Stork: Some people reject even the possibility of AI on religious grounds. Would you like to address that?
Dennett: I am fascinated by the resistance in so many quarters today to artificial intelligence on the grounds that it some how is reductionist and violates our sense of what a human being really is, of maybe having a soul, and this is essentially a religious objection. Well, other people who are religious scoff at this, and say, no, no, any sophisticated doctrine of the soul is quite consistent with what science teaches us. Well, I am not so sure that is true. I think that what we have come to realize in the 20th century, and now the 21st century, is that what a human being is, is a collection of some trillions of little tiny robots. They are called cells. They are not conscious. They do not even know you exist. They are several trillion cells that make up your body and no one of them knows who you are or cares, and yet you can put together that trillion member team of specialists and when it has the right organization it is a living, breathing, conscious, thinking, responsible, human person. Now that is a remarkable fact. And it is a fact. I think the doubts about that are eroding on every side. Coming to grips with that fact are not easy, and I can appreciate the many people who have deep religious convictions just refuse to accept it, but the evidence is pouring in.

Stork: What do you think or expect will be the first ethical problems that person on the street will ever have to confront related to artificial intelligence?
Dennett: I think that when futurologist try to imagine what the problems are, they almost always get them wrong, so anything that I say about the moral problems to be faced in the future will be I probably miss comically, but I will have a stab at it anyway. . . I think the main problems are going to be questions of who do you trust. When there is an expert system, product of artificial intelligence that consults its vast stores of information and says where going to remove this person's kidney and replace it with another kidney, or make some other truly life threatening decision, and the human being who has to endorse, sign off on this will wonder, what I am doing here? Am I just a rubber stamp? Do I have the moral stamina, the guts to say, that's what the machine says, but I say otherwise. I think we are on the edge of many, many situations in the world where we are going to have to face the fact that putting a human being there, as the final arbiter, the person that has to sign off, becomes more and more of a formality, becomes less and less important. We are going to start seeding the true moral decision making to some of these systems, and I think we haven't begun to come to grips to what that means as our own role as moral agents.

COG isn't conscious now. COG isn't close to be conscious. But COG has a lot of the pieces in place that necessary for consciousness. For instance, COG's eyes can't really see yet, but COG's eyes already have to solve the problem as to what to look at next, and that is which way should they dart, and in order to solve that problem you have to have something driving that, something like curiosity. You have to have a built in process that which gets restless with familiarity, that wants to move on, and another process that is a little bit, as it were, afraid of the novel and likes things that are familiar and recognizable. This competition between curiosity and security is already built in at the foundations, even though COG isn't conscious yet. Now this is just one example of many different control issues that have to be faced before you can begin dreaming about building a robot that actually impresses us of as conscious. Already COG can fool people. When COG looks into your eyes or tracks you as you walk across the room, it's a very eerie feeling. There is nobody home in there yet, but some of the furniture is being moved in, as we speak.

MOC: How would you define intelligence and do you think any of the current AI systems exhibit signs of it?
Dennett: Intelligence is one of those terms like health, or beauty, or life, itself which defies any simple explanation or definition, but I would say that intelligence is the capacity to improve your prospects in real time by taking information from the world and mining it for expectations. So we use our capacity to pull information out of the world to improve our expectations of what is going to happen in the near future then we act on those expectations. The task of making an intelligent robot is tremendously expensive and time consuming and would cost billions of dollars to do. Is there any reason to think that we will do it? I think there is every reason to think we won't do it. Compare it to the project of making a robotic bird that weighed less than a pound, that could catch insects on the fly, and land perched on a little twig. Is it possible in principle? Sure. I don't think there is anything miraculous or mysterious about birds. Is anybody going to do it? No. It would cost more than the moon program. It would cost a colossal amounts of money and we wouldn't get much pay off from it. It you want to study about birds, you study aeronautics, you study birds themselves, and you build models of different features. You don't have to make a robotic bird that can do what every bird can do. And, I don't think the same thing is true, more so, to human intelligence. We will learn about human intelligence, human consciousness, by making partial models, models that cost a thousandth or a millionth of what the full model would cost, and once we got that information, there won't be any justification for making the real thing.

MOC: So you don't think that having such a truly intelligent system would be useful for searching the web, helping you research the world?
Dennett: I have encountered one reason which might actually justify making a human intelligence and that is we now have an international web of telecommunications, the internet and the phone system and so forth, which is simply so big and so complicated that nobody can understand it. Nobody. And, no team of people can understand it. And, this leads the team of people, who are the custodians of this, people like AT&T and British Telecom and so forth, to worry about how they will ever get control of the very thing that they are responsible for. One idea is to make a humanoid intelligence, to make a sort of COG-like intelligence, and then release it from its body, put it on the web, as sort of like a Max Headroom investigator, to go around and find out what is actually going on, on this huge artifactual construction and come back and tell us, and, if it doesn't have humanoid intelligence we won't understand what it tell us. SO I suppose there is one scenario where real artificial intelligence would pay for itself, and that is, if we need super human intelligence, but it has to be humanoid enough so that we can understand it, then it might really pay to make an intelligent robot, and I suppose in a way, Hal anticipated that sort of need, so in that regard it is still fairly realistic.

MOC: You make it seem as it's the AT&T's and British Telecom's that are interested in this, but what about the average user? "I want a good web search engine."
Dennett: Yes, I suppose, that as we drown in the flood of information that's all around us, our need to have reliable filters, our need to protect ourselves from this flood will grow at the same rate or even faster than the flood of information itself, and probably the only filter that would do a good job would be one which is a real substitute for a human being and then some. So this is an area where there will be competition for better and better filters and so we may create a sort of truncated special purpose artificial intelligence which is simply an information filter tuned to individual needs.

MOC: Do you see this as a Pandora's Box, and should we be opening this Pandora's box of creating artifical intelligence?
Dennett: Here's a little fantasy to scare you. A virus lands on the planet and begins to take over and grow and grow and grow, and it need us to provide it with energy and repair, but it provides us with lots of services that we like, and it gradually makes us more and more dependent on it, to the point where we realize that many of the things that we hold most dear we can no longer accomplish without feeding this virus and helping it grow until it finally takes over the planet. Now that's a pretty scary scenario, and it's not true, in one regard, in that the virus did not come from outer space. It's called the information superstructure of the world or the web, internet, and it's already here, and we are already dependent on it. So, it's too late to go back. We're going to have to maintain whatever control we can over it and that control is already beginning to slip.

MOC: What are the ethical issues of creating sentient beings with intelligence and emotions and free will? Do we necessarily have to owe them the same rights that we have?
Dennett: No. It is an open empirical question whether we can make really intelligent filters that are nevertheless so truncated in their intelligence that they don't have to be given rights. I think almost certainly we can't. That if they are tuned enough to our needs and desires, if they are enough like us to do the jobs that we will need doing, they will also be enough like us, and we will see they are enough like us, that they themselves have rights. That they must not be mistreated; That we have enlarged the moral-sphere by creating them. I think that we will either only succeed in making relatively ineffective household appliances or we will make being that have rights that are comparable to our own.