Music & Emotion.

Anybody who thinks music cannot express emotion has obviously never heard this:

Just listen.

It doesn't have anything to do with listening.

First, music does not have emotions or intention. Music is just organized sound. Thus it is wrong to say that music expresses anything. Music is not the musican. Music is what the musician does.

This isn't just a minor point.

There's a world of difference between "music expreses emotion" and "a musician expresses emotion using music."

Music doesn't express emotion anymore than a hammer builds a house.

Musicians can use music to express emotion, but they do so in the context of their culture and their own sense of musical norms.

Music is as culturally bound as language is, in some ways even moreso.

A listener who has no cultural reference to the music will not understand the emotion the musician is trying to convey anymore than someone with no linguistic or cultural reference would understand what a poet is trying to say.

Music is an artform, as such it can and has been used to express the whole of human experience, just like any other art form. But you can't hear what is being expressed unless you already have some cultural reference to the artist.

Assuming no lyrics are used to set a mood: music on itself doesn't have emotion, nor is it a method of communication. However, our simple primitive brains do react in certain, rather consistent, ways to certain combination of notes, movenements and sounds. For example, making slow but big bends with an equally slow vibrato would be seen as highly emotion, whereas the same notes played with slides would be far less emotional. Playing the A-major chord sounds more happy and alive then the A-minor chord, even though there is no real reason why we react so predictably to such relatively minor differences in sound.

As such my personal believe is that ultimately the best music will be composed and performed by computers. And my estimated guess is that this will be happening within the next thirty years.

I have to wholeheartedly disagree with your last statement, Arjen. Computers have no heart and no soul, all they can do is spit out mathematical sequences that may very well sound like an intricate musical piece, BUT, IMHO, truly great music always has been and always will be a product of the human soul. I simply don't believe that a computer will ever make more beautiful music than humans have and will continue to produce.

When I initially read your post the first thing that popped into my head was a memory from school when I was about 8. We were told to put our heads on our desks and to close our eyes. Our teacher put on a recording of Sergei Prokofiev's "Peter and the Wolf". I hadn't heard or read the story yet, so I had no preconceived notions in my 8 year old mind. The teacher simply told us who the composer was and the title of the piece. We put our heads on our desks and closed our eyes and listened. It was the first time in my life that I was immediately transported to another place and time simply through the power of music. I experienced every emotion that's possible for an 8 year old girl to experience through listening to that one piece. I was exhilarated, frightened at times, fascinated, angry, sad and ecstatic and all of that was conveyed to me through music.

That emotion, I believe, could only have been stirred because of the passion that Prokofiev infused into his music. I don't think that, had it been produced by a machine, I would have had such a wonderful experience.

IMHO.

I don't really believe in absolutes, only perceptions. So if you take any arrangement of notes and attach and emotion to it, then it does have emotion. So I don't believe music expresses emotion, but people will attach certain emotions to it. Those emotions may be different from what the composer intended.

People have weird ways of rating music. I guess emotion is one of them. Another weird one is how hard it is to play. Apparently the more difficult it is, the better it sounds.

A daresay people thought 30 years ago that by now computers could support autonomous intelligences and be capable of creating thoughts and art. You may be a tad optimistic.

Then those people were wrong. Maybe I'll be just as wrong, maybe not. I personally believe that my expectations are very reasonable: a computer that could generate music based on a list of requirements the user created. When I look at rap, britney-pop, poppy-punk, trashmetal and many more genres I have to note that these are all, composition-wise, fairly simply structured and organised. For example, as a simple start: it can't be that hard to have a computer generate a chord progression, have that played by a piano using one of the usual ways to play them, add a drum patters with a fill every four measures and generate a melody over it. All the user has to do is add some words to the melody and your very own keane-song is made. Have the computer generate it in MIDI and you have the complete backing with some proper plugs done in less then a minute. This is already possible and I can't see why we can't expand on this. I don't mean that in thirty years computers can spit out album after album of excellent classical music or free jazz improvisations, but some genres should be well within our reach within thirty years.
Arjen, what you are describing is the emotional effects of sound, not music, on a part of our brain which is programmed to respond to such basic aural stimulii. A mother's anxious response to the cry of a child, being a classic example of this phenomenon.

Music itself is sound, and as far as I know we do not have seperate brain sections that deal specifically with music and no other aural stimuli. The way we respond is often just as fixed as with the example you gave.
You asked "Why does a minor chord sound sadder than a major?" The most widely accepted explanation is that it's simply because of the semitone clash between the minor third note and the root's 4th overtone, which is a major third. It causes not so much a feeling of sadness but a feeling of disturbance - as all dissonance does.

That's not an answer. A c-minor will sound more sad or melancholic then a c-dim or c-sus chord, even though they all contain dissonance, one way or the other. And besides, the question remains *why* we respond in such a way to dissonant and consonant sounds and how consistent that response is. If the response is consistent we should be able to make a grand database listing how humans generally respons to combination of sounds. If it is not consistent the entire idea of emotion is utterly pointless since any target audience is likely to respond in totally unpredictable ways. Or in other words: either there is a pattern and computers should be able to come up with something, or there is not meaning emotion in music happens by mere chance.
Why does a wide slow vibrato bend sound more emotional than a slide? Who knows? but whatever the reason, it will be just another simple physiological and psychological 'cause and effect' process.
This has little to do with music. The creation of music involves using those psychological and physiological responses as building blocks to create that which can be truly imaginitive, emotional, spiritually uplifting, and most of all 'human'.
A computer may give the illusion of producing music, just as a hypnotist can make you, or even an audience, believe you can play like Andres Segovia.
But the computer is not creating music - only musical sounds. Without an artist there's no art because one absolutely indispensible factor in the creation of all great art is the emotional response of the creator. If a computer is ever designed with true (not just simulated), emotion and imagination then I'll concede otherwise.

The computer needs no emotion for it. If there is a natural consistency in the human response to music then an artist/programmer could specify what kind of emotion is required, and the computer could come up with the desired sound. The music I think of would be vastly different from 'human compositions' but could appear just as powerfull and emotional. I don't think waiting for emotional computers is the key, the computers should just execute what emotional programmers want them to do.
Music, is interpreted completely differently by the brain, which, far fom being simple and primitive, is in fact the most complex entity in the known universe - by far.

What do you mean with this? Music is more complex then H2O, a superconductor or genetically modified banana? Why? What do you compare it with? And why is music interpretated differently?
I have to wholeheartedly disagree with your last statement, Arjen. Computers have no heart and no soul, all they can do is spit out mathematical sequences that may very well sound like an intricate musical piece, BUT, IMHO, truly great music always has been and always will be a product of the human soul. I simply don't believe that a computer will ever make more beautiful music than humans have and will continue to produce.

Well, to be perfectly honest, I don't believe that much in the concept of the 'human soul' either. But that's another discussion altogether.

Or in other words: either there is a pattern and computers should be able to come up with something, or there is not meaning emotion in music happens by mere chance.
Your logic is cold and unrelenting... exactly as logic should be :)

I agree with every word you've posted on this subject, but the above quote especially - it's the inalienable conclusion and closes the case completely, IMHO.

the same music can trigger a hundred different responses in people.
our understanding of the brain is relatively limited and we'd have to have a large # of volunteers all willing to allow themselves to be subject to these different computer noises, or else we'd have to have a totally accurate artificial brain that a computer could run a million simulations on very quickly
also, music depends a lot on what has come before it. today, early beatles sound like bubblegum pop, but at the time, it was revolutionary.
finally, and most importantly, a computer would have to have imagination. it would have to be willing to try new things unthought of by the programmers, or else it would merely be combinations and permutations of what we already have, which might in fact produce great music, but i don't think it would make anything totally original. there would have to be new inputs or new music being created elsewhere.

the same music can trigger a hundred different responses in people.
our understanding of the brain is relatively limited and we'd have to have a large # of volunteers all willing to allow themselves to be subject to these different computer noises, or else we'd have to have a totally accurate artificial brain that a computer could run a million simulations on very quickly
also, music depends a lot on what has come before it. today, early beatles sound like bubblegum pop, but at the time, it was revolutionary.
finally, and most importantly, a computer would have to have imagination. it would have to be willing to try new things unthought of by the programmers, or else it would merely be combinations and permutations of what we already have, which might in fact produce great music, but i don't think it would make anything totally original. there would have to be new inputs or new music being created elsewhere.

Thank you! I couldn't have said it better! :D

Or in other words: either there is a pattern and computers should be able to come up with something, or there is not meaning emotion in music happens by mere chance.
Your logic is cold and unrelenting... exactly as logic should be :)

I agree with every word you've posted on this subject, but the above quote especially - it's the inalienable conclusion and closes the case completely, IMHO.

There are plenty of paterns in any semiotic system.

However, computers aren't constructing viable poetry, painting great paintings, or even composing interesting short stories.

And linguistic systems are far simplier than musical systems when it comes to semiotic scope.

Even in artificial semiotic systems, such as computer languages, computers in general are not writting paritcularly compelling or nuanced computer programs.

The reality is that there is something deeper going on here. It's captured in the chinese translator paradox.

Consider if you have a computer that has an algorithm to perfectly translate chinese into english. You feed it in text in chinese, and out comes a perfect translation in english. The AI practicioner would claim, and Arjen would agree, that the computer understands chinese and english.

No, you put the computer in a big box.

Then you remove the computer and put a person in there who doesn't speak a word of chinese. You give them the same rules the computer runs by, feed them the same chinese text, and they follow the rules and faithfully produce an english translation.

Now, does the person understand chinese?

The same paradox is even more of an issue when it comes to semiotic structures in art. What constitutes "good" music is "just a pattern," but it doesn't follow that that pattern is algorithmically discernable.

Semiotic structures are not the simple pattern development that Arjen presumes they are.

A daresay people thought 30 years ago that by now computers could support autonomous intelligences and be capable of creating thoughts and art. You may be a tad optimistic.

Then those people were wrong. Maybe I'll be just as wrong, maybe not. I personally believe that my expectations are very reasonable: a computer that could generate music based on a list of requirements the user created. When I look at rap, britney-pop, poppy-punk, trashmetal and many more genres I have to note that these are all, composition-wise, fairly simply structured and organised. For example, as a simple start: it can't be that hard to have a computer generate a chord progression, have that played by a piano using one of the usual ways to play them, add a drum patters with a fill every four measures and generate a melody over it. All the user has to do is add some words to the melody and your very own keane-song is made. Have the computer generate it in MIDI and you have the complete backing with some proper plugs done in less then a minute. This is already possible and I can't see why we can't expand on this. I don't mean that in thirty years computers can spit out album after album of excellent classical music or free jazz improvisations, but some genres should be well within our reach within thirty years.

Arjen: No argument that computer generated music is possible already and it will advance; but this is somewhat softer than your original assertion of "ultimately the best" within an estimated 30 years.

As such my personal believe is that ultimately the best music will be composed and performed by computers. And my estimated guess is that this will be happening within the next thirty years.

our understanding of the brain is relatively limited and we'd have to have a large # of volunteers all willing to allow themselves to be subject to these different computer noises, or else we'd have to have a totally accurate artificial brain that a computer could run a million simulations on very quickly

All this means is that there has to be a substantial financial purpose behind developing the kind of program I 'envision'. See below for more about it.
also, music depends a lot on what has come before it. today, early beatles sound like bubblegum pop, but at the time, it was revolutionary.
finally, and most importantly, a computer would have to have imagination. it would have to be willing to try new things unthought of by the programmers, or else it would merely be combinations and permutations of what we already have, which might in fact produce great music, but i don't think it would make anything totally original.

99.99999% of music is totally unoriginal, whether we're talking blues, classical, rock, rap or pop. One dude comes with a nice idea and if it is a success the rest of the world will clone it. And for good reason, because most people really want more of the same once they like something. When you take a look at the big-earning labels, which are the organisations with the resources to develop such a program, you'll see the core of their income comes from the Britneys and Christinas, th 50cents and Eminems, the Keanes and Timberlakes. This is all music that is relatively simple, unoriginal, bland, generic and amazingly succesfull. It's the music where the big bucks are earned and it's the music that would be most suited for computers to generate.

Music is, sad but true, a product. The reason most music exist is because people earn their money with it, directly or indirectly. As such, people will do a lot to be able to produce more with less costs, as any businessman would do. So there is a need for such a program, the resources are availlable and there are no technical problems which should make it impossible, as far as I can see. As such it is only a matter of time before this situation will become real, the only thing that can hold it back is the cost of development, and that price will go down every year as science progresses.
there would have to be new inputs or new music being created elsewhere.

There always will be. And after this new genre is created it will take only one person to analyse it, structure and organise the way the song moves around and put the general rules into the program. After that you can pump out album after album of 'cloned' music that will not be original but will bring in lots of money with very little expenses.

Note: this 'basic' version of the program is not a supercomputer that can generate new and perfect music, that's something for the far future. But a program that would allow someone to create songs in a number of popular genres with minimal work and maximum efficiency is not that insane, I think.

Greg: That depends on how you read 'best music'. In this context it means 'most financially succesfull song' instead of 'most artistically interesting song' or such, since the former can be objectively measured and the second can not.
Semiotic structures are not the simple pattern development that Arjen presumes they are.

Maybe not. But as far as I can see most of the top100 songs right now are infinitely more simple then your average story or painting. And remember I am not looking for computer generated art but popular music, music that's already being 'generated' in a sense by human composers.

As for the paradox: music isn't a language, no matter what people say. If you strip the lyrics from any song there is no way anyone can ever get the story out of it. Music is like an inktblob, you hear or see whatever you want to hear or see. As such music has to follow general principles more then exact rules as most of the experience comes from the interpretation of the listener, more so then with any other art. As such there needs to be less of a literal understanding and purpose behind generic songwriting so in a sense the computer needs not understand it. (and besides that, I would not say the computer understands language, it's just able to process it in a usefull way)

Consider if you have a computer that has an algorithm to perfectly translate chinese into english. You feed it in text in chinese, and out comes a perfect translation in english. The AI practicioner would claim, and Arjen would agree, that the computer understands chinese and english.

I doubt either would make this claim.

As for the paradox: music isn't a language, no matter what people say.

It only needs to be a semiotic system.

If music can be used to communicate (convey emotion) then it is in some sense such a system.
I doubt either would make this claim.

Actually, it's a perfectly accurate presenation of the views of what are called hard AI proponents. Which, if you are claiming a computer can be creative, is the claim being made.

Arjen's original claim, to which I was responding, is that music generated by computer will be judged to be the best music available. That is most certainly a hard AI claim, as it is claiming that computers will be able to produce artistically meaningful works repeatedly.

That requires that computers be capable of manipulating semiotic systems. Problematically, syntax is not the same as semantics.

Arjen's original claim, to which I was responding, is that music generated by computer will be judged to be the best music available. That is most certainly a hard AI claim, as it is claiming that computers will be able to produce artistically meaningful works repeatedly.

No, what I'm saying is that ultimately the best music does not require originality or creativity but merely a flawless knowledge of the human brain and the ways sound is processed. Since we do not have such knowledge right now we depend on originality, inspiration and such to work around it, but results aren't very consistent. Now, it is very music possible that *you* won't agree it;s the best, but I don't think you agree with what the Grammy-jury decides each year either.

The really innovative stuff therefor won't be much in the program as in the research required to build up the knowledge database. The program just needs to process that according to pre-programmed rules based on what genre of music is required.

If music can be used to communicate (convey emotion) then it is in some sense such a system.
We're going in circles here...

There isn't going to be any major labels in 30 years, so why would we want to generate the crap they put out today?

Because tons of people like that 'crap'. And major labels will always exist because musicians are in general just too stupid to manage without them. Sad but true.

(yeah, had to spice up my 3500th post somehow ;))