Can computer music speak to us?

[selfAdjoint]

Can computer music "speak" to us?

Via the commenter whose handle is Fly, over at the gene expression site www.gnxp.com, I found this moving testimony by Douglas Hofstadter, on the effects of one piece of AI programming.

http://www.unc.edu/~mumukshu/gandhi/gandhi/hofstadter.htm

He finds the consequences of what he honestly reports to be shattering for his self-image. I wonder what implications folks here will find for their beliefs about consciousness?

 

[Amp1]

If in a like manner AI is eventually able to mimic what we concieve of as conscienceness, I think a few people may start to worry or maybe we all should be concerned. Thoughts of the movie Terminator springs to mind and I can't fortell what such an accomplishment would portend. Right now, I'm not worried.

 

[reilly]

As a working jazz musician, I'm not surprised that EMI can do so well, nor am I terribly bothered that it does. (Many musicians are far less romantic about music than music fans.) Good musicians and composers know how to reach an audience, to jerk the heart strings, to get feet tapping, to instill joy, fear, whatever -- it's part of the craft, the trade. So, simulation is not out of the question.

All that eMI is really doing is, after figuring out what makes Chopin, or Prokoffiev or Bach, is to apply the rules of the great masters -- just like students.

The trick, of course, is what are the rules? It's not terribly well known, but during Mozart's time, and Bach's as well, composers, the great one's included, composed according to mathematical formulas. There's a lot of formula and simple algorithms in music, and a lot of knowledge how to work the basic language of music to most any end. If you can tell Mozart from the Rolling Stones, then these days, so can a computer system, and you are off and running to profound, well maybe, machine-composed music.

Personally, I'll take a jazz piano trio in some small bar.
Regards,
Reilly Atkinson

 

[selfAdjoint]

nks for the post Reilly; it's really valuable to have the perspective of a working musician on this. I do think that the algorithm would work on jazz too. I recall John Coltrane's concern as a young man that he needed not to "sound like Bird", and many a sax player of today is eager to sound like Coltrane, or rather like some particular phase of Coltrane's evolving approach.

You are certainly right that the 18th century composers worked within what we now see as a limited algorithm. This was borne in on me on the 27th, his 250th anniversary, when the classical stations played all Mozart all day long. Boy was I glad when that was over! Don't get me wrong; Mozart is truly great, but nothing else for 18 hours rubs your nose in the algorithmicity of his work.

 

[hypnagogue]

A nice piece by Hofstadter, but for my part, this really has no bearing on my thoughts on consciousness. I've always thought those romantic-sounding invocations of being able to write a truly moving piece of music or poetry or whatever as the quintessential mark of consciousness are more wishy-washy sentimentalisms than strong criteria to be taken seriously. I fully expect every last human capability to be matched, and eventually outpaced, by artificial constructions like EMI. But I am not by any stretch convinced that human capabilities bear any necessary connection to subjective experience, so for me, there is no clear implication from one to the other.

I suspect that the mongrel nature of the word "consciousness" is once again causing some confusion or obfuscation here as well. As is evident in Hofstadter's writings, he regards EMI as having implications for something more like some uniquely human essence or 'soul' than something like subjective experience (or, if one will allow it, "qualia"). Some seem to equate "soul-like" notions with the word "consciousness," but I think this is more an obstacle to understanding what is arguably the core phenomenon (subjective experience) than anything else, and should be avoided.

As an aside, I do not think EMI and other programs like it are as devestating to notions of "unique human essence" as Hofstadter seems to. I would say it's one thing to ape established technique, and quite another to create it. That EMI can write a convincingly Chopin-sounding piece, complete with apparent meaning and feeling, certainly doesn't imply that subjective meaning and feeling were completely irrelevant to Chopin's own work. A more convincing (albeit probably implausible) demonstration along these lines would be if we could feed a program information about typical musical pieces and music theory and so on in Chopin's time, after which the program could independently synthesize a distinctive, novel, and affecting style all its own, just as Chopin did. (Of course, such a demonstration would only have teeth on the assumption that the resultant program would not itself be conscious-- certainly no guarantee.) I have absolutely no doubt that such a program could be created in principle, but I am also pretty confident that it would need to be rather more complex than something like EMI in order to pull this off. So, at least the apparent threat of a demeaning reduction of glorious human complexity to an exceedingly simple (relatively speaking) set of syntatical rules seems to be safe for now.

 

[russ_watters]

Just saw this thread...

As a working jazz musician, I'm not surprised that EMI can do so well, nor am I terribly bothered that it does. (Many musicians are far less romantic about music than music fans.) Good musicians and composers know how to reach an audience, to jerk the heart strings, to get feet tapping, to instill joy, fear, whatever -- it's part of the craft, the trade. So, simulation is not out of the question.

All that eMI is really doing is, after figuring out what makes Chopin, or Prokoffiev or Bach, is to apply the rules of the great masters -- just like students.

As a former high-school jazz musician, I agree with you(I'll ask my roommate too - he's a working jazz/classical bassist).

I hate to invoke Star Trek, but has anyone seen the episodes of TNG where Data plays music? People tell him he's expressive, but he points out that all he is doing is mimicing and mixing-and-matching styles of different artists. I see no reason why a sufficiently capable computer couldn't do just that and be convincing about it. Being truly original, though, would be a lot tougher.

 

[octelcogopod]

If the machine can't enjoy the music itself, why be worried?
If a machine was more capable of enjoying music than a human, then maybe I'd agree with him, but just because a machine can create music based on formulas does nothing to me.

As an electronic musician myself, I know that a lot of times, i will stumble upon something without really realizing it fully until after I've soaked it in for a little while.
Like a guitar riff, a bass line, a sample, and these could have been created by the computer.

As for consciousness and its uniqueness..
Music is math, as such the way we perceive music is b uilt into us physically, so I see no reason why a machine can't replicate it.

There's two ways to create beautiful music;
1. By following a formula of what you think others might like (and if studied long and hard enough succeed at it)
2. Following your own emotions and maybe as russ waters said, come up with some thing truly original.

I don't think a machine will be able to do the latter anytime soon. I'm not saying it's impossible though.
Maybe if it learned synthesis, harmonies in frequencies etc, it could create truly original sounds and thus create something no one had ever heard before.

 

[chroot]

Sunsets, the sound of waves crashing on a beach, the smell of flowers, and icicles hanging from the eaves are all beautiful to us, yet mankind was in no way involved with their creation. They're simply elements of a "set of beautiful things," probably infinite in size. Some elements are "automatically" shown to us by nature's physical processes, some others existed first in the mind of a gifted artist, and still others wait in the dark for someone with an appropriate flashlight to stumble over them.

In this case, the scientists have built such a flashlight in the form of a computer system, and are simply exploring parts of the "beautiful set" that have not yet been found by anyone else.

So what? We have no evidence that the "beautiful set" is limited in extent, or even that we humans are capable of exploring it completely with our minds. Why not use tools to achieve what we cannot with cerebrum alone? We've already seen that computers can make beautiful, immensely detailed pictures -- the Mandelbrot set, for example -- the likes of which had never been seen before without their computational prowess. The Mandelbrot set almost seems "programmed" into the nature of the universe, its endless complexity and intrigue arising from what is essentially a tiny lump of mathematical symbols. At least in the case of the Mandelbrot set, it is as if beauty itself can be distilled into an algorithm a child can perform.

Is it possible that all other elements of the set of beautiful things have equally simple underpinnings? Is it possible that Picasso and Mozart were unwittingly following patterns that can be adequately described by nothing more than a hard, crystallized formula? If so, does it really make their achievements any less extraordinary? I don't believe so.

This article strikes "close to home" for me, as I did a science fair project in high school in which I applied artificial intelligence to recorded music. I simply trained several artificial neural networks with the "sonograms" of various sorts of music, and then let them run on their own to see what kind of music they produced in vacuo. While I could only describe my networks' contributions to the musical world as "uninspired," they clearly sounded like the artist they had studied.

While my techniques were doubtless much less refined that those who produced EMI, it was still eerie to discover that the difference in "flavor" between Bach and Mozart could apparently be distilled into nothing more than a few hundred coefficients in a weight matrix.

Simple is beautiful, I suspect, and all beautiful things have, at their core, some crystallization of simple genius.

- Warren

 

[selfAdjoint]

For me the articles raises the question, if software can capture the "feel" of a composer's works with a statistical algorithm, then what other facets of our cherished personalities can be captured this way? That was the point of my "less there than we suppose" comment. It wouldn't take a very big parameter space to snow our ability to know each other. Recall how "realistic" the simple ELIZA software appeared before we "knew the secret". and that was not based on hard analysis at all.

 

[Moonbear]

Anyone who has studied any music at all, or even taken a music appreciation class, will realize quickly that music follows fairly clear rules, as does the emotive response to music. Finish a song on a major chord, and it feels complete, and satisfying. Finish it on a minor chord and you feel a bit unsettled. Include a lot of minor chords, and the song takes on an eerie quality.

You have to also consider that the reason you can identify a composer's works as all having his "style" is that there is a bit of a formulaic approach to it. It doesn't sound surprising at all that a computer could identify the patterns that give a particular composer their "signature" sound, and emulate it, sort of like a theme and variations work would do.

Every once in a while, just to amuse myself when I hear a new song (and when nobody else is around to listen), I'll try to hum along despite having never heard it before. It's not really that hard to pick up the patterns and predict where it's going. Sure, I'm not going to nail every note that way, but what makes it pleasing is when the next note is what your ears want to hear.

So, I don't really see much tie-in with consciousness here. It might be interesting to ponder why certain sounds evoke certain emotions, especially in such a predictable way that composers can take such advantage of this. It might even raise the question of whether a great composer is a talented artist, or really just someone who has a strong grasp of the rules and can readily adapt them to original pieces? Perhaps there's a more mathematical/analytical aspect to composition than just purely art.