ARaman@massey.ac.nz (A.V. Raman) (10/09/89)
Topic: Knowledge Representation
Aim: To find a system that represents knowledge as efficiently
as the human brain.
Introduction
~~~~~~~~~~~~
There is a widespread notion that the human brain is the most
intelligent system at present known to us. This is a direct
consequence of the fact that the human brain is also the most
efficient knowledge base known. In this article, we analyze the
above proposition objectively to find out whether the human
brain is really the most efficient knowledge base known, and
whether any other intelligent system, with at least as efficient a
knowledge base already exists, and if so, is known to man.
1) Another look at Information:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
What exactly is this thing called information that is so essentially
a part of knowledge representation? In what way can a data structure
not be called information? Simply said, information is structure.
We do not classify information as useful structure, because this is
not a discussion on philosophy.
Any structure conveys information of some kind or other. In what way
is a sequence of bytes in a file or RAM different from, say, a stack
of books piled up on the table? If we so decided, we could use the
stack of books to convey the same information that the sequence of
bytes conveys.
There is absolutely no way in which we can say that any structure
whatsoever, is information, that can be manipulated.
2) Another look at thought processes:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
What is the difference between a thought process as it is now known
and any operation that changes a data structure? Why do we say that
a book falling of the top of a stack is not a thought process, whereas
a rearrangement of bytes in a theoretical expert system is? What
makes us decide that free-relating objects or discrete data structures
is not the same as any 'random' event occurring on earth today.
What makes us think that discovering a new mathematical law is any
different from from the creation of zinc sulphate when zinc and
sulphuric acid are mixed? Neither existed before the 'event'. Each
depended on its predecessors. In the first case it was a set of
mathematical laws which led to the derivation of the next. In the
second case it was a set of chemicals. In either case an 'event' was
involved that resulted in the transformation. What makes us distinguish
these two events other than the fact that one 'creation' is a tangible
data structure and the other is not.
Any structure conveys information and any operation that changes a
structure is a thought process.
2.1) A thought experiment:
~~~~~~~~~~~~~~~~~~~~~~~~~~
If a man was reduced to the size of a photon and put inside the
human brain, would he be capable of deducing that his surroundings are
capable of thinking?
Would he observe the electrons flowing from one part of the brain to
another as a thought process?
Would he deduce that the structured arrangement of the molecules around
him are nothing but information useful to the body to which the brain
belongs?
Would he observe the external stimuli being passed on to the brain as
information inputs being sent to the processing unit.
Would he observe a lone electron falling through a neural network as
an attempt by the brain to rectify a faulty assumption?
Or would he try to discover laws to explain the phenomenon around
him, coin a word called "random" to account for the inexplicable,
and invent such a thing as the 'Nature' now known?
I'm not saying that there exists Omniscience around us. Further away
am I from saying that all intelligence is illusory. What I am saying,
though, is that haven't we given Nature a pretty raw deal in
our study of Knowledge Representations? Maybe we might benefit, at least
to a small extent, by studying how Nature itself stores information,
not within us but around us, and how Nature manipulates information
in our daily lives.
If what we call creativity is a measuring stone to intelligence,
hasn't Nature ever been as creative as we have. Hasn't it learnt
from mistakes, just as we do. Or maybe this article itself is one of
Nature's mistakes.
- &/..ok@cs.mu.oz.au (Richard O'Keefe) (10/10/89)
In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > What makes us think that discovering a new mathematical law is any > different from from the creation of zinc sulphate when zinc and > sulphuric acid are mixed? I'm no fan of sociology, but one of the insights which I think is valid is that categories like "knowledge" and "discovery" are socially constructed. For example, we distinguish between the first person to come up with a law (that's "discovery") and later people who go through the same reasoning steps (that's "rediscovery"). Think of the old declension; "I am intelligent, thou art shrewd, he is cunning". > What makes us distinguish > these two events other than the fact that one 'creation' is a tangible > data structure and the other is not. Er, which one is the "tangible data structure"? We non-discovers learn about mathematical laws by attending to symbolic schemes produced by other language-users (data structures); zinc sulphate may be tangible but it's hardly a data structure. There are quite a number of differences. Zinc sulphate can readily be turned back into its original constituents, indeed if they weren't telling my lies in Chemistry 1 it's happening all the time right there in the beaker (or whatever you do your chemistry in). But mathematical laws don't spontaneously decompose into the symbolic structures that went into their production. On the contrary; a long-standing criticism of mathematical writing and teaching is that you are presented with the end results and _don't_ get the original ``components''. Another important difference is that a mathematical law can be translated from one symbolic system to another (marks on paper, speech, hand movements, flashes of light) and still remain ``the same''. Roughly, the point is that zinc sulphate does not originate from a language-user, so it doesn't _mean_ anything. > If a man was reduced to the size of a photon and put inside the > human brain, would he be capable of deducing that his surroundings are > capable of thinking? Photons have a size? New to me. There is an upper bound on information density, and I'd be surprised if a photon could hold more than one or two bits. > Would he observe the electrons flowing from one part of the brain to > another as a thought process? What's the subjective time rate of the observer supposed to be? If it's commensurate with typical subatomic time scales he won't notice anything changing (the polarisation wave of neurons involves ion flow; brain events take place at "chemical" rates rather than "physics" rates). If it's commensurate with the time scales relevant to brain events, he's going to be so confused by his own motion he's not going to observe anything. > Would he deduce that the structured arrangement of the molecules around > him are nothing but information useful to the body to which the brain > belongs? Well, I hope not, because that isn't true. The vast bulk of the structured arrangement of the molecules in the brain is required to maintain the brain cells as living cells. E.g. the cell walls are not information, the blood vessels are not information, the water molecules are not information, ... > Would he observe the external stimuli being passed on to the brain as > information inputs being sent to the processing unit. If he's _inside_ the brain, how is he going to observe external stimuli at all? There are two separate things: X *is* symbolic/information bearing/meaningful X is *perceived as* symbolic/information bearing/meaningful All four possibilities occur. (There's a lovely bit in one of R.A.Lafferty's stories where a character tries to read an orange.) If PhotonMan couldn't perceive information in the brain, so what? A blind man can't perceive information in traffic lights; that doesn't mean they _aren't_ signals. > If what we call creativity is a measuring stone to intelligence, > hasn't Nature ever been as creative as we have. Intelligence and creativity are socially defined categories. The avant-garde of the late nineteenth century identified creativity and degeneracy; claiming that true art had to be "bizarre". IQ tests don't measure creativity. (What have IQ tests got to do with intelligence? Well, they've got "intelligence" in the _name_...) Nature can't be "creative" unless Nature is a language-user; unless the productions of the goddess Natura are capable of _meaning_ something they aren't "creative". > Hasn't it learnt from mistakes, just as we do. No. Unless she can form intentions, Natura cannot make mistakes.
dmocsny@uceng.UC.EDU (daniel mocsny) (10/10/89)
In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > Any structure conveys information and any operation that changes a > structure is a thought process. This reminds me of Wolfram's notion that every computer is a physical system, and every physical system is a computer. Some physical systems are computationally inefficient, meaning that a simpler computer can adequately mimic their behavior. Since we are currently a bit squeezed for computational power, the only physical systems we can profitably study in our diversion called "science" must be computationally inefficient (or, alternatively, very simple). > Would he observe the electrons flowing from one part of the brain to > another as a thought process? One minor quibble: if he did see such a thing (assuming that "part" refers to multicellular collections), then he would be mistaken. Nervous signals do not propagate down axons via bulk flow of electrons, but rather by a cascading reversal of membrane polarity. (Something like a "signal" passing through a row of falling dominoes.) See any elementary text on neurophysiology for details. Dan Mocsny dmocsny@uceng.uc.edu
dmocsny@uceng.UC.EDU (daniel mocsny) (10/10/89)
In article <2376@munnari.oz.au>, ok@cs.mu.oz.au (Richard O'Keefe) writes: > In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > > Hasn't it learnt from mistakes, just as we do. > No. Unless she can form intentions, Natura cannot make mistakes. How interesting to both personify and de-personify Natura in one sentence! Dan Mocsny dmocsny@uceng.uc.edu
kung@mips.COM (Kung Hsu) (10/11/89)
In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > Topic: Knowledge Representation > > Any structure conveys information of some kind or other. In what way > is a sequence of bytes in a file or RAM different from, say, a stack > of books piled up on the table? If we so decided, we could use the > stack of books to convey the same information that the sequence of > bytes conveys. > Just look at the evolution of information systems, we will get some hints. Sequential file, ISAM, Random access file, Relational database, Knowledge representation. e.g. why relational database is better? Because that it finds out that starting from the fundamental thing, *relation* between entities and attributes, we can ask pretty intelligent questions (complex queries). Then, what is the idea behind this AI knowledge representation? I believe what we called *intelligence* can be defined as the ability to induce common rules/ patterns/orders when observing a group of things, be it a group of objects, a group of existing/man-made rules, a group of known knowledges. Sometimes this is called abstraction ability. But intelligence is more than abstraction. We will elaborate this later. Knowledge representation today, as I see it, has the goal of making this intelligent induction easier. I am talking about what is intelligence, I do not know how. But we do believe human brain is the most intelligent system known. A lot research right now is try to understand how its information is organized. > What makes us think that discovering a new mathematical law is any > different from from the creation of zinc sulphate when zinc and > sulphuric acid are mixed? Neither existed before the 'event'. Each > depended on its predecessors. In the first case it was a set of > mathematical laws which led to the derivation of the next. In the > second case it was a set of chemicals. In either case an 'event' was > involved that resulted in the transformation. What makes us distinguish > these two events other than the fact that one 'creation' is a tangible > data structure and the other is not. > There are many different kinds of abstraction involved in discovering a new mathematical law. First of all, facing a group of math laws, you have to form an abstraction about what this math object is about, i.e. what is relavant and what is not relavant. This math object is abstract, some people fails to see this level of abstraction from the laws. Somebody will say you do not need this level of abstraction to discover, just do deductions from existing laws. But application of laws is not exaustive search, the abstraction about the math object, though not necessarily complete, helps to find out the goal. Then, one can use backward reasoning. This comes to another kind of abstraction, from the pure logical level, from the laws proved now, we form an abstraction of how and when to apply different kinds of techniques: forward, backward, couter- proof, induction, etc. These are all very high level and difficult abstractions , that's why we say doing math is intelligent. While compare with later case which require not abstraction at all. Intelligence form some time gets very subtle. There may be some intrinsic functions in the brain that can detect *commons*. Sometimes these commons are in higher level that yet can be described, or may be from a strabge angle that also hard to describe. These may called instinct or inspiration. These happens a lot. That is why chess Grand Master can beat computer chess with limited process power. > > If what we call creativity is a measuring stone to intelligence, > hasn't Nature ever been as creative as we have. Hasn't it learnt > from mistakes, just as we do. Or maybe this article itself is one of > Nature's mistakes. > - &/.. I will try to describe why and what is creativity using the concept of intelligence above. The essence of intelligence is *to form* a rules from a group observations. If once the rules is formed, to understand or to follow the rules are much less intelligent. This why creativity is the measuring stone as suggested. For example, Beatles' music is very creative, in the sense that within the domain of music and human aesthetics, they formed their own music phrase, music expression, music style (rules). That is also why comments like 'True art has to be bizarre' was made. One has to form it's own rules. Also, through the activity to sharpen ones own intelligence. Sounds OK, but there's one thing not defined 'human aesthetics'. What else, 'music aesthetics' is just the regularities in sound frequencies and intervals. This is so fundamental, like langugage, it may be built-in. I am not familiar with this area, but is there any theory describe why Beethoven, Bach or Chopin are better in a more *AI* way. I completely disgree intelligence and creativity are societal. In fact, today there's no doubt that Newton's Dynamics is smart, Van Gough's painting is great, Mozart's music is beautiful. Einstein's relativity is genius. Isn't this proves that there's something intrinsic, yet we don't understand, that is intelligence. AI and cognitive science should mtual fertilize in the area in the comming decade. -- Kung Hsu, MIPS Computer System, tel: (408)991-7762
muttiah@cs.purdue.EDU (Ranjan Samuel Muttiah) (10/11/89)
In article <357@massey.ac.nz> ARaman@massey.ac.nz (A.V. Raman) writes: >intelligent system at present known to us. This is a direct >consequence of the fact that the human brain is also the most >efficient knowledge base known. In this article, we analyze the >If what we call creativity is a measuring stone to intelligence, >hasn't Nature ever been as creative as we have. Hasn't it learnt >from mistakes, just as we do. Or maybe this article itself is one of >Nature's mistakes. > - &/.. First, distinguish rational acts from irrational ones. I happen to believe (at least for the moment) that what are rational acts now were irrational at one time or the other.
weltyc@ (Chris Welty) (10/11/89)
This reminds me of a genetic algorithms paper where the author, in an attempt to support his contention that mutation combined with genetic recombination was an efficient form of search which is capable of avoiding local maxima, stated that nature employed this technique sucessfully (look at the complexity of our brains and the dexterity of our hands....) Made me wonder how smart Nature really was, using the GA to search for something that would destroy it.... Christopher Welty --- Asst. Director, RPI CS Labs | "Porsche: Fahren in weltyc@cs.rpi.edu ...!njin!nyser!weltyc | seiner schoensten Form"
bwk@mbunix.mitre.org (Barry W. Kort) (10/11/89)
In article <2394@uceng.UC.EDU> dmocsny@uceng.UC.EDU (Daniel Mocsny) writes: > In article <2376@munnari.oz.au>, ok@cs.mu.oz.au (Richard O'Keefe) writes: > > In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > > > Hasn't it [Nature] learned from mistakes, just as we do? > > No. Unless she can form intentions, Natura cannot make mistakes. > How interesting to both personify and de-personify Natura in one sentence! Einstein observed that Man is part of Nature, too. Joseph Campbell suggests that we are Nature's Consciousness. This sentiment is also reflected in the Gaia Theory. So at least a part of Nature is personifiable and capable of intentional behavior. I sometimes make mistakes on behalf of Nature. --Barry Kort
ok@cs.mu.oz.au (Richard O'Keefe) (10/13/89)
In article <73510@linus.UUCP>, bwk@mbunix.mitre.org (Barry W. Kort) writes: > In article <2394@uceng.UC.EDU> dmocsny@uceng.UC.EDU (Daniel Mocsny) writes: > > In article <2376@munnari.oz.au>, ok@cs.mu.oz.au (Richard O'Keefe) writes: > > > In article <357@massey.ac.nz>, ARaman@massey.ac.nz (A.V. Raman) writes: > > > > Hasn't it [Nature] learned from mistakes, just as we do? > > > No. Unless she can form intentions, Natura cannot make mistakes. > > How interesting to both personify and de-personify Natura in one sentence! Evidently some of the articles in comp.ai aren't getting through here, because I didn't see Daniel Mocsny's article. I do *not* personify ``nature''; I used the feminine article and the form Natura to stress the personification that _Raman_ was using. > Einstein observed that Man is part of Nature, too. Uh, people have been putting statements like that in writing as far back as we can read the writing. I have some metal in some of my teeth, so that metal is part of me. Does that mean that I have all the characteristics of metal? Obviously not. The metal is part of nature. Does that mean that the scheme of things has all the characteristics of that metal? Again, no. So from (part-of X Y) and (has-feature X F) it does not in general follow that (has-feature Y F). For some features, it does follow (the properties of occupying space, possessing mass, and so on). From the rather banal observation that human beings exist, no conclusions can be drawn about whether the scheme of things can make mistakes UNTIL we have first established that "is able to form intentions" (or some such property) is one of the few properties which does transfer like that. This has not been done. In fact there is a rather bigger presumption in this woodshed. The assumption is that ``Nature'' refers to something. But what is the genus of ``Nature'' and what are its differentia? I have a strong suspicion that ``Nature'' as a ``thing'' may be about as sensible as a set of all sets. > Joseph Campbell suggests that we are Nature's Consciousness. Well, it's a lovely sentiment, but what does it _mean_? It certainly can't be using "consciousness" in the way that it is used when referring to people and animals. Consciousness generally includes the idea of awareness of ones surroundings, and what are the surroundings of ``Nature''? [It's a very anthropocentric claim too, but that's another story.] > So at least a part of Nature is > personifiable and capable of intentional behavior. Yes, but that demonstrates nothing about ``Nature'', any more than the presence of metal fillings in my teeth proves _me_ to be made of metal. > I sometimes make mistakes on behalf of Nature. How do you know this? To do X "on behalf of Y" is to assert that Y wanted X done and appointed you as an agent or representative to accomplish this in Y's interest. How does the goddess Natura tell you what she wants you to do on her behalf? Does she appear to you in a dream? Look, I can tell the nice policeman ''I took that money out of the man's pocket on his behalf'', but to back it up I have to show that he did in fact want me to act as his agent in this matter. To back up the claim that you "make mistakes on behalf of Nature" you must provide evidence that the goddess asks you to do it.