Hieronder vindt u een artikel van mij, in de vorm van een dialoog, die in 1995 is gepubliceerd in Kybernetes. Het is niet online beschikbaar, dus ik ben zo vrij het hier te publiceren:
METALOGUE: WHY ASK QUESTIONS?
A discussion on the concept of artificial intelligence
Joop Kielema
Daughter: Daddy, why do people ask questions?
Father: Hmm, you are asking me something you can only answer yourself.
D: Why is that?
F: Now you are asking me for the second time why people ask questions.
D: That’s because I want an answer.
F: You shouldn’t expect an answer from me. You have just answered the question yourself.
D: But things don’t work that way, do they? You cannot ask a question and answer it at the same time. In that case I would not have asked. And I don’t believe that each question contains an answer, do you?
F: Yes and no.
D: Now you don’t answer my question again. First you said I already knew the answer and then you gave an answer that has a contradiction in it. So you still haven’t answered my question.
F: Oh yes, I did.
D: Oh come on, don’t be silly.
F: All right, let’s start again. You ask me why people ask questions and by asking that question you already gave the answer.
D: OK, that might be the answer to my question, but you cannot answer all questions in that way?
F: You are right about that, that is why the second time I answered yes and no.
D: Oh.
F: You might say that every question contains the key to its answer.
D: I see. So when I asked you: “why do people ask questions?” you knew I had the key to the answer.
F: Exactly.
D: And I only needed you to find out that I had the key. But, if this is true, each question carries its own answer.
F: No.
D: Now I don’t get it any longer. You just said I was right and now you are telling me that I am wrong. Are you trying to confuse me, or did I ask something stupid?
F: No, it was not stupid and I don’t want to confuse you. When I said you were right I was quite serious. But remember I answered the second time with yes and no.
D: But you cannot answer a question in the affirmative and the negative?
F: Of course not.
D: But you just did.
F: The answer is again yes and no.
D: Now you are doing it again!
F: I did not. I haven’t answered your question completely. You asked me whether each question carries the key to its own answer and that is right. The point is that the key sometimes leads to the conclusion that there is no answer.
D: That’s again such a strange contradiction. The answer to the question is: there is no answer. But by posing that there is no answer you have in fact answered the question. Isn’t it better to reformulate the answer and say that to some questions we know the answer and to some questions we don’t?
F: No, that is something completely different. We answer any question. However, we never know the value of the particular answer. Everyone can give a different answer, without giving a wrong answer.
D: So every answer I give to the question posed by myself is right. In that case there is no need to ask them. That doesn’t sound very plausible, I mean...some people know more about certain things than others, that’s why you ask them questions.
F: You are right about that, some people know more about certain things than others and therefore it is good to ask them questions. However, we never know which answer is the best.
D: I really don’t get it, it is so confusing. The answer to some questions is that there is no answer and all answers are in one way equally valid, but in another they are not.
F: Well, yes...a fool can ask more questions than a thousand wise men can answer. Let us take an example that cannot be answered. Marvin Minski once put the question (and with him many others): “Are men machines?” They tried to find the answer by building machines that acted like a human being. They tried to construct machines that could pass the Turing test.
D: What is that the Turing test?
F: Alan Turing is one of the founders of modern computer technology. In 1936 he laid the mathematical foundation for the development of the computer. Later, when the first computers were build Turing asked himself whether computers could think. And he constructed a test he believed that, if a computer could pass the test, it could be concluded that computers can think. Turing named the test the imitation game.
In this game a man and a woman, sitting in separate rooms, are being interrogated, via a telex connection, by a third party. The third party can address questions to either room but has no idea which person is in which room. The object of the game is to determine which room the woman is in. . The woman must help the interrogator as much as possible through her answers, whereas the man must mislead the interrogator as much as possible. He does so by responding as he thinks a woman might. Turing was wondering what would happen if a computer would take over the role of the woman. He was convinced that within fifty years there would be a computer that could take the woman’s place with such good results that 70 percent of the people would not notice they are communicating with a machine instead of a human being.
D: Is there already a computer that can pass the Turing test?
F: No, but according to some people it is only a matter of time.
D: And then computers can think and the question of Minski is answered.
F: It is not that simple, remember I would give you an example of a question that cannot be answered.
D: Yes, but why are some people convinced about this particular answer whereas you claim there is no answer.
F: That is because their basic assumptions are not properly formulated. By explaining these basic assumptions I will explain to you the difference between questions that can be answered in principal and questions that cannot. Turing made a mistake in logical typing
D: What is logical typing?
F: Logical type is bringing a hierarchic structure in learning processes. Let me try to explain it through the ‘imitation game’. The theory of logical types states that the structures we create in processes are of a higher logical type then the separate processes. This means that the individual processes of the ‘imitation game’ (the interactions between the persons in the rooms and the interrogator) are of a lower logical type then the structure we recognize in the processes. The communication between a human and a computer is regarded as human if a human being doesn't realize computer is on the other side instead of a human being. The concept of communication is, however, of a higher logical type. It is the name we give to an interaction, for instance between humans. But of course we can also assign this name to the interaction between man and computer. However, we cannot draw the conclusion that computers can think because two logical types are confused. The name of the structure (communication) becomes an attribute of one of the separate processes of the actors involved (thinking).
D: So the question: “can machines pass the Turing test?” can be answered. The answer is yes, but this doesn’t mean that computers can think.
F: No, things are even more complex. We still haven’t answered the question: “Are men machines?” As we now know the question must contain the key to the answer. If men are machines (and not resemble machines) we must look for evidence that proves that we can build machines that are in every aspect as humans. But the problem is that these aspects are limitless. There is never an absolute end to this search. This makes the question as well as its answer fruitless. A question is only fruitful if criteria are given to which the question can be answered. If there are no criteria the question is needless, there is no key.
D: And does the question: “can a computer pass the Turing test?” contain such a key?
F: Hmm... the point is that even the Turing test does not contain such a key. You can ask the machine everything. Turing didn't realize that both speech acts and thinking (or maybe it is better to say the creation of ideas) are of the same logical type. They are of a higher logical type than the personal experience. We can say how we feel, but our speech act is never similar to the experience itself. By means of language we can describe an endless amount of feelings, experiences and events. This can be varied in the same way as we can vary the criteria for what is human. So the question: “Can a computer pass the Turing test?” is in fact the same question as “Can machines think?”.
D: But this means that science is not only about looking for facts but also about asking questions. Since you know, without any empirical evidence that a question like “ Are men machines?” cannot be answered, this also implies that there are questions that, eventually, can be answered, although you don’t even know the answer.
F: That’s right, as long as you can put the question in a context.
D: You mean, when I ask you: “how much is 2 plus 2?” you will say 4, because mathematics is the key to the answer.
F: Exactly.
D: But why are people asking questions that can never be answered?
F: And I ask you why you asked me why people ask questions.
ELUCIDATION METALOGUE
A metalogue is a figure of speech introduced by Gregory Bateson. He defines a metalogue as: “a conversation about some problematic subject. This conversation should be such that not only do the participants discuss the problem that the structure of the conversation as a whole is also relevant to the same subject.” [1]. In Bateson’s metalogues the dialogue is always between a father and his daughter. To make the reference towards Bateson stronger this metalogue is also a conversation between father and daughter.
In this metalogue several theories are discussed, especially the principal of logical typing that Bateson adopted from the Principia Mathematica of Whitehead and Russel (1910-13) and the Imitation game from Alan Turing. Both concepts are explained in the metalogue. My main reason for writing this metalogue and my point of view on artificial intelligence are discussed in the metalogue, so I will give here no further comments on the subject.
The conversation contains a few reference to literature. I made no explicit reference within the text. There are, however, some ideas within the dialogue from people that are not mentioned. Furthermore the explication of the imitation game is nearly an exact quote from Douglas Hofstadter [2]. Instead of making explicit references I will give a list of literature containing both the explicit and implicit references.
[1] Bateson, G. Steps to an Ecology of mind Collected essays in anthropology psychiatry evolution and epistemology., Jason Aronson Northvale, 1972.
[2] Hofstadter, D.R. “Metamagical Themas: A coffeehouse conversation on the Turing test to determine if a machine can think.” Scientific American may 1981: 15-36
[3] Bateson, G., Mind and nature A necessary unity. Bantam Books 1979.
[4] Stachowiak, H., “Erkenntnisstufen zum Systematischen Neopragmatismus und zu allgemeinen Modelltheorie” in: Stachowiak, H. (Hrsg), Modelle–Konstruktion der Wirklichkeit. W. Fink, München, 1983.
[5] Minski, M. The society of mind. Simon & Schuster, New York, 1985.
[6] Turing, A.M. “Computing machinery and intelligence” Mind Oct. 1950: 433-60 reprinted in: Boden, M. (ed.) The philosophy of artificial intelligence. Oxford University Press 1990: 40-66
Published in: Kybernetes, 1995, 24, (5), 58 - 62.
