mickey Mouse program but the point is a lot of people were fooled I am wise mom said that his secretary insisted on being alone with the program during the lunch hour and occasionally she'd come away in tears because of the pre-fund profound insights that she'd got in the program so you can get at programs to do all kinds of dumb things the question is what is the philosophical significance of that and I haven't answered that yet I'm telling you what these guys believe now I think their conclusion is false but the but the ideas are not false all this stuff that I've told you is all very powerful stuff the distinction with syntax and semantics existence of algorithms the fact that they're programmable on Turing machines and the fact that any computable function by any algorithm at all can be implemented on a Turing machine I think those are all true and all powerful ideas it's just the conclusion they draw is the wrong conclusion one more question that I'll go on yeah yes yeah yeah the idea is that any linguistic expression will be either a primitive that is undefinable or more likely for complex expressions you will be able to analyze it into simpler expression so you analyze complex feelings if you are suffering the auction of post-industrial men under late capitalism I don't suffer that honest but a lot of my friends think I'm obviously a Philistine because they do suffer that long stay then presumably you can analyze that into simpler feelings and the idea is you break it down into the simplest components and the earliest computational theories of language understanding were always like that they always tried to break down the complex linguistic ideas into simpler ideas the guy who did a lot of work on that was Roger shank at Yale it didn't work for a reason I'm going to tell you when we talk about language but that I yeah that was the idea that you're going to analyze as with any computer program you analyze the complex ideas in two simple elements okay let's go on the next great idea on this list of seminal ideas of the last half of the 20th century is the idea that in any of these systems there are levels of description and in any physical system you have to be able to think you have to be able to describe it at different levels at the top level you might say I own a Volvo station wagon but of course there are lower levels of description where you describe the body its shape the paint than their lower levels of description of that where you talk about the particular components of the I carburetor and then finally you get down to the lowest level which would be the physical particles which would be the specific molecules and subatomic particles that go to make up the engine any system of any complexity or at all will have different levels of description and the levels of description the idea that the same system admits of different levels of description is important because a system can be described in the same way as another system at a higher level even though at a lower level they're quite different so I you might have the same program as a computer and at that level of description you're implementing the program but at a lower level of description you have a different hardware structure the computer is made out of silicon I and you are made out of various types of organic molecules so the reason that levels of description is important for this particular discussion is because of an X notion I've run out of numbers here what am i up to we were up to about five four five well I'll let you figure this is this six levels of description okay then the reason that's important is because of multiple realizability the same higher-level phenomenon can be realized in different physical substrates and indeed there is a diagram that you always see in the textbooks that look looks like this at this level from A to B you have the same description we're implementing the same program but there's lots of different hardware's for implementing that there's C and B and E and F and G so at the level of description a to be these systems are all the same because they're all implementing the same program but they have different hardware's and you're all familiar with this both the the Macintosh and the PC can implement the word program this will be the word program and this will be the different hardware implementations and the idea now is that at one level of description you are implementing the same program as another person or as another computer even though at a lower level of description that has a different Hardware realization how is that possible it's possible because this of this wonderful feature of multiple realizability the same program can be realized in an indefinitely large range of different hardware's it can be realized in an indefinitely large range of different hardware's provided only that each of them has a level of description where they're implementing the same program so the physics can be different even though the computational level is exactly the same I hope you see where this is leading the idea is we have minds but our minds are computer programs at the level of description where we say I'm thinking about my income tax I am at the same level of description as a computer that I programmed to work on my income tax even though that level of description is implemented in me in one way and in the computer in another way it's got a different computer hardware so levels of description is essentially tied to the notion of multiple realizability now there's one other notion that's crucial I want to introduce that and then we'll then I'll stop for question that is I I love all these fancy expressions this is one is recursive decomposition how is it that we are so smart how is it that for example we can solve problems in arithmetic and the answer is because we are parts of a computer system or rather we have in our brain a computer system which can break complex problems down into simple problems and solve them one at a time so if you ask me how is it that I multiply 371 and 582 I mean if I was going to do that well the answer is I consciously go through a series of simpler steps but while I'm consciously going through those steps my brain is even going through even simpler steps remember at the rate of several million per second so if you ask how does the the your pocket calculator how does it multiply seven times a to get 56 well the first part of the answer might be well look it takes seven and adds it to itself seven times they always say eight times but that's bad arithmetic you start with seven anyway okay so you reduce multiplication to addition and how does it multiply seven plus seven well first it come converts seven into binary symbols it converts it into a bunch of zeros and ones and then it just operates on them on them one plus one plus one nothing more complicated than that so the complex task of multiplying seven times eight is reduced to a series of very simple tasks adding one plus one that's all the computer needs to do is simple operations on binary symbols on to two symbols and that means that the complex task can be decomposed that's the decomposition part into simpler tasks and you do it over and over you keep you asking how do you do that how do you do that until finally you get to the simplest level which is the level of zeros and ones and those steps whereby you go from the more complex to the less complex that's called recursive because you do it over and over again you go through the the reduction over and over so the reason we're so smart is because the actual complex tasks that we think are so smart like composing a work of philosophy or a you're not writing a term paper those decompose into simpler tasks and those decompose into simpler tasks until recursively we get down to where the work is really done 1+1 print of zero erase a one move one square to the left everything reduces by recursive decomposition to the simplest type of complex operations okay now unless you understand this stuff that I've been telling you you're really not a fully intellectual citizen of the other of the of the of the century now that doesn't mean you can forget about Baudelaire and James Joyce you got to understand them too but that's not today's lecture okay I'm not talking about yes Baudelaire oh you prefer Rambo okay Rambo and Baudelaire but in any case where I there's a whole lot of other stuff but this is necessary that you should understand these however though I think all this stuff I told you is well up to a point it's valid recursive decomposition it's got some problems the way I characterized it but here is the next step how does the mind work I have given you the machinery for answering that question the mind is a universal Turing machine programs or rather a set of programs implemented in the brain and the brain then is able to carry out complex tasks because the complex tasks are recursively decomposable into the primitive computational operations that are all and by arrey symbols are all zeros and ones we have a test for telling whether or not we have succeeded in creating mine it is the Turing test and any system that passes the Turing test will have a mind in exactly the same sense that each of us has minds and you don't have to worry about the specific hardware that you're in because I your mind is multiple-- realizable this is why Ray Kurzweil can say we're gone we ought all get uploaded onto a decent hardware and get out of this realization we got because it's messy and it decays and it dies all that and a better hardware won't so they're out of all of this comes an equation which was the guiding light of a certain conception of the mind and of cognitive science the equation was the mind is to the brain as the program is to the hardware we know how programs are interim nted in hardware's we know how the same program can be implemented in an indefinitely large range of different hardware's well that's the key to understanding the mind one of the many advantages to this theory is it turns out you don't have to know how the brain works to know how the mind works my friends who do artificial intelligence programming are rather proud of the fact that they can't fix the damn computer when it breaks down I you know I say them come on you're a professor computer science what's going on here that's for plumbers you know call an electrician I do it serious intellectual work I do work at this abstract level of the touring machine program so if it's right that if this equation is right that the mind is to the brain as the program is to the hardware then you can study the mind independently of knowing how the brain works and this was and there are a lot of this was where important ideas in the early days in the founding days of cognitive science because the question was hi going to do cognitive science if you don't know how the brain works and a comforting answer was you don't have to know how the brain works because the brain just happens to be the material in which we're implemented but any physical system would do we just happen to be implemented in a in a certain type of hardware system but any hardware system that can carry the programs I because of multiple realizability any hardware systems that could compare the programs it can carry the programs will be equivalent will have a level of description where it has the same cognitive capacities that I have or that you have now I just have a label I call this when I first wrote about I call it strong artificial intelligence artificial intelligence was the effort to create computer programs that could I do what in humans we would think of as intelligent behavior and strong AI I characterize as the view that as this view this is a summary of strong AI that all there is to having a mind is having the right computer program a computer program that can carry out the algorithms in such a way as to pass the Turing test but again I want to emphasize how this seemed to give a wonderful research program for cognitive science you design a computer program and if you get the computer to carry out what humans can do you've now got a theory you've got a theory of human cognition because the theory is what the eye computer is doing and what the brain is doing are the same at some level of description at the probe level how do you know you've got the right program because there are different programs that could carry out it could give this result and there you call in the psychologists and what they do is to try to see if the program that the brain is carrying out is the same as the program that the computer is carrying out and they did this with their favorite research tool at the time was called a reaction time experiments and the idea was if the delays in the human cognition match the delays in the computer got me pretty small delays for the computer but anyway if you get reaction time experiments that show that the type of reaction times that humans have I mimic or or isomorphic with or similar to the reaction times that the computer has you have good evidence that the computer gives you a theory of human cognition it gives you a theory of what the brain is actually doing so you've got a wonderful research program you don't have to know how the brain actually works as a hardware system who cares about the number of carbon rings in serotonin or I the structure of the Perkin J cells I mean that leave that to the plumbers leave that to those brain stabbers we will do serious hard-nosed our cognitive science and that consists in designing computer programs that can pass the Turing test and then we get our psychologists to do reaction time experiments to see if the humans are actually using the same programs as the computer if they are the computer program is more than just a program it's a theory we now have a theory of human cognition we can show you with actual demonstrations that this is how human cognition works I it is a wonderful idea it is false in almost every respect and I couldn't resist pointing this out in article I wrote in 1980 which you're going to have to read and I think at last I'm now beginning to convince these people I the evidence is this I have in the past year been invited to address two large international gatherings of artificial intelligence experts at one point I remember the artificial in Journal said the main purpose of artificial intelligence right now is to refute Cyril some years ago I thought well you know that's not bad have we got time I see people starting to but I'm by my o'clock we seem to have a couple more minutes I can cover a lot of ground a couple minutes so don't snap your notebooks yet okay couple more minutes okay so this is strong artificial intelligence and I now carried you through the great dramatic sequence that goes from Descartes to what people thought was the culmination of the philosophy of mind the computational theory of the mind I baptized it as strong artificial intelligence weak artificial intelligence I just said was a theory you can use the computer to study the mind the way you're using a beuter to study anything kind just about any department in this university in the sciences has to use computational modeling whether it's soil science or cell biology they use computers all the time so the definition of weak artificial intelligence was I you can I use the computer to study the mind as you can use it to study anything you don't make any strong claims to the effect that the computer isn't mine by the way as a piece of rhetoric I should never have labeled these strong and weak because none of these guys wanted to be said to be weak you know whatever they thought of themselves were not wimps so that was just meant when philosophers say strong and weak they mean the theory is more likely to be false if it's the stronger theory the weaker theory is the one that's likely to be true I think weak artificial intelligence is true strong artificial intelligence is false okay and I'm going to tell you on Tuesday why it's false but I got to tell you a lot about a crap first ok I'll see you on Tuesday

Channel: SocioPhilosophy

Views: 21,317

Rating: 4.9000001 out of 5

Keywords: John, Searle, Philosophy, of, Mind, University, California, Berkeley

Id: qivOmLRn5Oc

Length: 72min 47sec (4367 seconds)

Published: Sun Oct 09 2011