Stork: How many of the ideas came from you? How many came from him? What was the working relationship on the script?
Pinker: What were your impression of Hal?

Stork: What were your impression of Hal?
Pinker: Hal was one of the most interesting characters, the one with the most emotion in his voice, and the one you could relate to more than the monolith for example.

Stork: Tell us about some of the problems in cognitive science, and generally how that is influenced by or influences AI research?
Pinker: Cognitive science has always had a close relationship to artificial intelligence because: how we do even know what a theory of seeing, or a theory of thinking, or a theory of moving would even look like? One way is to figure out how to build a machine to do it and then at least we would know what it takes, what's the minimum that you could get away with to get an intelligent machine. It doesn't mean that what you build out of silicon and relays does it the way the brain would do it, but if you have a bunch of ways of doing it, yourself, then at least you have a hypothesis you can test about how the brain may do it, taking into account the huge differences between the silicon and brains. But just as in other sciences where you learn a lot by comparing man-made artifacts to natural systems: the eyes and the camera, the wing of an airplane and the wing of a bird, there are subject to the same laws of physics even if they work slightly differently. The hope is that you can get some insight about how the human brain works by seeing what it takes to have an artificial device that can do the same things that the brain does.

Stork: Can you be even more specific in the problem of language understanding?
Pinker: Take language for example. How does the brain, first of all, makes sense of the sound coming into the ears so that you know what words the other guy said, and once you have the words how do you decode the order in which they appeared to know whom did what to whom? We have no real intuitions about that can happen. We can't look at other animals. One way we can make sense of it is to say well what would we have to put into a computer program to get it to understand language, to figure out whom did what to whom, from that squiggly wave of sound coming in through a microphone. And, that gives us hypothesis we can test. Let's say there are two ways to build a machine that would understand language. We can then say well which of those two ways is the one that human beings use? It gives us hypothesis, it gives us an experimental research program, it gives questions that we can then answer.
For example, as soon as you try to build a machine that tries to understand language you discover this problem, namely, that in speech, there aren't these little pauses between the words the way there are white spaces on a printed page.
"People-don't-talk-like-this-with-a-pause-in-between-each-word". You only discover that when you try to build a machine to try to understand language because our brains work so well that they give us the illusion that we actually hear where one word ends and the other word begins but its nowhere in the actual sound. And you don't even realize what a fantastic thing the brain does until you stumble on this very difficult problem of building a machine to do the same thing. And, it's the same with something like recognizing a face, understanding the meaning of a sentence, picking up a glass. We do them so effortlessly that we think they are easy problems. Well what's the big deal with picking up a glass? I want the glass, I reach over and pick it up, and I don't drop it, I don't crush it. I bump into someone I recognize them from very subtle differences in how close their eyes are together, how thick their lips are, but it works so well, that it doesn't occur to me, here inside my brain, that the brain is doing this amazing process. But, now start out with a blank computer screen. Now design a computer program that's got to understand speech, or pick up the glass, or recognize a face, and you begin to get an appreciation for the amazing thing that the brain does without even realizing it.

Stork: Would you like to discuss some of the issues of apparently the higher level problems such as playing chess, that seem to succumb to AI techniques, where as these ones like recognizing a face and segmenting words from sentences haven't?
Pinker:In artificial intelligence they sometimes say that the hard problems are easy and that the easy problems are hard. That is the ones that we find hard, like playing chess or solving equations or predicting the stock market, are things that are not so hard to get a computer to perform at a human level. For the things that any four year old can do, like understand a command to walk over and pick up a book off the shelf, or to put away dishes from the dishwasher, or to run simple errands, got to the grocery store and pick up a few items. Those are easy, but those are the ones where we're not going to see computers doing that for quite sometime, even after computers can beat the world chess champion.

Stork: Tell us about the different ways in which the brain came to be compared to how machines came to be, development, evolution and so forth as opposed to programming, and what that might mean.
Pinker: There are two big differences between how the brain came to be and how a computer came to be. The first is that the brain evolved by natural selection, there wasn't a conscious designer who figured out what the brain was trying to accomplish, but instead it evolved by a process that simply picked the most successful replicator over millions of generations of replication. So the brain is really a device, if it wasn't literally designed, it has the illusion of having been designed for the purpose of survival and reproduction in the kind of environment in which humans evolved, namely the tribal or hunting and gathering environment that prevailed throughout human evolution. A computer, on the other hand, is designed to send out bills, or compute digits of Pi, or generate graphics on a screen, it's at the whim of a computer designer. The other big difference is that a given brain grows out of a single cell. It comes about division of the cell and subdivision and specialization of differentiation and some how, one of the great mysteries of science is: how these millions of know where to go in the brain, know how to wire themselves up with no supervisor, no one actually assembling it. Each cell just growing this way versus that way, and yet the whole thing manages to function coherently. For a computer, you have an assembly line, an intelligent engineer designed it, someone puts the chips into place, and when the whole thing is ready you plug it in, turn it on, and it works. So, they are designed by different processes, and they are assembled by different processes.

Stork: We've been discussing the hardware, the brain itself, and the machine; talk about mind and maybe software. The difference between brain and mind.
Pinker: Most brain scientist say the mind is what the brain does, or more specifically , the mind is the information processing activity of the brain, because the brain does lots of other things too. It metabolizes fat and gives off heat. But, of all the things the brain does, the one that is interesting in terms of psychology is the information processing, the aggregation of signals, comparing them against thresholds, following decision rules, storing and retrieving information, and many of the things that a computer does. So some of the activities that are familiar to us from the realm of computers that tell us what look for in the activity of the brain to understand thinking.

Stork: Tell us about the flexibility of the brain and mind compared to the flexibility of machines.
Pinker: Well the brain has a number of problems that the computer doesn't have to face. The brain has to control a body that is growing the whole time, the brain doesn't know if it is going to be in a big body or a small body, or a strong body or a weak body. It doesn't know if it is going to be placed in an environment that got has lots of food or barely any, with enemies mixed or not. So the brain is forced to learn all of the aspects of a problem that is going to have to solve during the life of a person. Where as for a computer, computers are sold for particular purposes, software is written for particular purposes, and its known beforehand the kind of world in which it is going to be used.

Stork: Can you take us back to the time of the release of 2001, 1968, and tell us about any developments in theories of mind or some of the most important milestones up to 2001 the year?
Pinker: Since 1968 there have been at least two big changes in the way we think about the human mind compared to the old general idea that the mind was kind of a general purpose, AI, program. One of them is that the mind has a lot of parts. There isn't going to be one general problem solver that gives us the answer to everything we are interested in figuring out. The brain appears to have different systems for reasoning about other human beings, reasoning about other physical objects, reasoning about space, reasoning about number. It seems to be somewhat modular, and the original plan of the magic algorithm that is going to do everything, I think, is unlikely to work out. Also, it's unlikely that the brain evolved to have one single, all-purpose algorithm for the same reason the body doesn't have just one organ, because you need specialization just as the body has to do different things, pump blood, digest food, the brain has to do different things, it has to be able to find your way, you have to be able to make friends and attract mates, you got to be able to fashion tools out of rocks and sticks, and it's unlikely that a system that's good at solving one of those problems is going to be good at solving all the others. The other thing is that the brain of made of different stuff than computers. It's massively parallel. It's build out of billions of units that are unpredictable and noisy and statistical in their operation It grows and grows connections rather than being assembled one little chip at a time. All of those are going to have an effect on the kind of computations that a brain can do as opposed to what a computer can do. And, those two things have changed our understanding of the brain.

Stork: Now you mentioned different modules for attracting mates and vision and so forth, certain of those separations are obvious, but what about individual functions? How do we know that we have separate modules for attracting a mate or for any of these modular problems?
Pinker: We believe that the brain is divided into subsystems, sometimes because we can actually see them. There are different parts of the brain, different organs, we know that the organ the hippocampus seems to be crucial for mental maps, for example. Sometimes we know it because if we try to devise algorithms that do what the brain does you need different algorithms for acquiring language versus acquiring a knowledge of space versus acquiring a knowledge of other people. In some cases we can tell from disease or pathology where a person might be missing one kind of intelligence but not being retarded across the board. An example is autism where people seem to lack the part of mind that is good at figuring out other people and what other people are thinking. To an autistic child every other person is a robot, just a machine without a mind. That suggests that the ability to attribute thoughts and feelings to other people might be a special part of the mind that didn't develop properly in this population.

Stork: Is this sense of self just another one of these modules?
Pinker: Even the sense of self may be a product of a specialization of the brain. We don't actually have a perfectly accurate knowledge of how we are. We kind of have a P.R. job that we use to package and sell ourselves to others, and we usually believe our own stories about ourselves to better convince others of it. Most of us believe that we are above average in any positive trait you mention: driving skill, intelligence, kindness, generosity, and so on, and of course, not everyone can be above average. We tell ourselves lies to make our stories consistent, a process known as "cognitive dissonance", where we twist around the truth to make it seem like we are rational and generous. So it's possible that even the sense of self isn't a little man standing back and examining the rest of the person, but rather a system of the brain that works in certain ways.