Path: utzoo!dciem!nrcaer!scs!spl1!laidbak!att!chinet!mcdchg!clyde!watmath!isishq!doug From: doug@isishq.UUCP (Doug Thompson) Newsgroups: comp.society.futures Subject: Re: AI and people Message-ID: <52.22A80968@isishq.UUCP> Date: 4 Jun 88 13:34:27 GMT Article-I.D.: isishq.52.22A80968 Organization: FidoNet node 221/162 - ISIS International, Waterloo ON Lines: 124 >From: jimmyz%oak.dnet@VLSI2.EE.UFL.EDU (Anubis The Psychic Chaos > >I don't think anyone is really attempting to build a computer >that is just >like a human. Rather, AI researchers are trying to UNDERSTAND >the mind by >building models of it. Models do not replace, or duplicate what >they model. Actually, while that is probably true for some AI researchers, if you believe what others are saying and writing, they *are* trying to make a machine that can do anything a human can do. The same goes for trying to "understand the mind". Some AI researchers are trying to do that, but for the most part AI has assumed it knew enough about the mind (at least in principle) to build a mechanical duplicate. If AI research is testing the hypothesis that AI folk understood the mind well enough to model it mechanically, then the results are in, AI folk failed the test. Look at chess programs, one of the classic AI projects. A chess program needs to mimic human thought in a very narrow and precise application, where a whole lot of mathematics and rules apply. Ideal for AI. Chess program research began by interviewing many chess players to find out just how the human mind did approach chess problems. The results of that line of research were dismal. No one was able to come up with much on how the human mind tackles chess problems. So the approach was changed, "never mind how the human mind does it, can we make a computer that will do as well using any technique at all?". Programs were written which actually appraise millions of possible moves and assign each a numerical value before choosing one. Such programs win a lot of chess games against humans. One of the reasons they do well is that they are (when well written) incapable of making catastrophic mistakes or overlooking obvious traps. The "best" chess programs AI has produced are very un-human in their operation, they solve problems by simply crunching numbers, unlike humans. They don't make mistakes, unlike humans, and they cannot be distracted so easily, unlike humans. If AI people wanted to understand the mind, why aren't they in philosophy? There is certainly more work being done in understanding the human mind there. >They are used to study what they model. A lot of advances of >computer software >engineering have been from AI research, primarily in those programs >that are >able to simulate a certain aspect of intelligence, such as inductive >reasoning. I quite agree that AI has infused a lot of new ideas into computer science, and the only idea I take exception to is the one that says you can duplicate human thought. Heck, it 's incredibly hard for one human to duplicate the thought processes of another, let alone get a machine to do that!! > >As for religion, spritualness, and other meta-physical "science," >that is >a matter of personal beliefe. Different people feel/thing different >things >about the universe and their place in it. But we can not assume >they are all >true. >THat is not the way to approach a problem scientifically. Of course not. However, an awful lot of human problems cannot be fully understood or adequately addressed "scientifically" because science is able to do little more than categorize people's "feelings" and "beliefs" about the universe and their place in it. Scientific knowledge is only one kind of knowledge. It is not all knowledge, and it is almost always relative knowledge, more a capacity to describe and predict natural phenomenon which are sensible and predictable. Science can do nothing much with phenomenon which are neither sensible nor predictable. People who "believe" in science tend to disregard phenomenon which are neither sensible nor predictable because science cannot deal with such things (or at least we haven't found a way yet). However a great deal of that which is most meaningful to most people is really quite unscientific. It is important to realize that science may never be able to say very much at all about "the meaning of life", or "the meaning of love" beyond describing what people say and do about it. I think the biggest mistake that laymen and scientists tend to make is that of overrating the scope and applicability of scientific method and scientific results. Science doesn't really have much to say about "the meaning of life", although every scientist has something to say. Science can hardly even ask the question "Does God exist?" and must immediately respond that whatever answer there may to that question is outside the realm of science. Yet, too often when we say "that problem is not scientific" we imply "that problem is not important". In an earlier age Theology was the queen of the sciences! The knowledge of theologians (or at least the questions they addressed) were considered of ultimate significance. Theology has been replaced by Physics, surely, and "computer science" and mathematics may take over the pre-eminent role in the public mind. Our questions may be thought of as the only important questions. We will have our heads inflated by the crowds that come to us with profound questions as to how to keep their cybernetic slaves in line - - - The public will forget, but we must not. The computer will never answer all questions, and it may not even be able to understand the most important of questions. There is more to human life than any of our intellectual artifacts or disciplines can ever reflect, and any discipline is only a way of representing knowledge, it is not knowledge per se. ------------------------------------------------------------------------ Fido 1:221/162 -- 1:221/0 280 Phillip St., UUCP: !watmath!isishq!doug Unit B-3-11 Waterloo, Ontario Bitnet: fido@water Canada N2L 3X1 Internet: doug@isishq.math.waterloo.edu (519) 746-5022 ------------------------------------------------------------------------