daemon@bartok.Sun.COM (03/18/90)
Music-Research Digest Sun, 18 Mar 90 Volume 5 : Issue 25 Today's Topics: knowledge engineering (3 msgs) Music-Research Digest Vol. 5, #21 *** Send contributions to Music-Research@uk.ac.oxford.prg *** Send administrative requests to Music-Research-Request *** Overseas users should reverse UK addresses and give gateway if necessary *** e.g. Music-Research@prg.oxford.ac.uk *** or Music-Research%prg.oxford.ac.uk@nsfnet-relay.ac.uk ---------------------------------------------------------------------- Date: 6 Mar 90 23:09:16 GMT From: Brad Miller <miller%rochester%uhura.cc.rochester.edu%sunybcs%zaphod.mps.ohio-state.edu@edu.ohio-state.cis.tut> Subject: knowledge engineering To: music-research@prg OK, lots of recent bashing of "knowledge engineering" but I'd like to clarify a few things.... while "ke" is overused and misused, and I'm not sure what the hell it means anymore one shouldn't necessarily retreat to behavioristic approaches (that is, an approach that models intelligent behavior as has been recently suggested here). What it does mean, is that once cannot do ke for music because there aren't any decent KRs yet (knowledge representations). I think developing a good kr for musical "rules", both in terms of basic music theory (e.g. what scales seem to sound good, or why modulating the tonic to the fourth and fifth is a good progression..) and for certain stylistic information is necessary before any real work can proceed anywhere. In some sense this is an argument against neural net approaches, etc. because while one may be able to come up with a system which happens to behave nicely, one cannot in any sense describe what the system actually "know"s. KE, for music, then, consists of taking this kr, from which one can theoretically come up with music charts, and actually encode the special (stylistic/performance) and general (theoretic) knowledge. The goal would not be, of course, to simply have a prolog-like program that could generate a valid musical piece, but rather to have a prolog-like program that could act as an intelligent arranger/compositional enhancement tool. (not that the UI would in any sense be prolog-like of course, I would simply hope to encode the knowledge in this fashon). Something of an expert system in the general sense, although the techniques would actually come more from the planning and natural language domains (since a peice of music is very much like a plan to be executed: one wants to acheive certain goals, they must be done in some temporal order, and things like duration are crucial to acheiving particular effects. My work is much too sketchy at this point to present anything too cogent here (e.g. the rule structures), but the point is, as someone in the AI research field, I don't think one should discard non-behavioristic approaches too quickly. They have the theoretical advantages of being able to build a system that can be well understood, both in terms of construction, programming, and interpreting of results. Connectionist approaches may well make a better music composer in the long term, but until problems like "appropriate motivation" for these systems are solved, I'm not convinced we will personally appreciate the output. (i.e. why do you think a system that doesn't enjoy music would write good symphonies?) Provocatively yours, ------------------------------ Date: 7 Mar 90 16:37:40 GMT From: Stephen Smoliar <smoliar%venera.isi.edu%usc%elroy.jpl.nasa.gov@com.dec.decwrl> Subject: knowledge engineering To: music-research@prg In article <1990Mar6.230916.21115@cs.rochester.edu> miller@CS.ROCHESTER.EDU (Brad Miller) writes: >OK, lots of recent bashing of "knowledge engineering" but I'd like to >clarify a few things.... while "ke" is overused and misused, and I'm not >sure what the hell it means anymore one shouldn't necessarily retreat to >behavioristic approaches (that is, an approach that models intelligent >behavior as has been recently suggested here). > First of all, I don't think we are talking about retreat here. The heart of Turing's original test it that you have a machine whose behavior cannot be distinguished from that of a human. If such a machine exists, then you have grounds for calling its behavior intelligent. Of course, if you want to make any concrete progress, you should probably begin by trying to slice off some particular ASPECT of that behavior as your focus of study. This brings me to my second point. If we are interested in music, then the behavior we wish to study amounts to what we would be willing to call "musical behavior." In his excellent paper, "Music Theory, Phenomenology, and Modes of Perception" (which I reviewed about a year ago on rec.music.classical), David Lewin gave a survey of the breadth of such behavior which, among other things, illustrated that what might be called "processing music notation" is a rather small subset whose significance may be over-inflated. Thus, while Brad makes a good point: >What it does mean, is that once cannot do ke for music because there aren't >any decent KRs yet (knowledge representations). I wonder if the problem does not lie in a myopic view of knowledge representation which is paying too much attention to the constructs of music notation. This seems to be the direction Brad is heading, and I just want to raise a flag of caution: > I think developing a good kr >for musical "rules", both in terms of basic music theory (e.g. what scales >seem to sound good, or why modulating the tonic to the fourth and fifth is a >good progression..) and for certain stylistic information is necessary >before any real work can proceed anywhere. In some sense this is an argument >against neural net approaches, etc. because while one may be able to come up >with a system which happens to behave nicely, one cannot in any sense >describe what the system actually "know"s. KE, for music, then, consists of >taking this kr, from which one can theoretically come up with music charts, >and actually encode the special (stylistic/performance) and general >(theoretic) knowledge. > We are now moving into the realm of a fundamental premise of artificial intelligence: Newell's "knowledge level" hypothesis. I just got done taking a skeptical swipe at this on the Artificial Life Mailing List; so I'm sort of "primed" to summarize my thoughts. Newell's hypothesis arises from a synthesis of his work in artificial intelligence with his work in computer architecture. In the latter area, he developed a stratified approach to describing computer systems in terms of a series of levels of increasing capabilities: device, circuit, logic circuit, register-transfer, and symbol. Each of these levels is based on some set of concrete objects and a set of rules which determine how those objects may be assembled into structures. Newell's hypothesis is that there is a "knowledge level" which sits on top of the symbol level; and thus, there are "objects of knowledge" and rules for the assembly of those objects. It basically rejects the view that knowledge is an epiphenomenon of our behavior. Furthermore, by being situated above the symbol level, Newell's hypothesis also asserts that these "objects of knowledge" may be constructed from symbols (just as circuits are constructed from devices). This hypothesis is appealing because it implies that we may "engineer" knowledge, just as we "engineer" circuits. It's simply a matter of developing our understanding of how to put things together. However, IT IS STILL ONLY A HYPOTHESIS. Certainly, the track record of expert systems offers no sound confirmation of that hypothesis. Indeed, the ways in which the behavior of expert systems deviates from that of humans may ultimately detract from the hypothesis. Nevertheless, we are going to be very reluctant to give that hypothesis up because we have a long intellectual tradition which allows us to reason about ASSEMBLING THINGS and inclines us to raise eyebrows at any mention of epiphenomena. Connectionism is definitely a step away from the knowledge level hypothesis, for precisely the reason Brad pointed out: you cannot talk about the knowledge contained in a neural net the way you can talk about the contents of a knowledge base. At the same time, Brad is also right in wanting to have a way to interact with any system about "what it knows." However, I don't think any of us would like a dump of a knowledge base in our attempt to find out what even a simple expert system "knows." In other words, we are going down a slippery slope of issues of "knowing about knowledge." Yes, Virginia, we are talking about CONSCIOUSNESS now. At this point I should step aside and let Eliot take over, since he is confronting such matters head-on in his doctoral research. However, I want to close out with two quick points: 1. We don't have to be afraid of epiphenomena any more. Once scientists became interested in the dynamics of non-linear systems (let's see if we can avoid the "C" word), they realized that they would need new terminology to describe the properties emerging from such systems. Those of us who deal with music may very well have insights which will contribute to the development of that terminology. 2. There IS research in the study of mechanisms which give rise to consciousness as an epiphenomenon. This is what Gerald Edelman's latest book, THE REMEMBERED SELF, is all about. What this means is that it should not be out of the questions to build a workbench to investigate these matters, even as they pertain to music. ========================================================================= USPS: Stephen Smoliar USC Information Sciences Institute 4676 Admiralty Way Suite 1001 Marina del Rey, California 90292-6695 Internet: smoliar@vaxa.isi.edu "Only a schoolteacher innocent of how literature is made could have written such a line."--Gore Vidal ------------------------------ Date: 15 Mar 90 04:36:30 GMT From: Mark Gresham <mgresham%artsnet@edu.gatech> Subject: knowledge engineering To: music-research@prg In article <12275@venera.isi.edu> smoliar@vaxa.isi.edu.UUCP (Stephen Smoliar) writes: >This hypothesis is appealing because it implies that we may "engineer" >knowledge, just as we "engineer" circuits. It's simply a matter of developing >our understanding of how to put things together. However, IT IS STILL ONLY A >HYPOTHESIS. Certainly, the track record of expert systems offers no sound >confirmation of that hypothesis. Indeed, the ways in which the behavior of >expert systems deviates from that of humans may ultimately detract from the >hypothesis. Nevertheless, we are going to be very reluctant to give that >hypothesis up because we have a long intellectual tradition which allows >us to reason about ASSEMBLING THINGS and inclines us to raise eyebrows at >any mention of epiphenomena. > >Connectionism is definitely a step away from the knowledge level hypothesis, >for precisely the reason Brad pointed out: you cannot talk about the knowledge >contained in a neural net the way you can talk about the contents of a >knowledge base. At the same time, Brad is also right in wanting to have >a way to interact with any system about "what it knows." However, I don't >think any of us would like a dump of a knowledge base in our attempt to find >out what even a simple expert system "knows." In other words, we are going >down a slippery slope of issues of "knowing about knowledge." Yes, Virginia, >we are talking about CONSCIOUSNESS now. I wonder, Steve, if the difference between what an expert system "knows" and its "knowledge base" is a parallel to the difference between "music" and a "theory of music" -- this all being and intuitive observation on my part. More intuition: what a system (or a person) comes to "know" can alter its "knowledge base" in the same way that one's cumulative experience of "music" can alter one's "theory of music." Cheers, --Mark ======================================== Mark Gresham ARTSNET Norcross, GA, USA E-mail: ...gatech!artsnet!mgresham or: artsnet!mgresham@gatech.edu ======================================== ------------------------------ Date: 15 Mar 90 21:37:11 GMT From: pa2253 <pa2253%sdcc13%sdcc6@edu.ucsd> Subject: Music-Research Digest Vol. 5, #21 To: music-research@prg In article <14531@phoenix.Princeton.EDU> roger@phoenix.Princeton.EDU (Roger Lustig) writes: >Tell me, who has forced computer music composers to ignore any >possibilities whatever? Where are such things happening? Possibilities can be threatened if your time is wasted on narrowly focused pedagogy. >counterpoint -- from the orchestral works of Felix Mendelssohn. There >we derived "models" of music, and hten applied them to pieces in all >kinds of different styles. Needless to say, our instructor had received >many years of training in such matters, as well as a thorough grounding >in 12-tone composition. Several members of my class have gone on to become >musicians of various types; some do electronic/computer music. So does the >instructor, Paul Lansky. First, I must acknowledge the potential benefits of theoretical analysis and synthesis. Your collective study of Mendelssohn and the synthesis of "models" of music may have been rewarding. You, however, attend Princeton University--a university I'd like to attend in the Fall. At such a university I can see the encouragement of possibility--whereas UCSD imposes a pedagogical theoretical curriculum which is founded upon rigid style analysis projects. These projects waste the student's time by demanding that a specific task be performed without making the demand for synthesizing the task into a larger philosophical or theoretical framework. By imposing upon the composers time, they succeed in dictating musical preferences. In another light, you have not provided a rationale for describing the necessity for a "thorough grounding in 12-tone composition." Why not a thorough grounding in the techniques and structures inherent in rap music? Polka? Blues? automobile transmissions? Clearly, you must see that agendas exist. Theoretical models exist only for a handful of musical styles. This absense is another stylistic imposition/suggestion. Instead of choosing Mendelssohn, why couldn't you analyze the works of Gordon Mumma? The reason is clear--obvious theoretical models would have little utility in describing the music of Mumma. >Seems to me that the vast majority of music being composed and performed >today, with or without computers, whether academic, mainstream >"classical," jazz, pop, or any of a thousand other kinds, has strong >roots in one or another musical tradition. Pop is especially strong in >its traditions, and most of the music being written TODAY in that genre >and others is directly connected to the music theory of the "past," such >as harmony, phrase analysis, and species counterpoint. A piece of music obviously has an implicit relationship to the collective history of music. However, that does not require an individual to compose for historically obvious instruments in historically delineated ways. Why should a student of composition be forced to learn and use someone else's musical tools when they are more comfortable and facile developing their own? I understand MY musical history better than an institution, and I have the motivation and desire to create my OWN music as much as it is possible. >Again, that's a straw man. Who imposes these agendas? Seems to me it >can only be the composer herself. Anybody else is just teaching theory, >or teaching crafts of composition. The real work of composition is >never hampered by such teaching, except if the composer allows this to >happen by losing sight of her own interests. That is absolutely ridiculous! I doubt that you have had the experience of working 30 hours a week, while taking prerequisite music theory courses and attempting to compose. Clearly in my case, music theory has been an imposition. I was never asked to synthesize the information I was given: that I did automatically and effortlessly; I was asked to submit to rigid projects that gave no creative latitude and, therefore, were completely meaningless to me. The courses did not promote discussion or evaluation. Although I gained a few useful concepts from the courses, these ideas could have been presented in a much more interactive and efficient manner. I have taken composition courses where professors literally "corrected" my scores. I had notes erased and replaced by notes that the professor felt were "more efficacious in this context." He encouraged us to compose in pencil. Hmmm.. I think that the composer "must" have imposed those corrections on himself by making such a foolish musical choice in the first place. Just as in Enlightenment Christianity, you have total freedom as soon as you acknowledge that there is only one choice of action |-(. >Seems to me that the issues of instrumentation and pitch system in >computer composition are simply of a different type from the issues of >"traditional" composition. There is no incompatibility there at all. >Moreover, since almost all composers who use hte computer are >interested in using traditional materials in some way, it is important >for them to have an inside-out knowledge of these materials. >COLLEAGUES? I want them to be able to communicate with their >audiences! Don't you think there might be some experience that a >thousand years of insightful composers might have picked up, that might >be applicable to musics other than their own? Sounds like musical homogenization to me. You have no right to dictate musical values. You assume that most computer music composers are interested in traditional materials. This is completely unjustified. I use the computer to mutate and transform the traditional. See, I am different. Must I conform to your vision of music as you wish it to be? What does it mean to have inside-out knowledge of musical materials? >From the rest of your statements I gather that you imply that there is only ONE such understanding, ONE such truth. Knowledge is relative; there is a multitude of perspectives. I tend to look at musical objects timbrally--I think and perceive in terms of density, energy, space, motion, spectrum, time, contrast, and semantics. I fully recognize that other composers may employ a different language to describe their music and hence they will establish a different musical context. My intention is to insure that the multitude of contexts are not systematically destroyed by the imposition of 'proper musical processes.' Someone else's experience is never automatically useful--knowledge must be integrated into the individual's own history and experience for it to have any utility. >Another good reason to learn "hackneyed" theory is EXACTLY its >simplicity. To do almost anything well, one needs to master simpler >versions of the task first; and complete mastery of old-fashioned >technique has always stood composers in good stead. Schoenberg taught >most of his students nothing but elementary harmony and counterpoint; >Brahms likewise recommended the archaic ideas of Fux and CPE Bach to his >students. Not coincidentally, it was those two writers whom Haydn used >in his self-education; they were both describing music that was outdated >by the time Haydn read them, and htey made him into one of the great >radicals in music history. Mastering them, that is -- not rebelling >against them, even! One does not have to first become a Nazi to rebel against fascism. This metaphor hopefully shows how ridiculous your statement is. You CAN become a Nazi first, but I choose not to. A composer does not have to 'master' a technique if it offends her, she can simply avoid it and employ another. I have chosen my own focus: digital signal processing. I find this path to be functional and satisfying. This is not by any means my sole compositional process, it is simply a powerful tool that has been very useful to my music. Digital signal processing is a subject extremely alien to traditional music theory; however, it can be extremely intimate with respect to music. Now I guess that you will say that such a focus is fine as long as I share my attention with traditional music theory. A human being has only finite amount of time and patience available to her; I intend to maximize the share of attention that my interests receive at the expense of copious frustration and disintegration. >I wish I'd taken more music theory. The last class I took was called >Composition for Musicologists. In that class, we were given little >snatches of 18th-century-sounding music, and told to complete them. My >thinking about all kinds of music was enriched by this exercise. I am glad to see that someone can benefit from such exercises. I grant that my exposure to such exercises has often enriched my musical thinking in positive ways, but so has rap music. To think about ALL music one must be exposed to ALL music. 18th century art music is a mere fraction of all available music. I am not going to ignore the flocks of birds, pneumatic drill grooves, the occasional F14 flying over my house, Jupiter Larson, wind chimes, the drums of a TGIF from five miles away, Robert Ashley, poetry readings, by preoccupying myself with my next analysis hoax. I feel that I can take responsibility for my education and I can justify myself with my musical compositions. The justification for my compositions can come from an explication of their impetus, and what methods I employed to realize them. I would prefer a world where justification was not extant, but I feel that I can justify my work from a variety of perspectives. Christopher Penrose penrose@do.ucsd.edu ------------------------------ End of Music-Research Digest