(no subject)
Hey there,
Hope you're well...
I'll send you another email about connecting this summer while
I'm in the US if you'll be around...
This email pertains to the AGI-16 / Human-Level AI conference
which occurs upcoming week in New York (July 16-19).... at the New School.
We are doing the last-minute nitty-gritty conference planning and
find ourselves short on cash for snacks and for drinks at the
conference reception / poster-session ... (the New School venue
proved more costly than planned... we chose New School as an old
college friend of mine is a prof and administrator there...)
So I'm wondering if you might be willing to do a US$3000
sponsorship (we have two other sponsors at the US$3000 level: Vicarious Systems,
and Demiurge Technologies).... (We don't actually need the $$ wired
immediately, a commitment is enough)...
We can then serve everyone "drinks and snacks courtesy of Jeffrey
Epstein Foundation", or whatever foundation name you prefer (or
just keep it quiet if you prefer)
This is not insanely critical, but it would be a help... I
believe these conferences are still quite valuable in terms of
keeping free and open dialogue about AGI going in the research
community, as opposed to having AGI work entirely sucked into the
proprietary halls of big companies...
thanks
Ben
--
Ben Goertzel, PhD
http://goertzel.org
Super-benevolent super-intelligence is the thought the Global
Brain is currently struggling to form...
Just re-sending with a more appropriate subject line...
Hmmm...
I have been thinking a lot about music too, but from a different perspective...
Do you know this paper on Combinatorial Music Theory?
paper:
http://andrewduncan.net/cmt/
video:
https://www.youtube.com/watch?v=OMDtp89Xqlw
It's extremely good, it clarifies some basics of music theory in a mathematical way without the confusing archaic terminology of standard music theory... and suggests various interesting new directions...
This paper is also related and somewhat worthwhile
https://www.math.washington.edu/~morrow/336_09/papers/Ada.pdf
though neither as easy-to-read nor as interesting...
What these papers are getting as are the hidden abstract-algebraic and information-theoretic structures underlying melody and harmony...
What you're talking about regarding symphonies is sort of like "narrative structure" in stories it seems like -- or "discourse planning" in dialogue -- i.e. it's higher-level
That level is important, for sure. But what the papers I've linked above are getting at is a bit lower-level --- more analogous to the syntax and phrase/sentence level semantics of language, I would say.
The algebra and information theory of melody and harmony have to work out at the level of individual chords and melodic passages, for the higher-level development of a symphony or other complex work to make sense...
About
I mashed all of the four symphonies together , > playing recordings of the 3rd 5th 6th 7th all overlayed on each > other,
I have done stuff like that before, but not with classical music...
1
EFTA_R1_01563511
EFTA02458928
with my own recordings ;) .... Do you have a sound file of your experiment? I'd be curious to hear...
Interestingly, the "plumbing" work we've been doing to connect OpenCog to the Hanson robots would also enable us to use OpenCog for music composition / improvisation experimentation ... (i.e. we've been dealing with real-time input and output, which we hadn't been doing before)
But at the moment (this opinion is of course subject to revision) I somewhat feel that the basic problem of "what makes a melody /harmony / rhythm pattern sound emotionally and cognitively appealing" has to be solved before one moves on to issues of symphonic structure....
I.e., we probably gotta solve folk music before we can solve Beethoven or Bartok ...
Meyer's old theory of "surprising fulfillment of expectations"
is basically correct IMO (if you don't know that book check it out, you'll love his classical music examples) ... but the subtle point is that the expectations and the surprise we experience are a subtle mix of
-- basic mathematical music-structure factors, as in the vein of the two papers I linked above
-- human psychology factors as modeled e.g. in Joscha's work and in the Component Process Model ( http://boccignone.di.unimi.it/CompAff2015_files/Phil.%20Trans.%20R.%20Soc.%208-2009-Scherer-3459-74.pdf) which we have introduced into the OpenCog / Hanson Robot codebase to fill in certain gaps in Psi
-- human body-rhythm patterns, to do with the way the rhythms of different subsystems of our physical bodies overlay on each other, which we perceive all the time sub and semi consciously and which influence our perception of music; obviously even classical music without a drumbeat is highly rhythmic...
-- more abstract cognitive pattern recognition
So a good melody packs up "surprising fulfillment of expectations" where the expectation and fulfillment are assessed according to a combination of the above 4 factors ... and if you leave any of the above 4 factors out you're going to get lots of "false positives" ...
All of the above factors are there in Chinese, African etc. music as well as Western music, but manifested in different ways...
I can pretty clearly see how to seriously investigate the above but I can't at the moment see a "quick and dirty" way to make huge progress quickly ... while seemingly simple, the question of "what makes a melody appealing" actually wraps up a lot of issues ... BODY, EMOTION, COGNITION ... plus MATH-OF-MUSIC ...
Fascinating stuff though ;)
If you want to design a research programme in this direction I'd be game to participate; I don't want to drop my other OpenCog R&D obviously but music just fascinates me at a heart as well as head level, so I'd be willing to put some time into this in parallel with making the Hanson robots work... I do have a feeling that to really crack the problem of what makes a melody sound good, one has to crack the problem of how feeling and body and cognition work together, which is a key problem for AGI generally speaking...
-- Ben
looking for AI and music. . a melody is a strange object, it is an inter
as well as intra relation with notes and time. . there are inherent
harmonies. and dissonance. . what i wrote to chonsky this moring
today I conducted an experiment encouraged by Noam's wholly justified
aggressive and detailed directives to joscha. . joshcha focused on layers
being developed in the brain . the timing for the development of each
layer being different per species .
I postulate that music might be a frosted window into that structure.
symphonies begin with their first " layer " a theme. in fact , there
might be more than one theme in the first layer , , the second part of
symphonic form is the complex development stage. where those themes
are inverted, deconstructed , reconstructed etc ,and the development stage
takes the most time . in the conclusion of the symphonic form the
recapitulation of all that has come before it forms a " phenenoma of the
piece " a whole ,made up of its smaller concepts . As opposed to
listening to music to record which neuron is firing, as most musciologists
attempt . I propose that the music may be the audible result of those
neurons firing, made possible by a select few who would attempt to notate
those neuronal firings. Beethoven for example.
The experiment . I mashed all of the four symphonies together ,
playing recordings of the 3rd 5th 6th 7th all overlayed on each
other, playing at the same time. - the way a brain might develop. I expected an
ordered noise but to the surprising contrary, IT WAS AMAZING. .
you can hear new "concepts " forming,
il wonder whether in the mind of a blind child , the " music" would
be created even without the visual referencial. but created none the less.
later when the visual can be tied to concepts , the anatomy may be
hijacked to produce sounds . that somehow relate to the concepts. .
I tried to mix music from different cultures- it didn't work. African
does not work with western europe,- chinese works with neither of the
other two. but within the same cultural music ( the brain of the local species
) the mash ups are beautiful.
I would note that computers engage in "parallel processing" only in
order to take a hard problem and break it into its component parts ,
working on each component separately, , here each problem Interacts
and the their resolutions interact in remarkable ways.