Tim Crane

1.) Talking about the 'philosophy of the mind' requires one to have a very open mind. How do you view the brain from a philosophical stance, a computational viewpoint, and a combination of the two?
I would not want to oppose the computational viewpoint and the philosophical viewpoint. 'Philosophical' means many things, of course, but computationalism may be one philosophical approach among others. An important question for philosophy is how to conceive, in the most general terms, of the relation between mind and brain. I think we know a lot about our minds before we even embark on an investigation of the brain. We know, for example, that we have thoughts, experiences, emotions, sensations and so on. One question that then arises is what the relation is between what we know and what we are beginning to learn from the various sciences of the brain. Some think that scientific discoveries will threaten our claims to knowledge about the mind. I disagree: I agree with Descartes that we are more certain of our knowledge that we have minds than we are of (most) other facts. So we must try and understand how the facts we have discovered are consistent with the facts we know. My own view on the mind-brain relation is that mental properties are emergent properties of brains (or, better, of people and other minded creatures) which are causally effective in bringing about behaviour. I therefore deny the identity of mind and brain; I assert rather a dependence. This question is independent of the question of the correctness of a computational view; for the reasons given in my 1995 book, 'The Mechanical Mind', I am sceptical that computationalism can be the whole story about the mind.

2.) In your book "The Mechanical Mind" you explain different ways that knowledge in the brain could be represented. Do you believe that knowledge IS explicitly represented, or do you take a more 'connectionist' point of view? Why?
This is a hard question to answer until more is said about what 'explicit' and 'implicit' mean. In general, I think that a lot of the debate about explicit and implicit representation has never made it clear what these terms mean. But, as I say in chapter 5 of my book, I do think that connectionist machines contain representations, even if they contain no representation of a rule. Other than that, I agree with Daniel Dennett ('A cure for the common code?' in 'Brainstorms') that the idea that the mind contains a representation for every belief sounds very dodgy, and collapses under close examination.

3.) What do you see as the major advantages/disadvantages of the bottom-up and top-down approaches to Artificial Intelligence?
I am a philosopher, so I don't want to air speculations (at least in public) about the methodology of disciplines other than philosophy! Philosophers have done this in the past and have ended up with egg on their faces. But nonetheless my views about the mind - that mental properties are 'emergent properties' of complex objects - would suggest that one would get further with a top-down approach, at least in the following sense: first figure out what the system is supposed to be doing, at a global level, and then try and figure out how you could put together something which did that.

On the other hand, I believe in a kind of dependence of the mind on the body which philosophers call 'supervenience' (this means: no mental difference without a physical difference). This implies that two creatures with exactly the same bodies would have the same minds. And it follows from this that if one could create a replica of someone's brain and body, that would have the same mental states (at least for an instant) as the person being replicated. In this sense, then, a bottom-up approach of a certain kind may work! But this appeals to the kind of remote 'logical' possibilities loved by philosophers, which are of little interest to the practising scientists of AI. This sort of replica building is not, obviously, a practical possibility.

4.) Do you have any predictions on the future of Artificial Intelligence - do you think it will succeed in creating understanding, emotional and responsive computers/robots?
Let me begin my answer to this question by reminding you what I said at the beginning of my answer to the last one! However, I must say that I am somewhat sceptical about the grander claims made on behalf of AI. 'Sceptical' here just means that I withold belief, and wait for the results. Let's see what they eventually come up with. In one way, this is just a result of a generally sceptical attitude: I am sceptical too about claims made on behalf of other 'gee-whiz' sciences. For example, that a theory which unifies the physical forces should be called a 'theory of everything', that physics can ('in principle') explain everything, that the gene contains a complete 'programme' to determine a person's life and character, that MRI technology can tell us what a person is thinking and so on. Of course these theories contain some of the greatest achievements of our culture; but it's important to understand precisely what these achievements are, since they are, after all, very precise achievements in the most precise of disciplines.

But independently of this general scepticism, I do also think that there are specific challenges to AI, in particular those made by Drefyfus in 'What Computers Can't Do', which have yet to be adequately answered by the AI theorists. For the critics of AI, AI can seem like a moving target. To its defenders, this constant shifting of the theoretical ground signifies progress.

5.) What would be your definition of understanding?
I have a definition, though I doubt whether it will be helpful in the debate about AI. The definition is this: to understand something is to know what it means, to know what its significance is. To understand a sentence is to know what it means, to understand a language is to know what its words and sentences mean, to understand a phenomenon more generally is to know its significance. What 'knowing significance' amounts to may be very different in different contexts: thus understanding a piece of music requires different things of us than understanding a sentence in a language we are learning, for instance. It would be useful, then, for theorists to look at the different kinds of understanding that there are, and examine them in detail and without prejudice, rather than looking for the essence of understanding.

6.) Searle's takes a very extreme stance on 'computer understanding' with his 'Chinese Room' analogy. What are your viewpoints on his analogy? Why?
Unlike some philosophers, I think that Searle is basically right in the conclusion of the Chinese Room argument: that syntax is not sufficient for semantics. The analogy itself involves some messy disanalogies: for instance, why should it be that the room does not understand because part of the room (i.e. the man in the room) does not? Why should the point about syntax and semantics be illustrated by the complex mental phenomenon of understanding as opposed to just the idea of symbols having meaning? After all, there is no question that the Chinese symbols have meaning; Searle's point is that one could not understand their meaning by manipulating them in terms of purely syntactic rules. But why should AI be committed to this last claim? Why shouldn't AI just say that the symbols have their meaning because they are manipulated? What it takes to understand them may be another matter.

So there are lots of questions about the actual cogency of the argument, many of which Searle has ably rebutted in responses to criticisms. But the essential point of the argument is, I think, correct, whatever the quality of the argumentative steps used in getting there: that no amount of syntactic transformations of meaningless strings of symbols will give them meaning. That has to come from somewhere else.

My view on how this conclusion bears on the AI project is expressed in Chapter 3 of The Mechanical Mind.

7.) Do you ever think philosophy will be of great help to Artificial Intelligence in reaching its goal? Or do you believe that the opposite will be true?
My impression is that a lot of AI is philosophy, in the sense that it is speculation fuelled by a general picture of how things are, or how things must be. Any new scientific discipline must have a philosophy of some kind behind it, and I think that there is not a sharp boundary between such a subject's own philosophy and philosophy as more traditionally conceived. This said, I do think traditional philosophy has had a useful role in raising conceptual and foundational issues for AI. If the extreme critics are right, then philosophy will not have frustrated the goal of AI, but it may have explained why that goal was unobtainable.

8.) It is often said that everything was in philosophy at some point until it is solved (eg, motion of the planets, composition of minerals etc). Do you think the same thing is happening with Artificial Intelligence (eg, Deep Blue and the problem of chess)?
As I said in response to question 7, I do think that emerging sciences do have philosophies, they need to have them, and the same is true of AI. As for chess, I think that the relevance of chess-playing machines (brilliant as they are) to the success or failure of AI has been greatly over-stated. After all, we should not be surprised if computers are good at chess, just as we are not surprised if they are good at calculation!

What would be surprising would be if a machine could play a simple child's game, like peek-a-boo. That would be really something. But why do we find it so difficult to conceive?

The Mechanical Mind - review of the book Tim Crane wrote.

Does the Bottom-Up or Top-Down Approach Closely Model the Human Brain? - The essay I wrote inspired mainly by the book.