11 Feb
2009
11 Feb
'09
7:09 p.m.
On Wed, Feb 11, 2009 at 10:03:15PM +0000, Chris Cannam wrote:
On Wed, Feb 11, 2009 at 9:01 AM, Frank Barknecht
<fbar(a)footils.org> wrote:
'Good' automatic accompaniment will require more than
the computer's ability to listen combined with some
condensed programmed knowledge of musical structure
and performance praxis.
By 'good' I mean that the computer should be able to
do it as expected, and without requiring specific
instructions for each particular case.
I suspect it would require 'training' of the same
form that a musician capable of the same would go
through.
Now if such ability is the result of training, can
it be called either 'manufactured' or 'simulated' ?
If yes, should we humans consider ourselves as being
manufactured or simulating ?
Ciao,
--
FA
Laboratorio di Acustica ed Elettroacustica
Parma, Italia
O tu, che porte, correndo si ?
E guerra e morte !
Beat and rhythm detection also is a hot topic for
many years in the
academic music scene as it's a necessary prerequisite to let a
computer play along with human performers, ideally in real time.
I actually find this goal more distasteful than the one Fons is objecting to.
I can understand why people are interested in automatic accompaniment,
but mangling human-made music to fit a specific timing map seems like
a more proper goal to me -- "manufactured" being somehow preferable to
"simulated". (I think this may be an "uncanny valley" situation --
if
you haven't met this term before, look it up.)