|
harmonizations from Ebcioglu's program with the Roland Werckmeister III and meantone temperaments. This piece is included in the enclosed Compact Disk.
Expert listeners cannot be fooled by this trick, but it appears that a substantial number of musically literate people identify any meantone performance to be by Bach and any Werkmeister III performance to be by the computer, irrespective of the authorship of the harmonization. This finding underscores the important distinction to be made between a musical composition as a rule-constrained phenomenon and a musical performance as a sound-enabled phenomenon, or, in other words, between the cognitive and perceptual domains of understanding. When the evidence is presented aurally, we can more readily identify a 'Bach sound' by the wheezes, squeaks, and dissonant overtones of traditional organs than we can a 'Bach texture' by the movement of inner voices in relation to outer ones. Were the evidence to be presented visually, then the sound would be immaterial and the false texture immediately recognizable. If Ebcioglu's textures were fed to Malinowski 's Animation Machine, for example, the differences between natural and artificial harmonization would probably be much more readily perceived. Not infrequently these harmonizations require simultaneous of sequential spans in the range of a 10th, 11th, or 12th. Some other attempts to automate musical processes include the work of Bill Schottstaedt at Stanford University to implement the rules of first- and second-species counterpoint. Like Ebcioglu, Schottstaedt had to implement an entire rule system, but he started with the extensive rule sets codified in the early eighteenth century by Johann Jacob Fux in the Gradus ad Parnassum. Schottstaedt had to create various algorithms for the resolution of conflicts between rules, another need that theory treatises rarely discuss. His procedures resemble chess algorithms, with the directional moves often masking the ultimately desired position. The development of expert systems rests on the assumption that, to paraphrase the English poet William Blake, wisdom can be put in a silver rod. Yet what expert systems in music may show us is that not all wisdom can be put in a silver rod, or else, as a different poet, the American George Santayana, put it: Only wisdom is not wise. The collection and identification of rules to improve our knowledge bases is an inherently worthwhile activity, but it appears that much more thought will be required to create convincing specimens of music automatically fleshed out from an originally natural skeleton.
New Paradigms (5): Musical Analogues of Neural Processing
The trial-and-error approach that lies at the heart of neural computing is an antidote to the rule- and knowledge-bank approach of artificial harmonization and artificial composition. If this approach has applications to the study of historical repertories, they |