eliot@phoenix.Princeton.EDU (Eliot Handelman) (03/22/90)
;Music-Research Digest Sun, 18 Mar 90 Volume 5 : Issue 28 ;Date: Thu, 8 Mar 90 10:27:08 EST ;From: Otto Laske <laske@edu.bu.cs> ;Subject: greeting in French ;To: music-research <music-research%uk.ac.oxford.prg@uk.ac.nsfnet-relay> I'll summarize the jist of this because I think it is interesting, and because I'm sure some people will appreciate it. This is very free. ; "Il est interessant d'observer que, dans deux "univers musicaux" ;apparemment opposes--celui des traditions orales et celui de la creation ;contemporaine--,la difficulte de verbaliser les mecanismes de perception, ;evaluation et production de la musique, contribue a renforcer une opinion ;courante (sic) selon laquelle l'activite musicale n'est pas aussi ;systematique que les musiciens le pretendent, autrement dit que toute ;tentative de modelisation de cette activite est vouee a l'echec. He is saying that the difficulty of verbally describing the perception, evaluation and production of music has led people to believe that musical activity is unsystematic and therefore can't be modelled. ;En ;premier lieu, cette conception reflete une mecomprehension de la demarche ;scientifique experimentale, ou l'echec est une source d'enseignement a ;part entiere; However, he says, science works by learning from its own failures; ;en second lieu elle encourage, du moins dans les societes ;traditionnelles, un reniement de methodes pedagogiques hautement ;sophistiquees au profit de ce qu'on peut appeler l'apprentissage par ;imitation (Kippen, The Table of Lucknow: a Cultural Analysis of a ;Musical Tradition, Cambridge University Press, U.K., 1988). and secondly the conception that music is not really systematic encourages, at least in traditional societies, the renouncement of highly sophisticated pedagogical methods in favor of learning by imitation. (I don't understand the point he's trying to make here.) ;Vraisemblable- ;ment pour les memes raisons on constate que tres peu de chercheurs en ;informatique musicale s'interessent a l'acquisition des connaissances, ;alors qu'un nombre grandissant de theoriciens projettent leurs espoirs ;dans le connexionisme en raison de leur interet presque exclusif pour ;les activites de perception et de memorisation. It seems true, he continues, that for the same reasons very few researchers in [computer music, music cognition, etc] are interested in knowledge acquisition, whereas an increasing number are putting their hopes in connectionism because they are mainly interested in perception and memorization. [I'm going to comment on this below.] ;On n'hesite pas aujourd' ;hui a identifier le processus de "creation" avec celui de "generalisation" ;qui est mis en oeuvre dans un resaut neuromimetique. One doesn't hesitate to identify the process of "creation" with that of "generalization" which is put to work in a neural net. [I assume "resaut" is a typo]. ;Ici encore Laske ;met en garde les musiciens contre une conclusion aussi naive, soulignant ;que la creation artistique est bien plus qu'un processus de reorganisation ;ou de generalisation des oeuvres anciennes, et que si un observateur croit ;voir ou entendre une oeuvre radicalement nouvelle creee par un reseau a ;partir d'oeuvres existantes, il faut y voir une limitation de ses facultes ;perceptives ou conceptuelles (reponse, music research digest 1990). Here again, says Bel, Laske points out that creation is much more than the process of reorganizing older works. Now say you make a piece in this way -- you get a net to churn out something based on older music. If somebody things that that this piece is radically new, you have to assume that this person's perceptual or conceptual faculties are limited, says Laske. (Laske is supposed to have said this in the MR digest -- maybe it was when he wrote about Todd's article in CMJ?) ;Bernard Bel, Considerations generales et methodologiques, Laboratoire ;Musique et Informatique Marseille, 1990 (janvier). And now for some comments. First of all, if anyone can see what connection there is between Bel's comments concerning the compositional use of nets and our discussion of KA, let them speak. As far as the matter concerning making music with nets goes, I so far have not noticed very many, indeed any, musicians monkeying around with these things and I'm not prepared to say what they're worth compositionally until musicians do start monkeying around. For to date the only people monkeying around compositionally with nets have been computer science or psychology department types, and no matter what technology these people use, their results are bound to be not terribly interesting. [Let me qualify that lest I get a barrage of email telling me about Kohonen: I have absolutely no interest in the kinds of things that amateurs find exciting, like generating tonal music, or music "in the style of," or ways of making one piece sound sort of like another piece, or trying to make music sound "authentic" or "expressive," or arranging chords so that your tune comes out moderately "happy" or "soul searching." I don't say this sort of thing shouldn't be done -- I'm only saying that I don't care about this kind of work at all.] Secondly, to Bel's remark that nets are only good for perception and memorization -- they may be good for memorization but I haven't yet seen any evidence that they are good for perception, unless one means "recognizing sequences on which the net has been `trained,'" which isn't what I would mean. Thirdlly, Bel seems to be suggesting something about the systematization of creativity through knowledge engineering, or seems to be dropping a hint in that direction. I'd be interested in seeing the point elaborated, assuming that was the point. E. Handelman --Princeton U., Music