This story at Wired by Eliot Van Buskirk is pretty wild, tickling both my interest in artificial intelligence and far out policy implications. Startup Zenph Sound is modeling past performances by well known musicians to understand their style and mechanics. Right now the applications are limited to faithful reproductions, useful for high end player pianos and producing clearer new recordings to replace age degraded sources.
While they cannot today produce original, new music that is part of the long term plan. One idea branching off from that would be to produce interactive versions of famous musicians with whom users could potentially jam.
Eliot does a pretty good job of digging into the copyright situation with the re-recordings and looking forward to wholly original new works by these virtual musicians. The approach Zenph is taking is pretty conservative, basically looking at what they can produce today as belonging to the heirs or assigns of the original artists. The latent ambiguities for future applications go far beyond the simple question of why Zenph is being so timid. In the past, similar innovations have defied original rights holders on the basis that entirely new markets were spawned that expanded value rather than taking away from existing revenue potential.
Far more interesting is the question he begs about licensing of musical style, something that today isn’t even addressed by copyright law on the books or common contractual practices. There may be a moral dimension to honoring the descendants of a great musician but thankfully Eliot added an update clarifying Zenph’s view. They are not suggesting or pursuing any ideas about a new right covering musical style. Whether the heirs agree in the long wrong remains to be seen.