eliot@phoenix.Princeton.EDU (Eliot Handelman) (02/26/90)
In article <132035@sun.Eng.Sun.COM> music-research writes:
;Music-Research Digest Sat, 17 Feb 90 Volume 5 : Issue 14
;
;Date: Mon, 12 Feb 90 09:43:48 EST
;From: Otto Laske <laske@edu.bu.cs>
;Subject: Reply Regarding Knowledge Acquisition
;To: music-research <music-research%uk.ac.oxford.prg@uk.ac.nsfnet-relay>
; I start from the idea that musicology is a discipline of knowledge engin
;eering,
The Otto Laske definition of musicology, at odds with everyone else in the
field. One may as well say that the study of history is really the study of
the military expertise of Napolean Bonaparte.
; I add the experience that knowledge engineering requires knowledge
;elicitation, knowledge analysis, and knowledge modeling, and that one
;cannot engineer knowledge without modeling it
"Nice neutral term that says nothing." -- OTTO LASKE, in Music Research
Digest. The word "knowledge" is definitely overworked, one might suggest.
; I proceed from the observation that traditional musicology was a
;hermeneutic science practicing "knowledge engineering in reverse",
;viz., starting from products of musical thinking,
Is Otto Laske aware that the ETA Hoffman's prose analysis of the 5th Symphony
of Beethoven had a decisive impact on the compositional ideology of Robert
Schumann? Perhaps "traditional musicology" (a nice neutral term that says
nothing) has implications for the study of the compositional process that
Laske has not been able to forsee. Let his criticisms therefore apply to
his own narrow understanding of history.
;and that contemporary
;computer music research practices "ad hoc knowledge engineering" starting
;from presumptions of what might be musical knowledge, but having no
;empirical theories to support the (implicit) claim.
Meaningless assertion in the absence of any concrete example. What research?
;Most of this ad hoc knowledg
;e is public knowledge taken from text books,some of it is shared
;(collegial) knowledge, little or none of it is personal, idiosyncratic
;knowledge.
Who is Laske talking about? Is it, as I suspect, the Scarecrow of Oz? Which
textbooks? Laske, as the scientist he professes to be, leaves us all in
the dark as to the motivations of this resounding (though neutral, hence
meaningless) critique.
; I conclude that we need a theory of musical action (or activity),
;and that,therefore, cognitive musicology is probably not a "mainstream"
;(newtonian) science but an "action science" that gears its solutions
;to a "community of practice" instead of working with the distinction
;between "theoretical and applied" sciences
;which is a
;somewhat esoteric topic, but should be emphasized.
Here's the solution to all of Laske's big problems: the "community of
practice" consists of all those unnamed people who practice "ad hoc
knowledge engineering." How's about that?
;THE PREDICAMENT (if they only knew it)
It's hard being a visionary, Otto.
; The predicament of getting stuck in public knowledge--which is
;verbal, not action knowledge--is shared by the computer music and expert
;system community, except that the latter is aware of the predicament,
;and the former is not. (This is a crucial difference which colors
;everything happening in "computer music", an anachronistic term anyway).
Just who are these people? I want to see instances of research cited that
demonstrates that the computer music COMMUNITY -- not just one or two
isolated instances of practice -- is "stuck in this predicament." Because
I don't see that "knowledge engineering" is even an issue to begin with
in the computer music community, let alone that the community is being
held up by the fact that "knowledge" is transmitted verbally.
;PRECOMP, implemented by Don Cantor
;and called K
;AIMU by him,
Since he wrote the code, I say let's all call this thing KAIMU. I wouldn't
want Laske renaming my programs.
;It is both a "precompositional tool" (therefore PREC
;OMP) helping a composer prepare a definitive score for instruments and/or tape,
;and an (element
;ary) knowledge acquisition tool, in that it monitors, in an original and interes
;ting way, the expert's deliberations about how to find structure
;in, and segment, an "event list" (produced by the generator), in order
;to define a definitive score.
Hold on a second. Acoording to Otto Laske, KAIMU is "original and interesting."
According to Otto Laske, this is knowledge and that isn't. According to
Otto Laske this is empirical and everything else ain't. Maybe Otto Laske
is going too far as an arbiter or interestingness, knowldege and empricism?
How's about, Otto, we submit to the community and see whether anyone
agrees?
; Don Cantor's originality lies in having designed pictures of
;the analytic structure of event lists, and having thought up and
;implemented the "segmentation advisor" as a set of rules whose slots
;are set by the expert, who is being monitored when doing so. See
;his dissertation.
Yes Otto. Who is Don Cantor, what does he know about composition, and
why should his pictures count as anything more than his pictures? What
makes them empirical, for example? Does anyone who has drawn pictures
of the analytic structure of event lists qualify as a knowledge engineer?
Does Schenker, for example?
; There are many limitations to PRECOMP in its present form. Its rule
;set is primitive; its protocol generation is passable; it does
;nothing about how to evaluate protocols, manually or automatically;
;it is concerned only with a single event list (score section), while
;the composer/designer is concerned with many such sections at the
;same time; in many instances it repeats insights that can be gleaned
;from pictures of event list analyses; etc.
Laske, this doesn't sound like "knowledge engineering" at all. Sounds
more like "data acquistion." Might be easier to use something like
the sort of history mechanism used in emacs, that way you can look
at "what the expert did next." Good luck.
; (1) that to capture music knowledge means to capture action
; knowledge
You haven't demonstrated that. You don't actually have "knowledge" on your
hands. You have a bunch of data that you can't figure out how to analyze.
So let's not call this "capturing knowledge." Let's call it "recording
keystrokes."
; (3) that we need a theory of knowledge representation for music
; knowledge -- a major concern of the annual A.I. and Music
; Workshop (1990: ECAI, Stockholm,Sweden)
Chairman: Otto Laske.
; (4) that we cannot afford ad hoc stipulations of what is musical
; knowledge but need empirical theories,
Maybe it's time to get down off your empiricist high horse, Laske. Composition
is not an empirical process, nor have you succeeded in showing that it's
even an observable process. In any case I wonder how useful empiricism is
in modelling cognitive processes. One wants to incorporate facts if there
are any, but composition is such a high level activity, intrinsically
molded by one's own conception of what it is, that it's hard to see if there
is even such a thing as one compositional FACT that empirically obtains,
let alone a whole theory of what the thing necessarily must be. Laske's
aiming far too high, partly, I think, because his view of composition
is too narrow. Equivalently, one might try to find the facts concerning the
process of writing novels.
; (6) that composers are experts as others, inspite of the romantic
; Western European tradition treating them as special, geniuses,
; and the interest of most composers to hide their knowledge, in
; order to seem more unique and mysterious; that they are designers;
; and that without their cooperation, as that of living musicians
; generally, musicology is a dead field.
Laske lets off another resounding and odious fart. Laske says this is so,
ergo it is so. Laske's concept of composers as "designers" completely
overlooks the entire sociological history of music, that music evidently
has an importance that goes far beyond whatever aesthetic niceties it
may have to offer. Music can incite to rage and violence, it can tranquilize,
it is the most powerful instrument of propaganda and worship known,
it was thought in the 19th C to be the unique vehicle of the sublime,
it is an activity powerfully charged with its own history and accomplishment,
of its influence and influences, its acheivement whether explicit or
merely ideological.
All this Laske reduces to a simple behavioristic formula: it is "design."
Composers should not be regarded as spacial or geniuses, though there can
be no doubt that those composers who count are obviously special, and are
obviously able to do something that others are unable to do. If they would
only hand their knowledge over to Laske, he would de-mystify the entire
consciousness of mankind, show that composers are merely experts in design,
and that programs written for the Mac are able to capture this expertise.
But they won't, because then they won't be regarded as special any lonmger.
This is about the most twisted and stupid version of mankind that I've ever
come across. In comparison to this, L. Ron Hubbard could almost be called
interesting. His picture of the mind is a little bit more realistic.
--E. Handelman
Princeton U., Dept of Musicedhall@rand.org (Ed Hall) (02/26/90)
In between Eliot's personal attacks on Otto Laske lies a lot of commentary I must humbly agree with. There should be a special place in hell reserved for the person who coined the term ``knowledge engineering.'' Like so much in AI, a term which started as a metaphor has been erroneously accepted as a factual entity and then further perverted into marketing hype. The techniques which fall under the rubric ``knowledge engineering'' are clever tricks for organizing the development of computer programs, but hardly represent the revolution that has been claimed for them. Successful projects have tended either to be solutions to toy problems or to involve subject areas with innately limited complexity. This makes it especially silly to see it proposed for a field as profoundly open-ended as musical composition. At the very best, Laske will come up with yet another technique of computer-aided musical composition. This is just fine by me. But insofar as musical creativity derives from the peculiarly distinct experiences and abilities of human beings combined with their relationships to a kaleidescopically changing world, his project will necessarily result in a frozen and incomplete representation of the creative process. -Ed Hall edhall@rand.org
smoliar@vaxa.isi.edu (Stephen Smoliar) (02/27/90)
In article <1990Feb26.091939.3171@rand.org> edhall@rand.org (Ed Hall) writes: >_s{w3 There should be a special place in hell >reserved for the person who coined the term ``knowledge engineering.'' (Wasn't he from Rand, Ed?) :-) >Like so much in AI, a term which started as a metaphor has been erroneously >accepted as a factual entity and then further perverted into marketing >hype. > All kidding aside, I think Ed has taken an important step in clearing the air here. Those who are interested in the relationship between artificial intelligence and intelligent behavior know that knowledge engineering is little more than a form of software engineering developed for the symbol manipulation capabilities of rule-based systems (and, when all is said and done, it does not amount to particularly GOOD software engineering). The term is used most heavily by those who are still trying to pawn off expert systems as a manifestation of intelligence, as opposed to simply a new approach to programming which had been utterly foreign to all those poor souls who thought that programming was delimited by the capabilities of FORTRAN and COBOL. After these pitch-men, the term is used primarily by people who want to "talk" the game of artificial intelligence but are pretty much incapable of "playing" it. Most of these folks are in management, but there is also a high contingent of non-technical types with a great urge to wax philosophical about intelligence. As the evidence rolls in, it seems as if Laske falls in this latter category. I would like to believe that if he rid himself of the urge to use everyone else's jargon and use his own choice of simple language, he might have something to say; but I fear I'm still waiting to be convinced. Meanwhile, I suggest that those of us who are more interested in the intelligent behavior side of the story forget all about the snake oil of knowledge engineering and debate more serious issues, such as what is an appropriate methodology to accommodate both the observation and the modeling of such intelligent behavior as it pertains to making 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