Handle: Engineering Artificial Musical Creativity at the “Trickery” Level - Computational Creativity Research: Towards Creative Machines - page 278

14.1 Introduction and Plan

We briefly articulate a “middle-standard” position on machine creativity as measured
against human creativity (§ 14.2 ), and report on the status of our decade-long attempt
to concretize that position in the realm of music, specifically in the form of a multi-
talented intelligent agent: Handle. Our middle-standard position falls between what
we see as two extreme positions on machine creativity: on the one hand, Cope's
position, which (for reasons we briefly explain below) is acutely “low-standard,”
and on the other, a position articulated by Bringsjord et al.: a “high-standard” one
according to which it's probably impossible that any standard computing machine
could be creative.
After placing our stake in the “middle ground,” we summarize three versions of
Handle, one for each of three types of muscial creativity from among the full array
we ultimately seek to reach. Our ultimate goal is simply stated: For every significant
and determinate type of musical creativity seen in the associated human sphere of
music, and every level reached by humans in each type, Handle will match the type
and level—in the “middle-standard” sense of 'match.'
As to the three types discussed in the present, short chapter: First is Handle as
conductor (Handle cond ), introduced at C3GI 2012 in Montpellier, France (§ 14.5.1 ).
Handle cond is designed to be able to interpret and provide feedback on a human
performance of a score-based, classical solo piece of music. Second is Handle as
jazz musician (Handle jazz ), summarized in Sect. 14.5.2 , engineered to able to join
humans in the production of improvised, “free” jazz. Third isHandle as filmcomposer
(Handle film ); here our engineering, still very active andmore suggestive thanmature,
centers around the composition of music to suitably accompany narrative expressed
cinematically. Handle film is summarized in Sect. 14.5.3 .
One of the important hallmarks of the Handle project is an unwavering com-
mitment to exploiting the power of formal computational logic within the highly
intuitive domain of music. Accordingly, the music calculus ,

, a first-order sorted
modal logic, has been invented as a means of modeling certain aspects of musical
creativity and cognition. A significant portion of the sequel (viz., § 14.3 ) is devoted
to the presentation of parts of this calculus.
We turn now, as planned, to staking out our aforementioned “middle standard.”

M

14.2 Our “Middle-Standard” View of Creativity

In earlier work, Bringsjord, joined by Ferrucci and Bello, argues that while machines
can't be genuinely creative, at least in the literary sphere they can nonetheless be
engineered to seem to be creative [ 8 ]. This two-part position is partly philosophical in
nature (based as it is upon a priori reasoning), and partly engineering-centred (based
as it is upon producing a computational artifact capable of generating compelling
short-short stories [ 8 ]). On the philosophical side, in order for a machine to be
genuinely creative

( creative G ) , it would need to pass the so-called “Lovelace Test”