I think of Dan Dennett as a great fountain of ideas, and he's like a fireworks display for me. On every page of his you read, you constantly put ticks in the margin. I'm never quite sure why he's classified as a philosopher rather than as a scientist; he seems to me to do the same kind of thing I do in a somewhat different field, and I greatly admire the way he thinks, the way he uses metaphors to try to get his points across. And they're elegant metaphors; they really make you feel he's hit the nail on the head. My complaint about him is that his books set you thinking so hard that you have trouble turning to the next page, because you're so busy thinking about what's on the current one.

I think of Dan Dennett as a great fountain of ideas, and he's like a fireworks display for me. On every page of his you read, you constantly put ticks in the margin. I'm never quite sure why he's classified as a philosopher rather than as a scientist; he seems to me to do the same kind of thing I do in a somewhat different field, and I greatly admire the way he thinks, the way he uses metaphors to try to get his points across. And they're elegant metaphors; they really make you feel he's hit the nail on the head. My complaint about him is that his books set you thinking so hard that you have trouble turning to the next page, because you're so busy thinking about what's on the current one.

I've always been interested in Dennett's work, because he's interested in the main scientific questions I deal with — namely, how the mind is engineered; how the kinds of abilities we all take for granted, like recognizing a face or using common sense, get executed by mental software. His perspective of seeing psychology as reverse engineering is one I share in my day-to-day experimental work.

In forward engineering, you start off with an idea of what your machine is supposed to do and then you go out and design the machine. Biology, including cognitive science, is a kind of reverse engineering: you start off with a machine — namely, the human being — and you have to figure out what purpose it was designed for. The main impediment in getting other scientists to understand the complexity of intelligence is that people... [+]

I've always been interested in Dennett's work, because he's interested in the main scientific questions I deal with — namely, how the mind is engineered; how the kinds of abilities we all take for granted, like recognizing a face or using common sense, get executed by mental software. His perspective of seeing psychology as reverse engineering is one I share in my day-to-day experimental work.

In forward engineering, you start off with an idea of what your machine is supposed to do and then you go out and design the machine. Biology, including cognitive science, is a kind of reverse engineering: you start off with a machine — namely, the human being — and you have to figure out what purpose it was designed for. The main impediment in getting other scientists to understand the complexity of intelligence is that people have minds that work so well that they're apt not to be suitably impressed by what their minds are doing, in the same way that they're apt to be unaware of what's going on when they digest food.

I enjoyed, but disagree in some ways with, Dan's discussion of consciousness in Consciousness Explained. I like it because Dan challenges us to come up with an argument for why we should believe that there exist some kind of raw feelings, or qualia, or subjective experience. He argues that there isn't any substance to the idea: a person with what we think of as consciousness and a zombie who behaved in the same way would be indistinguishable, as far as science is concerned.

I agree that the qualitative experience is not the key to understanding intelligence from a scientist's point of view. The scientifically tractable aspect of consciousness is not the fact that there are people or animals subjectively experiencing it, but the fact that some kinds of information are mutually accessible and others are not, and that there is therefore a portion of mental information processing that has a different status than the rest of it. That's one sense of consciousness: information that's accessible to a particular body of information-processing involved with the current environment, and which in humans can interface with the verbal apparatus. So one can study, for example, why some information — say, an overlearned skill, like operating a stick shift when you're an experienced driver — is beneath the level of consciousness, whereas other kinds of mental processing (like how to operate a stick shift the first time you're learning to drive), where you have to reason something out consciously, step by step, is different.

Many people say that Dan's book should have been called Consciousness Explained Away instead of Consciousness Explained. (People congratulate themselves for that supposedly telling witticism, not realizing that Dan used it as a heading in the book!) The reason the book is in some way unsatisfying is that there's another aspect of the problem of consciousness for which no one has yet come up with a satisfactory explanation: why there's a clear intuitive difference between one organism that feels pain and another organism that acts as if it feels pain but doesn't really feel it, or one organism or system that has the experience of seeing red when a red object is in front of it, whereas another acts identically in every way but does not have the experience. Until one addresses the problem of why that's so compelling an intuition, one isn't going to have a completely satisfying account of consciousness.

I read Dan as saying that we've been misled into thinking there's a real question there. According to Dan, there isn't. That's where I disagree: I suspect there's a real question and that it's not just an error in the way we conceptualize the problem. Perhaps our minds are simply not designed to be able to formulate or grasp the answer — a suggestion of Chomsky's that I know Dan hates. But the intuition that qualia exist is real, and as yet irreducible and inexplicable. For one thing, all our intuitions about ethics crucially presuppose the distinction between a sentient being and a numb zombie. Putting a sentient being's thumb in a thumbscrew is unethical, but putting a robot's thumb in a thumbscrew is something else. And this isn't just a thought experiment; the debates over animal rights, euthanasia, and the use of anesthetics in infant surgery depend on it.

One other area in which I disagree with Dan is the explanation of human intelligence in an evolutionary context. Dan makes heavy use of Richard Dawkins' concept of the meme — an idea that replicates, mutates, and differentially spreads in the medium of brains in the same way that a gene replicates, mutates, and differentially spreads in the medium of bodies. This is Dan's main way of placing cognition in the context of evolution, rather than having it appear by magic; thoughts are created by a process analogous to the process of natural selection. But there are many other ways of explaining the emergence of human intelligence in a nonmiraculous way. I think it's much more plausible that evolution designed a brain that's a kind of computer that can generate complex ideas, in ways that need not be analogous to the operation of natural selection itself.

There's a big difference between gene selection in the design of organisms and meme selection in the design of mind and culture. For organisms, undirected variation followed by selection is the explanation, and the only explanation, for complex design. In contrast, because the brain is a complex machine that was itself designed by selection, "mutations," or ideas, are virtually always directed, and meme dynamics need not be the design source (though I agree it plays a big role in the demography of ideas: how many copies are out there). Memes such as the theory of relativity are not the cumulative product of selection of millions of random, undirected mutations of some original idea, but each brain in the chain of production added huge dollops of value to the product in a nonrandom way.

Here's another way of putting it. I think Dan thinks that the parallelism between genetics and memetics is profound: that it's the key to exorcising the hated idea that the mind came from nowhere, that it's a magical, miraculous "skyhook," hanging in midair. According to Dan, the power of the theory of selection of replicators is that it can explain organisms and culture in the same way. Maybe so (in my view, that would be an interesting coincidence), but then again, maybe not, and if not, it shouldn't matter to Dan's larger argument — namely, that the mind is a product of evolution. Perhaps, as the anthropologist Dan Sperber argues, the formal mechanisms that explain cultural evolution are from epidemiology, not population genetics — ideas spread like contagious diseases, not like genes.

Dennett's concept of relational order in relation to the brain is something I find extremely interesting. He suggests that the properties of mind aren't material properties, they're relational properties. That leads to the strong AI position. I tend to take a similar view with respect to artificial life — a view similar to the strong AI position, the idea that you can actually get intelligence in systems that aren't constituted of molecules and cells. You can get life in computers.

Dennett's concept of relational order in relation to the brain is something I find extremely interesting. He suggests that the properties of mind aren't material properties, they're relational properties. That leads to the strong AI position. I tend to take a similar view with respect to artificial life — a view similar to the strong AI position, the idea that you can actually get intelligence in systems that aren't constituted of molecules and cells. You can get life in computers.

Dan Dennett is my favorite philosopher, because he takes the trouble to understand things. I get annoyed at the traditional school of philosophy, whose members believe that they already understand things, so they pontificate on artificial intelligence without having the slightest idea of what the work is. Dennett, although I often disagree with him, takes the trouble to read the technical literature, and understand what people are doing in areas like linguistics, artificial intelligence, biology. His philosophical ideas are informed. They are sometimes wrong, but at least they are informed wrong as opposed to uninformed wrong.

Dennett sometimes is a sucker for a reductionist theory that seems to explain something. Maybe he inherits that disease from biologists. For instance, I think he's been suckered a little bit by Richard Dawkins' view... [+]

Dan Dennett is my favorite philosopher, because he takes the trouble to understand things. I get annoyed at the traditional school of philosophy, whose members believe that they already understand things, so they pontificate on artificial intelligence without having the slightest idea of what the work is. Dennett, although I often disagree with him, takes the trouble to read the technical literature, and understand what people are doing in areas like linguistics, artificial intelligence, biology. His philosophical ideas are informed. They are sometimes wrong, but at least they are informed wrong as opposed to uninformed wrong.

Dennett sometimes is a sucker for a reductionist theory that seems to explain something. Maybe he inherits that disease from biologists. For instance, I think he's been suckered a little bit by Richard Dawkins' view of genes as the central player, because it appears to explain things. People might even argue that he's been suckered by the simple theories of AI into believing that they explain too much about the mind. Fundamentally, he's a reductionist, and he does believe that the phenomena we see in the mind are the result of fundamental physical principles. That's a philosophical standpoint I'm basically comfortable with. It maybe makes him more popular among the scientists than among the philosophers, because if he's right then all philosophy is just a matter of science that hasn't been done yet.

Dennett's ideas are compatible with the notion of science that there's a reality out there; it's understandable; it's based on some simple underlying laws, and we just need to understand what those laws are and the connection between them and what we see. Philosophers have always felt that there's a set of things that don't fit that paradigm. People used to say, "Well, the laws may apply on Earth, but they're not true for the heavenly bodies." Then, after Galileo, they said, "Well, that might be true for physical bodies, but it's not true for biological organisms." After Darwin, they said, "Well, that might be true for our bodies, but it's not true for our minds." And so on. We are backing the philosophers into a corner and giving them less and less to talk about. In some sense, Dennett is cooperating with the enemy by helping us back the philosophers into a smaller and smaller corner, and I like that.

While Dan focuses on the cognitive level, my own approach is to think about all levels, perhaps because I'm influenced by the broad idea of nonrepresentationalist knowledge. In my reality, knowledge coevolves with the knower and not as an outside, objective representation.

Dan is against the idea of experience bearing on science. I'm not very fond of doing psychological readings of people. I do have this distinct impression from a long discussion with Dennett, who, unlike Minsky, is somebody you can engage in conversation and who will read the other person's point of view. It's a delight to have a debate with him. For reasons I still don't understand, he has an absolute panic of bringing experience and the subjective element into the field of explaining consciousness.

Dennett doesn't deny that people have minds.... [+]

While Dan focuses on the cognitive level, my own approach is to think about all levels, perhaps because I'm influenced by the broad idea of nonrepresentationalist knowledge. In my reality, knowledge coevolves with the knower and not as an outside, objective representation.

Dan is against the idea of experience bearing on science. I'm not very fond of doing psychological readings of people. I do have this distinct impression from a long discussion with Dennett, who, unlike Minsky, is somebody you can engage in conversation and who will read the other person's point of view. It's a delight to have a debate with him. For reasons I still don't understand, he has an absolute panic of bringing experience and the subjective element into the field of explaining consciousness.

Dennett doesn't deny that people have minds. He says those minds can be useful only if you treat them as overt behavior, as an anthropologist does with a foreign culture. You take them at face value. If you tell me you're in pain, I believe you. Then I note it down in my book. Then I consider it as overt behavior. That's what he calls heterophenomenology — or, more classically, the intentional stance. He treats you as if you're something capable of intentionality.

I find that far too weak to support a theory of consciousness, because it's just one leg. The other leg, which is the real phenomenology — that is, the "as is," firsthand, direct account of the quality of experience, is irreducible. To the extent that it's irreducible, his whole enterprise just falls short of getting down to the tack we need to get down to. On the positive side, what Dennett has done, probably better than anybody else, in terms of theorizing and writing, is to eliminate what he calls the ghost of the Cartesian Theater. He argues that you have to take this distributed phenomenon of the emergent properties of the brain in order to account for consciousness. In that sense, it's quite brilliant.

The other thing I credit him for is that he's introduced in philosophy of mind a style of discussion that's very rich: he comes into it with a philosopher's discipline, but he takes into account results from empirical research. You can't say that of people like John Searle, who keeps talking about philosophy of mind in a very dry, abstract, armchair way. I like Dennett's pulling up his sleeves and going into the lab with people. He's done something quite revolutionary, which is to steep himself in the scientific literature.

Dan's a purist, who can be tough-minded to a fault. He's wedded to the way of looking at things he learned from Gilbert Ryle, at Oxford. Its roots are in logical positivism and behaviorism. Basically it prescribes what you can talk about and what you can't: the meaning of statements lies in the way you would verify them by observation, and if you can't offer any sort of verification, forget it. Dan got trapped by the beauty of this approach. And if it meant denying the reality of things we all know are important — like sensations, raw feelings, all the qualitative aspects of consciousness — too bad. You have to be brave to be a philosopher. You have to follow where your arguments take you, until you get proved wrong. And since no one has proved Dan wrong, he's still out there.

Of course, part of Dan is uneasy about... [+]

Dan's a purist, who can be tough-minded to a fault. He's wedded to the way of looking at things he learned from Gilbert Ryle, at Oxford. Its roots are in logical positivism and behaviorism. Basically it prescribes what you can talk about and what you can't: the meaning of statements lies in the way you would verify them by observation, and if you can't offer any sort of verification, forget it. Dan got trapped by the beauty of this approach. And if it meant denying the reality of things we all know are important — like sensations, raw feelings, all the qualitative aspects of consciousness — too bad. You have to be brave to be a philosopher. You have to follow where your arguments take you, until you get proved wrong. And since no one has proved Dan wrong, he's still out there.

Of course, part of Dan is uneasy about where his theories have taken him. He's much too sensitive not to be. He realizes there's something missing. When his critics point out what they think are the weaknesses, he hates it and demands they say just what they mean. Often as not, they're reduced to mumbling, because it really is very hard to fault Dan's theory on its own terms. But I suspect that if anyone is aware of the problems, Dan himself is. It's just that he's not going to surrender to people who haven't understood him to begin with. He's not going to give way to people who challenge him on wishy-washy metaphysical grounds.

Dan's book Consciousness Explained is tremendously original, and it's already having a huge impact on cognitive psychology. He's produced the best account yet — a brilliant, funny, beautifully written description of the inner processes underlying thought. But while it's so good on the question of thinking, it's much less good on the question of feeling.

If you're going to explain "consciousness," you have to come to grips with the kind of consciousness that really counts with ordinary people. What do people want to have explained? What do they mean by consciousness? Or rather — since they may mean different things at different times — what is it they really care about?

If you listen to the kinds of questions people ask about consciousness — "Are babies conscious?" "Will I be conscious during the operation?" "How does my consciousness compare with yours?" and so on — you find again and again that the central issue isn't thinking but feeling. What concerns people is not so much the stream of thoughts that may or may not be running through their heads as the sense they have of being alive in the first place: alive, that is, as embodied beings interacting with an external world at their own body surfaces and subject to a spectrum of sensations — pain in their feet, taste on their tongue, color at their eyes.

What matters in particular is the subjective quality of these sensations: the peculiar painfulness of a thorn, the saltiness of an anchovy, the redness of an apple, the "What it's like" for us when the stimuli from these external objects meet our bodies and we respond. Thoughts may come and thoughts may go. A person can be conscious without thinking anything. But a person simply cannot be conscious without feeling.

Here's the paradox, though. What figures so strongly in ordinary people's conception of what matters about consciousness figures hardly at all in Dan's account of it. In Consciousness Explained, there's hardly anything about sensory phenomenology. Once when I said that in print, Dan pointed out to me in no uncertain terms that I'd ignored the several passages in the book where he does talk about sensations. Well, O.K., it's there if you look for it. There are some passages where he talks about sensations and feelings as complex behavioral dispositions (which is, I think, on the right lines, provided you allow that their complexity may mean that they're qualitatively in a different league from anything else.) But my point is that for Dan, the question of sensory phenomenology is no more than a side issue, never the central mystery it is for me.

Dan Dennett is the AI person's dream philosopher. We had all those years of putting up with philosophers like Hubert Dreyfus, who felt the need to attack AI without any attempt to understand it. Dan has made a real effort to understand AI and cognitive science, and he is the consummate philosopher in our world. I always enjoy listening to him; he always says clever things; he's one of the great fun people in our field.

What philosophers are doing is trying to put into perspective things that other people have thought. Dan does more than that, of course. He has his own thoughts, too. But it's not as if there's stuff that an AI person is likely to learn from a philosopher that will help them in AI. It's interesting to read philosophy, but it doesn't give you something you could somehow put into a program.

Dan Dennett is the AI person's dream philosopher. We had all those years of putting up with philosophers like Hubert Dreyfus, who felt the need to attack AI without any attempt to understand it. Dan has made a real effort to understand AI and cognitive science, and he is the consummate philosopher in our world. I always enjoy listening to him; he always says clever things; he's one of the great fun people in our field.

What philosophers are doing is trying to put into perspective things that other people have thought. Dan does more than that, of course. He has his own thoughts, too. But it's not as if there's stuff that an AI person is likely to learn from a philosopher that will help them in AI. It's interesting to read philosophy, but it doesn't give you something you could somehow put into a program.

Dan Dennett is obviously somebody who'll listen to arguments. The title of his book Consciousness Explained is overstated, however. I certainly don't believe that those ideas explain consciousness. He's exploring what I call "point of view A," in the list of four viewpoints I discuss in Shadows of the Mind, which I call A, B, C, and D. A is the strong artificial- intelligence viewpoint: that is, that mentality is to be understood in terms of computation. It doesn't matter what's doing the computation; a computer or a biological structure would be equally good.

Point of view B — which is more the philosopher John Searle's viewpoint, as I understand it — is that you could simulate the action of the brain, but the simulation wouldn't have mental attributes, so there's something other than... [+]

Dan Dennett is obviously somebody who'll listen to arguments. The title of his book Consciousness Explained is overstated, however. I certainly don't believe that those ideas explain consciousness. He's exploring what I call "point of view A," in the list of four viewpoints I discuss in Shadows of the Mind, which I call A, B, C, and D. A is the strong artificial- intelligence viewpoint: that is, that mentality is to be understood in terms of computation. It doesn't matter what's doing the computation; a computer or a biological structure would be equally good.

Point of view B — which is more the philosopher John Searle's viewpoint, as I understand it — is that you could simulate the action of the brain, but the simulation wouldn't have mental attributes, so there's something other than computation involved in conscious thinking. That differs from point of view C, which is my own point of view and asserts that you can't even simulate conscious activity. What's going on in conscious thinking is something you couldn't properly imitate at all by computer, according to C.

Point of view D asserts that you can't understand mentality in terms of science at all. So I'm saying, "Yes, it's science, but it's science of a kind that eludes computation." Dennett belongs to the A point of view, of which he's one of the best exponents. Another person representing this point of view is Hans Moravec, who's written an interesting book, where he takes this point of view to its extreme and argues that within thirty-five years or so the computer will achieve our level and then race beyond us.

There are at least two different kinds of arguments you can use against A. One is the John Searle type of argument, which is that just because something carries out computations, that doesn't make it capable of being aware of anything. That argument has quite a lot of power to it. But my argument is different, because it argues against both A and B. It's a stronger argument, because it says that you can't even properly simulate conscious actions. If something behaves as though it's conscious, do you say it is conscious? People argue endlessly about that. Some people would say, "Well, you've got to take the operational viewpoint; we don't know what consciousness is. How do you judge whether a person is conscious or not? Only by the way they act. You apply the same criterion to a computer or a computer- controlled robot." Other people would say, "No, you can't say it feels something merely because it behaves as though it feels something." My view is different from both those views. The robot wouldn't even behave convincingly as though it was conscious unless it really was — which I say it couldn't be, if it's entirely computationally controlled.

Dan Dennett is our best current philosopher. He is the next Bertrand Russell. Unlike traditional philosophers, Dan is a student of neuroscience, linguistics, artificial intelligence, computer science, and psychology. He's redefining and reforming the role of the philosopher. Of course, Dan doesn't understand my society-of-mind theory, but nobody's perfect.

Dan Dennett is our best current philosopher. He is the next Bertrand Russell. Unlike traditional philosophers, Dan is a student of neuroscience, linguistics, artificial intelligence, computer science, and psychology. He's redefining and reforming the role of the philosopher. Of course, Dan doesn't understand my society-of-mind theory, but nobody's perfect.