Can a chess-program compose music? [Archive]

nickysnd

12-05-2006, 05:36 PM

Quite an inciting issue, this chess thing, thanks Frakir - I like this sort of debates. :)

Well, that’s what they have reached so far: programming basic (and more advanced) rules of music theory.
No, they are not programming music, they just think they are :) . As I said, first they should be able to define music, because computer only work with definable concepts (Yes, No, Run, Stop, Zero, One, etc.) Teach them music? Impossible. I mean - What is music? After the programmers will understand for themselves and clearly define what music is, then they may start to think whether or not it can be programmed. Till then - nix, nada, they program something that only they will call "music".

The rules we can also fall back to if we have a very uninspired moment. Like Steedmans set of recursive rules which generate well-formed transformations of basic I-IV-I-V-I twelve bar blues chord sequences.
What is an uninspired moment? A moment in which one is less able to focus on the required tasks? Well, then, in such moments, *you* just leave it for some other time, or take everything a bit slower, that's all. There's no excuse for lowering *your* musical standards while blaming inspiration for not paying *you* a visit! :) .

Like the first-generation chess programs, Ludwig uses these rules to calculate all possible variations (just a matter of computing power), and then uses another set of rules to pick the best variation from the lot.
As SergeD pointed out: "A chessboard has physical limits, limited rules and one mission to accomplish. Music has no limits, no rules and nothing to prove." Brilliant and concise! So "Ludwig" (offending choice of name :mad: ) can apply whatever rules and compute anything he might be able to - the result will not be music.

I know where this misunderstanding comes from: it comes from the myth that music is made up with notes and various characteristics of notes. This is a common misunderstanding even among musicians. And it's very easy for binary minds, like programmers ;) , to buy and use this myth. In fact, very few people come to understanding the fact that composers do not make music out of notes, just as well as playwrights do not make up dramas or comedies with letters. Period. Is just as simple.

Eg. - There are about 30 letters and punctuations, and very few grammatical rules in English, or in any language (few - considering the computer ;) ). So - Can *you* make a computer invent a new joke? Can *you* make a computer to write something that would make millions of people feel, in the same time, pity and admiration for poor brave Juliet :( ? Impossible. Now, do you think that music is easier than English? Because I don't. So - no way a computer can make music. Dixit.

So far so good, but the result lacks a spiritual and emotional dimension, as you correctly put it.
Yep.

For the same reason
Nope. It's for a very different reason.

a first-generation chess program couldn’t beat a grandmaster, who also understood the rules of the game but made the difference with creativity.
There can be no analogy between music and chess. None at all. See SergeD's quote above.

Creativity? Chess has absolutelly nothing to do with creativity, chess is nothing but controlling various combinations of 64 little square fields by moving 32 pieces, following a small number of simple rules. So - chess is all about memory storage, fast access to any point of that memory, and computing power to combine those limited variables. A perfect game for computers and for binary minds ;) . It's not the creativity that makes a chess champion, it just takes a computer-like memory plus combinatorial skills. Composers have a whole set of total different intellectual qualities, which computers will never have, I'm pretty sure of that.

So how do chess-programs beat them now? The answer is that, next to the rules, they now have a complete database of tournament games played by the masters to backup their decision making process.
You're right, that's exactly what happens! I have just give a similar answer to how computers beat humans - it's their type of game, requireing binary computations. While music is not a game and it's wholly human.

I guess they will try the same thing with music programs.
What same thing? Applying rules? Music is not about rules, although at times music might use some rules, some other times it ignores all rules - hard to understand how come that some times rules work well, some other times everything sounds just great in the absence of any rule, and some other time one can follow all the rules and it just stinks! Now tell a computer that :p !

Computers already can analyze music,
No, they can't. Computers can't analyze what it's not clearly defined for them. Music can't be clearly defined.

and all music is digitally available.
Not correct - it's the sound that is digitized, not the music. They are different things. And computers can't even understand a much easier concept, as sound. Eg.: I can focus and listen to the second flute line in an orchestral tutti, while a computer simply can't distinguish it and isolate it. Now tell me - how can a computer make sense out of music when it can't perform such a basic task as selective listening? Huh? :)

This way computers can learn from us very fast, tapping from what is called ‘the Wisdom of Crowds’…
No, they learn s:pt from us, they can learn far less that we can comprehend and operate with. As I said, composers operate and are able to make sense (spiritual, emotional) in a language without words, letters, digits, dots, symbols or any other depict-able signs. Computers cannot function in the lack of such symbols. They are just tools - tools can't create anything, as they have no personality, no self-conscience, they are worse than animals, in fact they are just our prolongations. About computing power - I can do all the computations required to write a symphony as I go with it, with pencil and paper. Also, computers cannot take decisions of their own, they only act according to pre-determined rules such as: "once you reach A, if x is present go to B; if y is present go to C", only this sort of pre-written things. Conditioning makes creativity impossible. Composers are not conditioned machines.

Also, composers may often have strong reasons even when they have no words or symbols to depict them. And then, listeners can resonate and sense those reasons, again - even though there are no words or symbols to translate them.

Music can't be compared to a game, I feel embarrassed that a fellow human can possible think that. :o

Music is the most otherworldly presence on this planet, and definitely the only supernatural presence in the known universe.

gstitt

12-05-2006, 06:41 PM

Well said nickysnd.

So how do chess-programs beat them now? The answer is that, next to the rules, they now have a complete database of tournament games played by the masters to backup their decision making process.

Computers can play chess well only because they can evaulate billions of possibilities. This works for chess, because there is usually a well-defined objective function for how good a particular move is.

Like others have said, there is no objective function for measuring how good music is. Ask 20 different people to rank 10 songs in terms of goodness, and you are almost guaranteed 20 different answers.

Sure, a computer can compose. I can write a program in 30 seconds that could randomly outputs different pitches and rhythms. But, this isn't music. Of course, I could refine this program to follow traditional composition practices. Still not music. I could then use random behavior to stray from traditional rules to create new types of melodies, harmony, etc. Still not music. With enough effort, I believe computers could be made to make decent compositions, that are based on common techniques. However, there is no creativity involved here, just a rehashing of what has been done in the past with some randomness thrown in. Unless musical ideas originate from emotion, or for the direct purpose of evoking a particular emotion, in my opinion it is not music.

Computers already can analyze music,
Computers can perform signal processing, which at this point is no where near being able to analyze music.

To do so would require huge advances in AI, which would likely need to be preceded with an understanding of how humans actually process sound. The state of the art in AI is not even close to what the general public thinks it is. AI generally implies massive searches of an extremely large exploration space. So, intelligence in many cases just means "exhaustive search". These massive searches can be effective for clear objective functions, but again it is impossible to define an objective function for something that is very subjective.

josejherring

12-06-2006, 01:14 AM

Deep blue only beat Kasparov when the computer was fed every single one of Kasparov's games to analyze. So in fact Kasparov was defeated by a his own skill not by the independent thought of a computer.

Computers can't generate a spontaneous thought. It can't originate any kind of communication and I doubt that that's going to change any time soon. So a computer will only spit back out what it's programmed to do.

You could program a computer to follow the rules of counterpoint fairly easily. You could also teach it to randomly generate any combination of pitches and rhythms and dynamics and articulations all which are pretty finite as well and can be thought of mathematically. What you can't teach is beauty and ugliness, emotion, degree of emotion, ect. All that takes aesthetic judgment.

I do see over the next 5 years computers being able to generate random compositions that then one can choose the best one suited for the use. That's not hard to see happening. What is hard to see is a computer sitting down and saying, "wow, that's amazing!" Or, knowing when the rules of counterpoint should be broken or not followed at all for dramatic or emotional effect. Personally I think that we're pretty safe as long as we continue to invent.

Chess is a limited game. Music has limitless possibilities. Chess is a game where one can plan the move down to the last remaining pieces. Anybody whose studied chess will soon realize that intellectually it has it's limits and that so many games have been analyzed that patterned moves govern like 75% of the game these days. Music has no such bounderies and those who rely on patterned styles of music soon find themselves not doing much.

fongi

12-06-2006, 02:33 AM

No machine,software or hardware has that magic ingredient.........feeling. ;)

SergeD

12-06-2006, 06:26 AM

Eg. - There are about 30 letters and punctuations, and very few grammatical rules in English, or in any language (few - considering the computer ;) ). So - Can *you* make a computer invent a new joke?

Yes, using endless combinations...
But will the computer laugh ? :confused:

SergeD

gstitt

12-06-2006, 08:35 AM

Yes, using endless combinations...

That is actually correct. I truly random music generator would eventually come up with the Star Wars theme, or Beethoven's 5th, etc. It may take sixteen billions years before doing so, but it would eventually happen. Actually, it would eventually come up with every piece ever written in addition to every piece that will be written. The problem is, 99.9999999999% of the things it creates would be complete crap and obviously it is not feasible for one to sift through all the junk to find something decent.

nickysnd

12-06-2006, 09:00 AM

That is actually correct. I truly random music generator would eventually come up with the Star Wars theme, or Beethoven's 5th, etc. It may take sixteen billions years before doing so, but it would eventually happen
I don't think so, for the following simple reason: music is not a combination of events, just as well as a comedy or a tragedy is not a combination of letters.

Composers do not combine notes, just as well as playwrights do not combine letters.

When I listen to Beethoven 5th (to take your example) I do not hear notes. I do hear a living entity telling me something that can't be said in words. There is a spiritual dimension that is everything about music. OK, there can be also an emotional dimension. But these two are both out of programmers reach. Because - let's be honnest, we're not talking computers here, we're talking programmers. Computers are only extensions of their programmers.

Music can only be created by composers. Never by machines. Period.

Computers are only storage devices spinning up their information acording to combinatorial/mechanical rules. It's silly to think that a machine can create anything. Music? No way.

nickysnd

12-06-2006, 09:08 AM

Yes, using endless combinations...
But will the computer laugh ? :confused:

I have to say again - no, a machine can't invent a new joke.

You first you have to input your application with the following basic info: what is a joke? Is it a combination of words? No - it is a ineffable cultural *thing* that escapes combinatorial rules.

And it's clear that machines can't laugh. The real issue is that people won't laugh to a joke invented by combining letters.

nikolas

12-06-2006, 09:10 AM

Well... you know...

Endless combinations and so on... they can work realyl really nice with aleatoric music, or serial music. :d

I'm not sure that I can hear much emotion in the case of Boulez... :-/
Or some variations on the piano by Webbern, where he has 1 row, and he has a cetral point (A I believe), and everything is mirrored. Given the idea is original everything else is calculations and nothing else...

nickysnd

12-06-2006, 09:17 AM

nikolas

12-07-2006, 01:53 AM

hehe... depends on your definition of music then. :D

I think that it's pretty "easy" to define some rules on which the program will define pitch and duration, thus providing, for example a midi file, or a piano piece. Dynamics could also be defined, if you apply the program "rules" on phychoacoustic, and mainly music structures. But the idea of emotions, and the idea of an idea behind music will be missing.

So for me the next step would be to try and define somehow some basic emotions on whcih the program will base the structure and the whole music. (which either way partly you are doing it by providing rules on dissonance, melodic structure etc...)

For the record if anybody remembers my track "perniciosus", the middle part, which sounded realy mechanic, was made entirely on the computer, without anything to do with me. I just added some basic dynamics and nothing else (which were already shown on the page, according to what the program was giving me...) the rest of the track was based on those two and a half pages.

It is highly mechanical, but I didn't ask the computer to compose, I just asked him to provide 22 rows, with specific characteristic. Should I wanted, I guess I could've added more rules and make it more "composition wise". The rules were as thus:

* in the 22 rows, make sure you use all the notes equaly. That means that in the end you need to end up having used all pitches the same ammount of times... That was the part that was impossible to do...
* Each row consists of 22 pitches.
* There is no problem is repeating notes in the same row, twice, thrice, or however time you want.
* In the row, ALL intervals must be used ones. So while the pitches can be used as many times as the program "feels like" (<-tee-hee), still only once it can make a melodic motion of a 2nd major upwars (and one downwards...)
* It is "forbidded" to end up in an octave interval within 2 intervals. So no perfect fourth and then perfect fifth.
* It is "forbiden" to use the "negative" interval in continuation. So no perfect fourth up and then perfect fourth down.

Note that I didn't care for any duration, as I wanted it to be as mechanical as possible, so all notes were 16ths.

Also note that of course, it took the program (with my limited programing skills) 2 nights to come up with a result. The reason? It's pretty easy to end up in a dead end. Actually it is impossible to end up in a non dead end. The last row, will be impossible to be created, and reach the same ammount of pitches for all priches... :-/

So note, that after a week of trying, I tweaked it and it stoped a couple of rows before hand. That made the program do things in less than a minute :D. So after some tries, were I found satisfactory rows, I went on and put my own last 2 rows, as close as I could to the original rules. The missing notes from this whole process (the pitches that were missing from making 42 of every pitch that is...) ended up as a new theme for the last part of the track.

Now... is this composition?

It did provide me with a solution (and some research on that), on what "melody" to use.

Either way I do think that you cannot alwasy rely on your emotion, and music abilities to come up with "ideas", or "melodies". There will come a time that you will be completyl dry, but yet have to write.

So in the end, I don't care if Deep Blue can compose. But he sure cannot communicate the way that I can with my audience. And no matter what he will be able to "emulate" something the programer tells it, but no more than that.

Sorry for the little O-T. Hope it was interesting at least... :P

johncarter

12-07-2006, 02:48 AM

"I`m going to write loads of codes into the music; I`m going to go crazy with the Fibonacci sequence [a mathematical pattern that is among the clues in 'The Da Vinci Code'].

One of my guys here built me a computer program that just spits out notes with the Fibonacci sequence. But it was boring. It wasn`t a piece of me. It wasn`t real. It had no soul.
"

Hans Zimmer

nickysnd

12-07-2006, 07:45 AM

hehe... depends on your definition of music then. :D
There is not and there cannot be a definition of music. It would be a contradiction in terms. You see, definition, to define, means to limitate. Music is limitless, so it can't be defined. qed

Nikolas, your post was tremendously interesting for me. First - it provided a first-hand experience of a composer using a machine's help, and second - your findings don't contradict my own theoretical approach on the matter.

Now... is this composition?
Yes, of course it's composition, because it was created. The machine just gave you some answers, some raw material that you decided to incorporate in your piece. The machine only had outputs, all the inputs were yours, as well as the final decisions as how to incorporate them in your piece were totally your decisions.

I don't care if Deep Blue can compose. But he sure cannot communicate the way that I can with my audience.
You touched a crucial point here - audience. Music is created to match a cultural context of an audience. Now define "cultural context" to a machine!... :p It's the same reason for what a machine can't invent a new joke.

Machines can't invent anything, generally. They can play chess though... (although I doubt they much enjoy their victories :rolleyes: )

And no matter what he will be able to "emulate" something the programer tells it, but no more than that.
Here's another good point - if a composer decides to program a machine, i.e. to give it inputs (of the sort you gave to your program: use this, leave that out, etc.) - and then the composer/programmer decides to leave as it is the machine's output, in this case he/she takes full credit for the result, and that would be a human composition. Because: 1. the inputs were the composer's choices, and 2. the final decision (whether that piece represents his/her personal message to the audience) is also the composer's.

nikolas

12-07-2006, 08:18 AM

Just a small comment on the definition of music.

I'm not sure I could find an official one (and won't go googling or wiking right now), but for my own needs I've decided that:

Any kind of created system, which includes sounds, is music.

So 4 helicopters arranged by Stockhausen is music to me (but very unhealthy to listen to :D)
A quartet is music of course.
Boulez makes music (but ugly...)
A sound designer makes music. But he usually breaks almost all "rules" of musis and involves very little or no pitch elements, and this is why we (I) tend to mention it as sound designer and not musician.

And so fourth...

also, thanks ;)

nickysnd

12-07-2006, 08:39 AM

Any kind of created system, which includes sounds, is music.
With your definition you limit music to certain things. As I said, music is limitless, therefore it can't be defined.

Also, I can understand "created" and "system", but what do you call "sound"? Air vibrations? What if I *silently* play something in my head? Would that be also "sound"? Then how would you define that thing inside my head?

I would call that music, even if it's not sounding as "sound" or as anything else.

Also, when I read a musical score, sometimes I can "hear" what's happening there, yet sometimes it makes sense to me at a completely diferent level, not audio, and more than visual, as kind of a weird type of understanding that I don't have words for. I would also call that understanding - music, even if no sounds are involved.

Software

12-07-2006, 10:41 AM

This seems to be music composed by a computer:

http://www.synestesia.fi/

Software

12-07-2006, 10:47 AM

Read more about a remarkable lecture held in the National Art Gallery in Bonn, where ChessBase introduced their latest software development: Ludwig. Da-da-da-daaa (http://www.chessbase.com/newsdetail.asp?newsid=3522)!

The music composed bu Ludwig seems to be quite similar with this

http://www.saundi.fi/

also using database approach.

nickysnd

12-07-2006, 11:52 AM

Software, I won't make any remark on your name ;) , but I can't believe you fell for those frauds... :p

gstitt

12-07-2006, 07:55 PM

I don't think so, for the following simple reason: music is not a combination of events, just as well as a comedy or a tragedy is not a combination of letters.

Composers do not combine notes, just as well as playwrights do not combine letters.

When I listen to Beethoven 5th (to take your example) I do not hear notes. I do hear a living entity telling me something that can't be said in words. There is a spiritual dimension that is everything about music. OK, there can be also an emotional dimension. But these two are both out of programmers reach. Because - let's be honnest, we're not talking computers here, we're talking programmers. Computers are only extensions of their programmers.

Music can only be created by composers. Never by machines. Period.

Computers are only storage devices spinning up their information acording to combinatorial/mechanical rules. It's silly to think that a machine can create anything. Music? No way.

Don't get me wrong, I completely agree with you. You're right in the sense that music is experienced as much more than a combination of events. But, at a low-level it is respresented by clearly defined events. Therefore, a computer (or program to be more correct) could accidentally stumble across a good composition. Actually, it is mathematically guaranteed to come up with every song every written given a purely random number generator (which doesn't actually exist, although some are close) and one hell of a lot of different possibilities (which could take billions of years given near infinite combinations).

I understand that this certainly isn't composition, and this certainly isn't music (except for the rare coincidence that randomness happened to create something good). All it is a dumb use of mathematical properities to try and imitate what humans already do.

Music can only be created by composers. Never by machines. Period.

Until someone can create a program that passes the Turing test, this statement will be true. We certainly don't have to worry about this happening anytime soon.

Software

12-07-2006, 08:49 PM

Until someone can create a program that passes the Turing test, this statement will be true. We certainly don't have to worry about this happening anytime soon.

There can't be The Turing Test for music. Yes, there could be if you limit to a certain style. And that test has been passed years ago. See e.g. David Cope's experiments:

http://en.wikipedia.org/wiki/David_Cope

nickysnd

12-07-2006, 10:10 PM

Alan Turing and David Cope, some wise guys...

Here's Turing the Impenetrable -
Argument From Consciousness: This argument, suggested by Professor Jefferson Lister states, "not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain". Turing replies by saying that we have no way of knowing that any individual other than ourselves experiences emotions, and that therefore we should accept the test. (from http://en.wikipedia.org/wiki/Turing_test)
And here is Cope the Joggler -
My rationale for discovering such instructions was based, in part, on the concept of recombinancy. Recombinancy can be defined simply as a method for producing new music by recombining extant music into new logical successions. I describe this process in detail in my book Experiments in Musical Intelligence (1996). I argue there that recombinancy appears everywhere as a natural evolutionary and creative process. All the great books in the English language, for example, are constructed from recombinations of the twenty-six letters of the alphabet. Similarly, most of the great works of Western art music exist as recombinations of the twelve pitches of the equal-tempered scale and their octave equivalents. The secret lies not in the invention of new letters or notes but in the subtlety and elegance of their recombination. (from http://arts.ucsc.edu/faculty/cope/experiments.htm)
What did I tell you? - Music is a combination of notes, just as well as drama and comedy are combinations of letters. :p

And Mona Lisa is just a combination of colored dots (or pixels... :rolleyes: )

Long live binary minds! :D

gstitt

12-07-2006, 10:56 PM

Great references. :) Which one showed that a turing test had been passed? I am aware of several tests that worked for a while, but failed after some time. I'll ask some AI people tomorrow, but I seriously thought the test had never been passed.

Also, I wasn't trying to imply that there was a Turing test for music. I just meant that until programs can imitate humans well enough so that a human doesn't know the difference, there is no possible way for programs to compose. Sorry for the confusion (and possibly the complete change in direction of this thread). :)

Regardless, we don't have to worry about this happening anytime soon. I'll check out Cope's work, it sounds very interesting.

Frakir

12-08-2006, 02:18 PM

Thank you very much gstitt, nickysnd, software, nikolas, johncarter, SergeD, fongi and josejherring for your reactions so far! I'm sorry I wasn't able to join this very lively discussion, but I enjoyed reading everything that was written here on this subject. Very interesting!

Untill now, most of the discussion was about the present (what are computers capable of today). And I think we can all agree that today's computers can not compose music and that we don't have to worry about this happening anytime soon. But what will they be capable of after a few decades?

It's very interesting to read what Ray Kurzweil (http://en.wikipedia.org/wiki/Ray_Kurzweil) thinks about this subject.
Ray Kurzweil is a pioneer in the fields of optical character recognition (OCR), text-to-speech synthesis, speech recognition technology, and electronic keyboard instruments, and the author of several books on health, artificial intelligence, transhumanism, and the technological singularity . His work often leads into the realm of futurism.

In this article (the coming merging of mind and machine (http://www.kurzweilai.net/articles/art0063.html)) from the Ray Kurzweil Reader, a collection of essays by Ray Kurzweil on virtual reality, artificial intelligence, radical life extension, conscious machines, the promise and peril of technology, and other aspects of our future world, he writes:

"... Within several decades, machines will exhibit the full range of human intellect, emotions and skills, ranging from musical and other creative aptitudes to physical movement. ..."

Also have a look at the timeline (http://www.kurzweilai.net/meme/frame.html?main=/articles/art0274.html)from his book, The Age of Spiritual Machines (a must read if you ask me):

"- 2009: Human musicians routinely jam with cybernetic musicians
- 2019: Virtual artists, with their own reputations, are emerging in all of the arts.
There are widespread reports of computers passing the Turing Test, although these tests do not meet the criteria established by knowledgeable observers.
- 2029: Although computers routinely pass apparently valid forms of the Turing Test, controversy persists about whether or not machine intelligence equals human intelligence in all of its diversity.
Machines claim to be conscious. These claims are largely accepted."

I would love to read your views on the (or our) future!

nickysnd

12-08-2006, 02:45 PM

Untill now, most of the discussion was about the present (what are computers capable of today). And I think we can all agree that today's computers can not compose music and that we don't have to worry about this happening anytime soon. But what will they be capable of after a few decades?
No, it was all about one thing as old as humanity: machines are just tools, like scissors and chisels. And that's not going to change, they'll never be more than that, no matter they're fancy future look.

I am a long reader and admirer of Ray Kurzweil. His inventions and tools have helped many many people, and many people that desperately needed help, like blind people. Also, Kurzweill keyboards are the best sounding synthesizers ever built. I also admire his original way of expressing himself - I love his articles, but I just take them for what they are: brilliant speculations expressed in excellent English writting. Plainly said - that's SF literature, brilliant one, and unique in its own way. But I don't take RK for what he's not, and he is definitely not an oracle. Yes, some people seem to prefer to take his words ad litteram, but c'mon - that's just poetry, like Nostradamus. :p :)

I would love to read your views on the (or our) future!
Even if I'd like this type of mind games, now - comparing to RK articles, any speculation on future coming from me will sound utterly dull! :o :D

gstitt

12-08-2006, 06:11 PM

No, it was all about one thing as old as humanity: machines are just tools, like scissors and chisels. And that's not going to change, they'll never be more than that, no matter they're fancy future look.

I tend to agree, but theoretically in the future we could potentially gain an understanding of exactly how the human brain works, in which case a program could be made to function the same way. I know it is a crude thought, and I don't mean to offend anyone by stating this.

Theoretically, everything we are discussing is pure speculation. Nobody can prove that human-like intelligence is impossible to imitate (without some large assumptions) and similarly nobody can prove that it is possible.

Many AI researchers argue that at some fundamental level humans are just machines (again I'm not trying to offend people's beliefs, just stating other people's thoughts). If this is true, then any Turing-complete machine should be able to recreate it. After all, neurons in the brain work very similar to transistors.

In my opinion, the current state of AI shows how little we actually understand about humans. Take image recognition for example. Humans can obviously identify objects in very short times. Even with powerful computers, the current state of the art of image recognition takes forever and doesn't always work if an object is rotated a different direction, has different proportions compared to the training set, etc. No amount of processing power will significantly improve this. Image recognition sucks because of a fundamental lack in understanding of how human's do it. What is needed is a completely novel algorithm. If algorithms like this can be created, then who knows what computers will be able to do.

Either way, this is a very fascinating discussion. :)

nickysnd

12-09-2006, 02:56 PM

theoretically in the future we could potentially gain an understanding of exactly how the human brain works
Yes - and theoretically in the future we could potentially grow wings on our shoulders and fly like birds.

Nobody can prove that human-like intelligence is impossible to imitate
Yes - and nobody can prove that galaxy-like structures and atom-like structures are impossible to imitate. Also nobody can prove that it is impossible to achieve impossibility.

Many AI researchers argue that at some fundamental level humans are just machines
Many AI researchers already are little more than this.

After all, neurons in the brain work very similar to transistors.
Yes - that's why humans behave very much like tv sets.

In my opinion, the current state of AI shows how little we actually understand about humans.
Yes - and considering that AI employees and advocates are the most intelligent creatures on this planet, we can see that the state of AI is the sheer barometer of human scientific advancement.

Either way, this is a very fascinating discussion. :)
No offense, but aren't you little too fascinate-able? :p :)

Fascinating is bad, it means to cast a spell, to enslave, to subjugate - exactly like fascism.

Here's the naked truth - AI is a myth, and some people like to believe in this type of myths, just as well as some guys like to take advantage on people's beliefs, and to nurture those beliefs. There's nothing scientific to it, it's all total bull spread out towards naive people who take for granted everything that they are told of on the subject, everything they want to believe. It's a sort of a religion, nothing to do with science. AI discourse uses scientific terminology to put people in the haze and to gain diverse advantages as research funds, jobs, titles, political influence, etc.

Before they make the slightest whisper about Artificial Intelligence, those guys should prove
what intelligence really is in the first place,
then what qualifies them to talk with authority about it,
then they should prove their skills to replicate it,
as well as their own intelligence. :p

To me, so far they have only proved their abilities to fool people, something that even plants can do (I mean the carnivore ones of course :eek: :p )

nikolas

12-09-2006, 03:21 PM

Nicky, future is far far away. It's not exactly "right" (<-you know you like that word :D) to be so absolute! ;)

The thing with AI, atm is that it cannot be done with the current HARDWARE. But hardware is evolving as well... We already have machines that repair themselves... Just wait a little while.

Of course AI, is not what Terminator is telling us, but just proclaiming that it's impossible to do -the end-, seems a bit far off.

Certainly AI is no bull. Sorry I won't buy that. :) (Of course it does not exist atm, but still it's not bull!)

nickysnd

12-09-2006, 03:26 PM

I agree Nikolas - the future is opened to any and every possibilties, why don't you give me some funds? - I'd like to research some of them!
:p :D :)

nikolas

12-09-2006, 03:52 PM

I think you're right!

Research can lead to unexpected results: for example!

I take a huge grant to research on computers composing music.
After 5-6 years ofresearch I claim that I failed. :D

What is it impossible for a theory not to be correct after all? :D

nickysnd

12-09-2006, 03:59 PM

After 5-6 years ofresearch I claim that I failed. :D
Wrong! You didn't fail, you're on a good path, so after 5-6 years you put them in the haze with your technologico-musicological mumbo-jumbo, and fool them into doubling their grant, so you can research for yet another 10-12 years! :cool: :D

gstitt

12-09-2006, 04:13 PM

Here's the naked truth - AI is a myth, and some people like to believe in this type of myths, just as well as some guys like to take advantage on people's beliefs, and to nurture those beliefs. There's nothing scientific to it, it's all total bull spread out towards naive people who take for granted everything that they are told of on the subject, everything they want to believe. It's a sort of a religion, nothing to do with science. AI discourse uses scientific terminology to put people in the haze and to gain diverse advantages as research funds, jobs, titles, political influence, etc.

I'm not trying to argue, but why is it a myth? Can you prove it is a myth? Can you provide evidence that AI researchers blindly accept everything they have heard in previous work (given that I know some AI researchers, they would find this offensive)? Are you familiar enough with AI research to make claims about their work? Can you define what intelligence is? I can't either, that is why I am not making any claims about whether it is possible to recreate or not.

Again, I am not taking a stance for one side or another, so please don't be offended. I am only responding because your post makes a very bold claim, and without any supporting evidence you sound like one of these blindly accepting people you were referring to (just with the opposite belief).

Also, I'm not sure why you say it has nothing to do with science. Every paper I have read on the subject is pure science, which proposes ideas and then supports those ideas with experiments. Which studies are you referring to? I'm sure there are plenty that are actually complete crap.

Regarding the research funds, from what I can see, AI is a minimally funded area. I just searched the NSF website, and could only find 5.5 million dollars of funding for the entire field. To put this in perspective, I have received close to this amount for my own research (which is not AI by the way). Edit: Also, none of this money goes into my pocket so it's not like I have incentive to waste research funds like you claim other researchers do.

Before they malke the slightest whisper about Artificial Intelligence, those guys should prove
what intelligence really is in the first place,
then what qualifies them to talk with authority about it,
then they should prove their skills to replicate it,
as well as their own intelligence.

I completely agree, and this is in fact what many AI people, biology people, and psychology people are trying to do.

Fascinating is bad, it means to cast a spell, to enslave, to subjugate - exactly like fascism.

Ok, replace fascinating with interesting. Better? :)

gstitt

12-09-2006, 04:20 PM

What is it impossible for a theory not to be correct after all?

This brings up an interesting point. The inability to prove a theory does not disprove the theory, which it seems like what some people may be suggesting.

nickysnd

12-09-2006, 05:05 PM

and without any supporting evidence you sound like one of these blindly accepting people you were referring to
You got it! :)
That's exactly the point!
With one crucial difference between me and the AI advocates: I am conscious about my own incompetence on the matter, while they try to mask it. Plus, I publicly affirm my incompetence! :p

And yes, a myth is sort of a fairy tale, like Coca Cola's Santa Claus.

HO HO HO !!! :D :D :D

gstitt

12-09-2006, 05:35 PM

LOL, there is actually another difference. You seem to have a good sense of humor. :)

nickysnd

12-10-2006, 11:53 AM

Well, thanx, gstitt. That's a good point with the humor. Emotions in general can't be deciphered, they just can be mocked to some extent, with puppets for example. But without the feel of a human presence behind the puppets - there can be no emotion. Have you seen Being John Malkovich? ;)

In chess, emotions are to be avoided, they are counterproductive. In music, emotions are everything - music without emotions is a contradiction in terms.

Machines are wonderful when it comes to save us from the curse of physical working - and that's clearly the direction things are going. As about artistic expression, inventing, thinking, choices, decisions, this sort of things are exclusively human, they just can't be automated. There is also the issue of self, related to the above (expression, decisions, etc.) Self it is an exclusive human feature. It is also called free will. Can you allow around you an object capable of free will? :eek: (even if that was possible... which fortunately is not. :) )

I thinks this is the word: automation. An object can do only what it is told to, even if that means to tell it: move randomly. The question is how much "random" automation we want to allow to an object.

So much on the mystique of "intelligent, thinking, and creative" objects.

SergeD

12-10-2006, 04:15 PM

Reading this thread about AI reminds me the dove scene, in Blade Runner, where the robot talks on life and death just before passing away. The day Fritz and friends are afraid of being disconnected may be a good start to express their anxiety.

SergeD

Software

12-14-2006, 09:10 AM

nickysnd

12-14-2006, 09:25 AM

Well. Humans are just chemistry!
Living, thinkink, capable of art and love and humor, self-conscious "chemistry". :)

jimmymac

12-18-2006, 03:16 PM

Living, thinkink, capable of art and love and humor, self-conscious "chemistry". :)

Exactly. And if we continue with Software's logic:
A symphony = just vibrations
Devinci's Mona Lisa = just oils
This overly reductionist philosophy (and it is JUST a philosophy, :p ) makes me chuckle.

Anyway,
The only thing that truly amazes me about this is how people get suckered into this AI nonsense.
(edit: Or more precisely, the hype surrounding the AI field in the mainstream media. I know a few AI programmers, they're good guys and honest about thier work's limitations).
But then again, I shouldn't be surprised. How else would cults and such things arise? :mad: And a lot of people swear by astrology.

Regarding Fritz:
I wish people would be more honest about this kind of stuff.
The real composer behind the music that Ludwig 'composed' was actually the programmer (oh look, he a musician himself, how convenient).
That’s why the example tracks sound like "musak". They just set the program to randomise within parameters, or more precisely: within very stringent templates of established musical styles.
They sound like a million other tracks with no personality because, to a certain degree, they ARE just generic 'morphs' of a million other tracks :)

This type of stuff has been around for years. And you certainly don't need brute force processing power to do it or an intelligence.

So could 'ludwig' produce a track in the style of John Williams?
Of course. You could feed in a bunch of his compositions in the form of midi sequences (nice and easy for the program because they come pre-parsed by humans) and it can then regurgitate tracks.

On the surface, many people would be fooled that the computer is 'composing like John Williams'. Just as they are fooled into thinking that the computer is 'composing music' in any creative way.
But it's just a trick. An illlusion.
Sleight of hand by the programmer with the computer merely a stage for his act.

It's interesting to note that you don't see 'joke' generators in any serious capacity because unlike music it is almost impossible to fake on the superficial level with a program.
Whereas a lot of people take music for granted and just accept that the computer is composing music.
I mean, how can we peons doubt the 'experts'?

I work in computer games as an artist. If there were a randomisation feature for Photoshop actions (which would be entirely possible), I could draw a library of human body parts and arrange them in many separate layers within a PSD.
Then if I made the script complex enough (enough variables) and had done a good job on my library of 'construction kit parts' then I could have the action arrange them within the parameters to make a differently posed human figure each time the action script was run.

I've no doubt that it would appear to a layman that the computer is "drawing human figures!!!" (OMFG ITZ SO COOLZ, THE COMPOOTER IS ALIVE!).
Especially if my agenda was to trick people into thinking the computer is doing something smart.

In reality I have simply created a process of rearranging the body parts as per my script. The computer is not drawing anything and it's certainly not creative.
But I bet that people like Kurzweil would delude themselves into thinking that it was a "cybernetic artist in action". Actually, it should be noted that what I described above HAS happened.

Ray Kurzweil purchased a program that in many ways does what I just described above (called "Aaron") The program was made into a screensaver and you can download it if you search for it,
then you can see for yourself how awful it is.
If you were to believe the hype that Ray says on it's website then you would believe that the program is the next coming of Jebus. It's pathetic. :o

Just recently at work we had to take many renders of hundreds of character head models with a multitude of different backgrounds and with variance in camera angles.
So one of the R&D artists wrote a script that randomised within preset parameters in order to render off hundreds of these little pictures.
It loaded up each head model, randomly picked one of the backgrounds I created then randomly picked out of the camera positions I had setup and just spurted out the images, named automatically with a simple convention.
He could spend several months if he wanted to and he could make a much more complex script with many more variables and base components to work from.
It's not magic, and it's not intelligent computers. He's a smart artist who can also code and thats all.

Regarding chess:
As the computer programmer Jaron Lanier said after deep blue defeated kasparov ....
"There was much talk about whether human beings were still special, whether computers were becoming our equal. This way of framing the event is unfortunate.
What happened was primarily that a team of computer scientists built a very fast machine and figured out a better way to represent the problem of how to choose the next move in a chess game.
This accomplishment was performed by people, not machines, and its character was intellectual just as much as it was technical.
The Deep Blue team's central victory was one of clarity and elegance in thought."

It's quite funny to me that if you actually read what the creators of Deep Blue said about Deep Blue, you find that they are quite honest about how it is not intelligent and it is just brute force calculation using a database of kasparov’s games.
They deseve respect for that.

And finally, Ray Kurzwiel may have been a pretty good inventor of keyboards.
But the guy has truly cracked in recent years. Anyone who takes anything he says in his new age 'science' books as proof of anything is being misled.
Ignoring the bad pseudo-science he peddles (such as: more processing power automatically equals intelligent machines), did you guys know that he takes 250 pills per day in an effort to hold of mortality because he thinks he is going to upload his mind into his PC in 20 years? I'm not joking. :rolleyes:

I wouldn't mind if these were all just harmless delusions. Just as I have no beef with anyone’s beliefs as long as they don't harm anyone else. But it's like some new cult is in the making, for example:
http://www.theregister.co.uk/2006/10/27/bbc_horizon/

Throughout human history, people have had the tendancy to anthropomorphise inanimate objects. Whether it be faces in the clouds or spirits in the wind, this is possibly one root of religion.
Yet here we are in the 21st century and people who would probably turn thier noses up and laugh at religion don't quite see the irony as they simply substitute one religion for another (such as Ray Kurzwiel's techno bible).

Now since Ray is so fond of extrapolating trends, he should probably look at the past failed predictions of people like him. The good old climbing tree's to get to the moon fallacy.
I won't hold my breath though.

/Rant over. ;)

nikolas

12-18-2006, 04:22 PM

sh*t that's a big first post! :O

First of all, let's see the differences between some things:

A computer is composing music! Can this be done atm? Not really, or very lousy at most. But is it creating anything? Of course not! It's been programmed to give some parameters enough to be regarded as composition!

So a computer, of course is a totally dumb box, it can be programmed to write something towards music. That's entirely possible!

Let's now examine the fugue! The scholar fugue, as I was taugh some years ago! It's full of rules, strict ones, fomr wise, counterpoint wise and so on! It is soooo strict, that I can bet you whatever you want, if you gave a theme to a computer, it would come up with a wonderful (lifeless) fugue. Computers follow rules! They don't make them atm.

You are attacking AI and all people trying to figure this thing out, for no good reason, for all I care. Of course there is no AI atm, (and Spielbergs movie was BAD beyond words!), but still why total nonsense? If you're looking for a HAL in the near future you won't find one, but step by step the scientist are making progress.

I really don't care about a computer composing music. But I really care about a tool helping me out when I'm running out of time. Is this so bad to consider? I don't think that anyone claimed "I woke up today and saw my monitor open! HAL asked me if I wanted coffee, and then asked me "how about a game of chess"! I told him to piss off and he started laughing, mentioning that he's just a machine, so he can't piss off!" But I can accept claims, that someone can program a computer to provide certain aspects of composition, without cutting out the middle man and without claiming that it was done without the help of human.

Furthermore who would you liketo try and program this? An artist? HAHAHA! (lol). A composer would be the candidate to provide the rules (<-programming) on which the computer will react and give certain results!

But certainly if there is anyone believing that atm computers can "think", or "create" or anything simmilar, they are very much fools, exactly like believing in astrology...

Just can't understand the above huge rant (and so I responded :D)

jimmymac

12-18-2006, 04:32 PM

Hi Nikolas,
I'm not really having a go at anyone... other than Kurzwiel perhaps. If my post comes across as agressive then accept my apologies. ;)
Is it so bad to consider the computer as a tool to help you? no of course not. :D

But that was my point. It's a useful tool. No more, no less. And scientists have made no progress towards making it anything more than a tool. Despite some CLAIMING the opposite.
I just get annoyed when I read the hype surrounding AI. Thats all. So again, sorry for the rant. :D :p

In a way I was getting it all out of my system, if you can understand that. But since the topic was "Can a chess-program compose music?". Consider it my introductionary essay! :)
But regarding the field of AI, there are quite a few people working in the field who are at least honest about their results. And so those people I respect. It's the mainstream pop-science bullsh**ing that gets my goat.

Cheerio,
James

(i think we are both saying the same thing though, only you did it in a more succinct manner :) )

nickysnd

12-18-2006, 04:40 PM

But certainly if there is anyone believing that atm computers can "think", or "create" or anything simmilar, they are very much fools, exactly like believing in astrology...

Just can't understand the above huge rant (and so I responded :D)
I think the first phrase proves that you have understood the main idea, and you do agree with it. It so seems to me.

Jim, thanks for make it so plain! And I think you were a bit too short, in fact... :p

Anyways, coming from a person who works with computers, your arguments have weight.

Welcome to the forum! :)

nikolas

12-18-2006, 04:51 PM

Yes James, we are saying the same thing... (It was the word "hype" that was missing from your post to help me understand ;)). Other than that all is fine.

Oh...

And welcome ;)

And you did teach me a new word: succinct! haha!

jimmymac

12-18-2006, 05:00 PM

Thanks for the welcome guys,

And don't worry, I limit my rants to about one per year these days. :D Or else I would't get any art, music, eating or sleeping done.

nikolas

12-18-2006, 05:01 PM

Yes but the calender year ends in a couple of weeks... ;) We may get a new one on 2007! :D

jimmymac

12-18-2006, 05:06 PM

Yes but the calender year ends in a couple of weeks... ;) We may get a new one on 2007! :D

Blimey, I didn't think of that! :eek: :D

nickysnd

12-18-2006, 05:20 PM

May I draw a few conclusions? The topic was "Can a chess-program compose music?", so -

1. What can a chess program do?
- A chess program can play chess.

2. Who can compose music?
- Composers and only composers can compose music.

3. What are computing machines? What can they do?
- Computing machines are objects, useful computation tools, like abacus. When attached to memory storage devices, they can do calculations with the information present there, and combine it upon request.

4. Can machines help a composer write music?
- No, machines cannot help, they are objects. But a composer may feel the need of some external calculations and combinations, at some point. That's where those tools may prove handy.

.02

jimmymac

12-18-2006, 05:41 PM

Well, from my humble POV:

1: Not in the same way as a human does. In a way it plays chess the same way as a computer composes music. A computers will win chess by doing what a computers does best: rapidly searching a problem space (the same stuff that allows it to rapidly search my harddrive when I give it file parameters).

Humans win chess by doing what humans do best, fuzzy matches and heuristic theories. Computers tend to fail when the problem space is expanded beyond 32 pieces on a 64 square board. :)
It's been said that there is a pyschological aspect of chess where you stare down your opponent. Well, poor kasparov certainly got stared down in a bad way by that lifeless automaton and made some obvious mistakes in order to lose.
He got spooked! And the machine was never at any risk of that. You can no more spook a computer than you can spook your car, or your fridge freezer.
The interesting thing is that a slower computer (your pc, or even an old 386) could play the same quality game of chess as deep blue or fritz (with the same code of course). It would just take much longer to calculate. It's all down to how good the programming is (the human element). Processing speed just helps get the results faster, which of course leads to more productive work from the people using it.

2. I'd say that they can 'rehash' previous compositions using a formula. And thats not enough to be considered composing IMHO. Though some people would argue that James Horner these days does much the same. ;) Even so, it's just a display of programming skill rather than a 'composing computer' :) It's unfair if someone gives the computer credit. So, It can compose music, only if we lower our expectations of composers to 'hacks'.
Sort of OT, but if you guys fancy a really interesting interview with some of the folks at Pixar:
http://video.google.com/videoplay?docid=4617804532425813974
Brad Bird (animator and director of 'the incredibles') has an awesome rant near the end about how people keep thinking the computer does the work! (and you all know how much I like rants) ;)

3. Yep, agreed.

4. Yep, i'd probably agree. I do think the computer is one hell of a tool. I love them and i've used them all my life. A computer will help enhance a humans creativity but is not creative. The error that some people make is assuming that the computer is where the magic happens. Here we have all these programs, samples, synths and yet John Williams with his piano, pencil and paper still outwrites 99 percent of everyone else these days. :)

Here is what one IBM computer manual said once that sums it all up...
"A computer is not a giant brain, in spite of what the Sunday supplements and science fiction writers would have you believe. It is a remarkably fast and phenomenally accurate moron. It will do what you tell it to do - no more, no less."
Thankfully, most of those chaps at IBM know what they are talking about. :)

And with that I think I will stagger off to bed.

SergeD

12-18-2006, 08:17 PM

Throughout human history, people have had the tendancy to anthropomorphise inanimate objects. Whether it be faces in the clouds or spirits in the wind, this is possibly one root of religion.

Funny is in it ?

Mankind at last does not change much

The worst thing that could (or will) happen is that biological life through nanotechnologies open a door of some uncontrolled rise of self reproduction.

The main goal of life is to survive.

SergeD

Composing_automat

12-18-2006, 10:47 PM

4. Can machines help a composer write music?
- No, machines cannot help, they are objects. But a composer may feel the need of some external calculations and combinations, at some point. That's where those tools may prove handy.

Pieces are just dot's on paper. The listeners are humans. It doesn't matter what those dots are. The feelings of listeners is what matters. And if using living instrumentalists, human instrumentalists.

Composing_automat

12-18-2006, 10:56 PM

2. I'd say that they can 'rehash' previous compositions using a formula. And thats not enough to be considered composing IMHO. Though some people would argue that James Horner these days does much the same. ;) Even so, it's just a display of programming skill rather than a 'composing computer' :) It's unfair if someone gives the computer credit.

You are right if every piece is entirel programmed afreshly. Actually human composers just do it that way. BUT composing programs can compose pieces as many times as wanted without any reprogramming. What matters is which kind of "seeds" are used to start the generation of music (=composing)

jimmymac

12-19-2006, 01:56 AM

You are right if every piece is entirel programmed afreshly. Actually human composers just do it that way. BUT composing programs can compose pieces as many times as wanted without any reprogramming. What matters is which kind of "seeds" are used to start the generation of music (=composing)

Hi Composing_automat,
I don' think anyone was saying that they couldn't do that, but it's actually slightly beyond the point that was being made imho. The mistake I think is that people asscribe the process of music composition to the computer as if it has some creativity. When all it is doing is following a human made process. Even entirely fractal music (which tends to sound a bit like static noise on occasion :) ) is just a process. A factory line, only this time the factory line has a lot of branching conveyer belts.

There are procedural tree generation prorgrams and you can get a different shaped tree each time you change the seed of the program. But it's still just a formula, a process designed by a human. And the quality of the tree's depend entirely on how well the procedural tree generation is coded and how well modelled the template branches and elements are (even if they remain described in code only).
Some guys I know are making a procedural building generation system for putting together some of the building's of a game's level. It would be able to come up with a new combination of building block parts each time it is run.
Just change the seed. That does not make it a building block designer, or a building designer. It's simply doing as it is told and is a handy tool for doing a first pass on populating a level.

The computer program won't come up with any new kind of musical style on it's own. So if a novel track comes out of it then that is because of the programmer has come up with a novel track style and coded it.
The programmer defines the rules... the computer plays by the rules. It can do nothing else. :)

It's a wonderful tool. In the right hands.
James

jimmymac

12-19-2006, 02:05 AM

Pieces are just dot's on paper. The listeners are humans. It doesn't matter what those dots are. The feelings of listeners is what matters. And if using living instrumentalists, human instrumentalists.

Taking a human instrumentalist and getting them to play the piece sort of skews the results, as the human player will add his/her own musicality to the performance and elevate it as a result. Not that I'm suggesting that the piece would be unlistenable, but it would be based on rules and rules alone. How creatively those rules are defined could help of course.

I think what I am trying to put across here is that some people make the misguided assumption that computers are going to replace human composers. Not on this site of course.

But the attitude does exist in some people, which is silly.

nikolas

12-19-2006, 02:17 AM

James, the problem lies beyond the composers.

You see it's our clients who matters what they think. There are already rpograms which claim that can compose, many genres and so on... If a director decides to give it a try, a composer will loose a potential job. ;)

So it's not much to what the composers think, but what the general public thinks, and especially game developers, film creators, documentarists, ads creators and so on. And since all these people are constantly trying to save money by finding cheap or free offers, and most of them do think that music is just a background drop in a film or a game, things could be looking rather bad for the future... :(

Software

12-19-2006, 05:24 AM

Hi Composing_automat,
I don' think anyone was saying that they couldn't do that, but it's actually slightly beyond the point that was being made imho. The mistake I think is that people asscribe the process of music composition to the computer as if it has some creativity. When all it is doing is following a human made process.

What is the difference between computer creativity and human creativity. Both start from some impulses coming from somewhere!

Just change the seed. That does not make it a building block designer, or a building designer. It's simply doing as it is told and is a handy tool for doing a first pass on populating a level.

Human creativity is also based on "changing seeds".

The computer program won't come up with any new kind of musical style on it's own. So if a novel track comes out of it then that is because of the programmer has come up with a novel track style and coded it.
The programmer defines the rules... the computer plays by the rules. It can do nothing else. :)

How do you define "style". Most living composers don't have any "style", because taking different pieces no one could recognize the composer. Most modern art music composers define the rules for their own pieces. If a programmer creates SW that is deterministic, but parameter selection is based some seed(s), you get different rules for every piece without any reprogramming. That what creativity is all about.

But don't forget the role of the listener! Listening is also (or should be) creative.

Software

12-19-2006, 05:26 AM

... that computers are going to replace human composers.

That claim is logically absurd. Like claiming that one composer is going "to replace" another composer.

nickysnd

12-19-2006, 06:37 AM

What is the difference between computer creativity and human creativity. Both start from some impulses coming from somewhere!
First, no such thing as computer creativity, where did you hear that? Computers only execute commands that were given to them. And what impulses, creativity is not about impulses. Here's the picture: - There is a need for music for a film sequence. The composer watch that sequence several times, trying to figure out what music would be appropriate for that situation, also trying to come up with some kind of music that some audience would emotionally respond to. The composer is trying several versions, making some options, taking some decisions, many of them being whimsical, or say: inefable. It is all about discernment and judgements about appropriateness. Now tell that to your intelligent/creative computer from my part. :)
How do you define "style".
No you can't define style, it slips through your fingers. But you can have a vague feeling about it instead. Tell also this to your intelligent/creative computer, from my part. :)
Most living composers don't have any "style", because taking different pieces no one could recognize the composer. Most modern art music composers define the rules for their own pieces.
Music is not made of notes, rhythms, rules, etc. Unfortunately that is a very extended missunderstanding. This subject has been discussed here before - music has no limits so it cannot be defined or quantified.
If a programmer creates SW that is deterministic, but parameter selection is based some seed(s), you get different rules for every piece without any reprogramming. That what creativity is all about.
No, creativity is not about applying rules, on the contrary: it is about breaking the rules to make something new, something that someone would appreciate. It's about taking decisions out of the blue. Tell also this to your intelligeny/creative computer, from my part. :)
But don't forget the role of the listener! Listening is also (or should be) creative.
That is a great point that opens up to the third dimension of music. (the first two being: the composer's, and the interpreter's) This deserves a special thread I guess - how is music re-created in each individual's mind? Eh?? How do each person experience music? First find this out, and then teach tour intelligent/creative computer to make music. :)

jimmymac

12-19-2006, 07:56 AM

That claim is logically absurd. Like claiming that one composer is going "to replace" another composer.

I agree, but it's been said.

Software

12-19-2006, 07:56 AM

First, no such thing as computer creativity, where did you hear that? Computers only execute commands that were given to them. And what impulses, creativity is not about impulses. Here's the picture: - There is a need for music for a film sequence. The composer watch that sequence several times, trying to figure out what music would be appropriate for that situation, also trying to come up with some kind of music that some audience would emotionally respond to. The composer is trying several versions, making some options, taking some decisions, many of them being whimsical, or say: inefable. It is all about discernment and judgements about appropriateness. Now tell that to your intelligent/creative computer from my part. :)

That's planning. There is no such thing as humabn creativity. Humans just behave acconding to the rules of their DNA and the cultural programming. Humans use some triggers coming outside. Everything just like computers.

No you can't define style, it slips through your fingers. But you can have a vague feeling about it instead. Tell also this to your intelligent/creative computer, from my part. :)

Was asking the general definition of "style". But computer music definitely has style, leaving out those fractals that have never worked, computer music can be coherent. Coherence is just one property of style.

Music is not made of notes, rhythms, rules, etc. Unfortunately that is a very extended missunderstanding. This subject has been discussed here before - music has no limits so it cannot be defined or quantified.

Right. There are the musicians also.

No, creativity is not about applying rules, on the contrary: it is about breaking the rules to make something new, something that someone would appreciate. It's about taking decisions out of the blue. Tell also this to your intelligeny/creative computer, from my part. :)

Wrong. Creativity is just about creating new rules. In music, in painting, in all arts.

That is a great point that opens up to the third dimension of music. (the first two being: the composer's, and the interpreter's) This deserves a special thread I guess - how is music re-created in each individual's mind? Eh?? How do each person experience music? First find this out, and then teach tour intelligent/creative computer to make music. :)

There is a big contradiction above. Human composers has to find that out, also. Many are not willing to do that. Many others a happy with cliches, John Williams kind or even worse. Even in art music.

jimmymac

12-19-2006, 08:18 AM

We'll all probably never reach an agreement on this, since this is really all down to philosophy. But it's fun, eh guys? :)

What is the difference between computer creativity and human creativity. Both start from some impulses coming from somewhere!

Sure it comes from somewhere. We have no idea where though.

Human creativity is also based on "changing seeds".

Since we don't actually know that to be fact, it would be unfair to state that as a truth :)
Again, it's being overly reductionist and reducing something we don't know about.

How do you define "style". Most living composers don't have any "style", because taking different pieces no one could recognize the composer. Most modern art music composers define the rules for their own pieces. If a programmer creates SW that is deterministic, but parameter selection is based some seed(s), you get different rules for every piece without any reprogramming. That what creativity is all about.

Thats simply not true imho. Now, in the area of film music I can almost always spot a track done by a specific composer. There will be a film on the tv that I have never seen and I will often say to my self (for example) "Thats good old Alan Silvestri, i bet"... and it often is. They have very recognisable styles. How they orchestrate, thier melodies. It's all very personal.

But don't forget the role of the listener! Listening is also (or should be) creative.

Well, there are just the passive listeners but yes I agree there.

To be honest though. I actually have no worries that musicians will be out of work due to computers. It's a silly notion entertained by those who really don't know much about music other than "it's just notes".
I recall reading an article back in '94 said that by 2000 or so there would be nothing but generative music radio stations.

Likewise, check out this article:
http://www.discover.com/issues/sep-06/departments/murderofmystery/

"Last week I had a jarring conversation with one of the most influential figures in Silicon Valley. Me: I wish more kids were learning to be musicians. He: In 10 years computers will be able to use a combination of artificial intelligence and massed data from the Internet to generate music better than human musicians. We can already use these techniques to choose hit songs more accurately than record executives. Musicianship will be an obsolete profession by the time today's kids grow up. There might be good reasons to teach kids music, but creating a new generation of professional musicians is not one of them."

What an ignorant thing to say. But the article is about refuting it.

jimmymac

12-19-2006, 08:32 AM

That's planning. There is no such thing as humabn creativity. Humans just behave acconding to the rules of their DNA and the cultural programming. Humans use some triggers coming outside. Everything just like computers.
Again, thats just a philosophy you have chosen to accept. (I used to myself to be honest, so I can understand it). The brain is not a digital computer. There are fundamental differences. Most we cannot even grasp yet.
Of course there are mechanisms going on, but they are NOT the same as my PC. And why is that a bad thing? Computers are not like us and THAT is what makes them useful.
The brain=computer metaphor has been slowly dying, it's just the diehards that seem to be keeping it alive.
It's interesting that most biologists don't agree with it. It's just the pie in the sky computer scientists who like saying things like that. No doubt it sounds exciting as a concept. (Perhaps they like to think "I'm making LIFE!!")
Maybe one day, they will be. But lets not lower our expectations of creativity just to accomodate a computer and make them seem smart.

Was asking the general definition of "style". But computer music definitely has style, leaving out those fractals that have never worked, computer music can be coherent. Coherence is just one property of style.

Nobody has suggested that it could not be coherent. Thats all down to how well the programmer codes the rules.
But as has been mentioned the stuff it produced is very generic. It has no personality because it does nothing but follow the rules. The human mind is far more mallable than any computer. That is it's strength.

Wrong. Creativity is just about creating new rules. In music, in painting, in all arts.

Yet the computer is incapable of making new rules. Thats the point. Thus, it is not creative.

There is a big contradiction above. Human composers has to find that out, also. Many are not willing to do that. Many others a happy with cliches, John Williams kind or even worse. Even in art music.
But thats just lazyness on thier part :) You cannot tar ALL human creativity because of that. That's sort of a strawman. :)

gstitt

12-19-2006, 09:15 AM

The brain is not a digital computer. There are fundamental differences. Most we cannot even grasp yet.
Of course there are mechanisms going on, but they are NOT the same as my PC. And why is that a bad thing? Computers are not like us and THAT is what makes them useful.
The brain=computer metaphor has been slowly dying, it's just the diehards that seem to be keeping it alive.
(Perhaps they like to think "I'm making LIFE!!")

There are in fact many obvious differences between the architecture of a human brain and a modern CPU. However, the real issue is not the architecture, but the process that executes on that architecture. AI seeks to determine that process, which could then theoretically execute on any architecture (biological, CMOS, or even mechanical). Whether or not this execution is "life" or not is a completely philosophical discussion.

AI is actually a poor name for the field, especially considering that the state of the art involves absolutely no intelligence at all. In fact, like nkysnd has said, the only intelligence involved now is the human intelligence involved with coming up with the heuristics and algorithms used in AI.

I think a more appropriate name for the "AI" we are discussing would be something like "human imitation". This field is completely independent of computer science, and involves research into the understanding of how humans work. If this can be completely understood (which I think is unlikely), then the human "process" can potentially be implemented on any computing device. I agree that the brain/computer comparison is a poor analogy. However, I think the underlying question is whether or not the process executed by the brain (and all its peripherals) can be captured so that a machine can execute it.

It's interesting that most biologists don't agree with it. It's just the pie in the sky computer scientists who like saying things like that. No doubt it sounds exciting as a concept.

This is a very aggressive claim (do you know most biologists)? :). While I'm sure many biologists disagree with the brain/computer analogy, I'm would doubt that most biologists would claim that replicating human behavior is impossible, especially since nobody can even come close to proving or disproving it.

Also, I do not know any computer scientistists who claim that computers can imitate human intelligence. They are interested in determining whether it is possible, but they certainly don't blindly accept any conclusion (they wouldn't be very good scientists if they did). So, I don't think it is really computer scientists that are the real problem, I think it is Hollywood's portrayal of AI which has really caused the problem. I took a class years ago from a leading researcher in AI/data mining and he started on the first day by defining AI as a bad Spielberg movie. :)

Software

12-19-2006, 09:19 AM

Since we don't actually know that to be fact, it would be unfair to state that as a truth :)
Again, it's being overly reductionist and reducing something we don't know about.

That was resonde to this claim: "Human creativity is also based on "changing seeds"."
Just read any text of how to boost creativity and its about finding proper "seeds".

Thats simply not true imho. Now, in the area of film music I can almost always spot a track done by a specific composer. There will be a film on the tv that I have never seen and I will often say to my self (for example)

Film music is just based on cliches and the greatest one have created their own to be followed by others. Film music has a kind of clichee vocabulary. My context is art music.

Software

12-19-2006, 09:23 AM

Yet the computer is incapable of making new rules. Thats the point. Thus, it is not creative.

Absolutely wrong. New parameters using proper triggers can creat quite new rules. Giving results one one could have guessed. That's computer creativity. Having no essential diffference with human creativity.

Software

12-19-2006, 09:28 AM

AI is actually a poor name for the field, especially considering that the state of the art involves absolutely no intelligence at all. In fact, like nkysnd has said, the only intelligence involved now is the human intelligence involved with coming up with the heuristics and algorithms used in AI.

Going again and again back to failed AI is wrong. Now we are talking about creativity.

I think a more appropriate name for the "AI" we are discussing would be something like "human imitation". This field is completely independent of computer science, and involves research into the understanding of how humans work. If this can be completely understood (which I think is unlikely), then the human "process" can potentially be implemented on any computing device. I agree that the brain/computer comparison is a poor analogy. However, I think the underlying question is whether or not the process executed by the brain (and all its peripherals) can be captured so that a machine can execute it.

The only limit is what the listeners can scope. If some "imitation" is needed, it is just because of the listeners. That is the problem, not the existing creativity of computers.

gstitt

12-19-2006, 09:39 AM

Going again and again back to failed AI is wrong. Now we are talking about creativity.

I think you may have misunderstood my intentions. I never said anything about the wrongness of AI and my discussion had nothing to do with creativity. Edit: In fact, my post actually seems to support the possibility of machines replicating human behavior. If you read earlier, I pointed out that the failure to prove something does not disprove it. Is this what you meant?

The only limit is what the listeners can scope. If some "imitation" is needed, it is just because of the listeners. That is the problem, not the existing creativity of computers.

My post was not specific to music, it was a general statement regarding the possibility of computers executing human behavior. Again, I never mentioned anything about creativity (or the lack of) being a problem. I was simply trying to say that most current AI research has nothing to do with what we are talking about. Sorry if my post was not clear. :)

Software

12-19-2006, 09:41 AM

I think you may have misunderstood my intentions. I never said anything about the wrongness of AI and my discussion had nothing to do with creativity.

My post was not specific to music, it was a general statement regarding the possibility of computers executing human behavior. Again, I never mentioned anything about creativity (or the lack of) being a problem. I was simply trying to say that most current AI research has nothing to do with what we are talking about. Sorry if my post was not clear. :)

OK. Sorry.

jimmymac

12-19-2006, 09:54 AM

There are in fact many obvious differences between the architecture of a human brain and a modern CPU. However, the real issue is not the architecture, but the process that executes on that architecture. AI seeks to determine that process, which could then theoretically execute on any architecture (biological, CMOS, or even mechanical). Whether or not this execution is "life" or not is a completely philosophical discussion.

But that is also making an assumption: That the brain is hardware and that there is 'software' running on it. We don't know that. For all we know, the process depends on the architecture. But of course, thats going to be hard to find out. I'm not ruling out anything at all. I'm just trying to state what we have today is not intelligent or creative. It's is the humans beings behind the programming that deserve credit. Not the automaton.

AI is actually a poor name for the field, especially considering that the state of the art involves absolutely no intelligence at all. In fact, like nkysnd has said, the only intelligence involved now is the human intelligence involved with coming up with the heuristics and algorithms used in AI.

Thats what I've been saying all along too. Software claims that the computer is creative. I was saying that the programmer is being creative. :)

I think a more appropriate name for the "AI" we are discussing would be something like "human imitation". This field is completely independent of computer science, and involves research into the understanding of how humans work. If this can be completely understood (which I think is unlikely), then the human "process" can potentially be implemented on any computing device. I agree that the brain/computer comparison is a poor analogy. However, I think the underlying question is whether or not the process executed by the brain (and all its peripherals) can be captured so that a machine can execute it.

Agreed. My point was that claiming that it IS possible to execute it is premature.

This is a very aggressive claim (do you know most biologists)? :). While I'm sure many biologists disagree with the brain/computer analogy,

I see your point and I apologise for that. It's probably not fair to make blanket statements without a backup.
Here is one interesting article on line here...
http://www.his.atr.jp/~ray/pubs/kurzweil/
I don't want to fill up this thread with too many links though. It could be construed as spamming. :)

I'm would doubt that most biologists would claim that replicating human behavior is impossible, especially since nobody can even come close to proving or disproving it.

Thats fair enough. But replicating is a tricky word. Because just because something acts like a human, doesn't MAKE it alive in the same way as a human. I think it would be perfectly possible to make a robot that walks and talks like a human in distant future. On the surface it might trick people for a while, but it's still an automaton. It's going to be hard, thats for sure. The turing test gets far too much mention as a valid scientific method.
If we use a comparable turing test on whether there is a god, then we must accept that there IS a god because people believe there is a god. :)

Also, I do not know any computer scientistists who claim that computers can imitate human intelligence. They are interested in determining whether it is possible, but they certainly don't blindly accept any conclusion (they wouldn't be very good scientists if they did)
In that case I'll repost a link to this...
http://www.theregister.co.uk/2006/10/27/bbc_horizon/
BBC's flagship science program was doing exactly what you describe. With people who claim to be scientists.

So, I don't think it is really computer scientists that are the real problem, I think it is Hollywood's portrayal of AI which has really caused the problem. I took a class years ago from a leading researcher in AI/data mining and he started on the first day by defining AI as a bad Spielberg movie. :)
Fair enough. :)
Again though, if you read my very first posts in this thread. I say that I am only against the HYPE and bullshi**ting that is very common in this young field. I mentioned several times that I respected the honest AI researchers. It's just they don't seem to get as much press time as the fantasy prone ones. It would not sell books I guess.

(blimey, I forgot how time consuming posting on a forum is!)

gstitt

12-19-2006, 10:32 AM

Excellent post, jimmymac.

But that is also making an assumption: That the brain is hardware and that there is 'software' running on it. We don't know that. For all we know, the process depends on the architecture. But of course, thats going to be hard to find out. I'm not ruling out anything at all.

Very good point. At least in computer science, any turing-complete machine can execute any application, which establishes a complete indepedence between processes and architectures. I agree that we really don't know what the brain is doing, so I really have no idea if computer science theory can be applied to it.

I'm just trying to state what we have today is not intelligent or creative. It's is the humans beings behind the programming that deserve credit. Not the automaton.

Agreed, %100.

I see your point and I apologise for that. It's probably not fair to make blanket statements without a backup.

No need to apologize, I was really just joking.

I say that I am only against the HYPE and bullshi**ting that is very common in this young field. I mentioned several times that I respected the honest AI researchers. It's just they don't seem to get as much press time as the fantasy prone ones. It would not sell books I guess.

I feel the same way. I checked out the BBC Horizon link and that truly is sad.

jimmymac

12-19-2006, 11:12 AM

Thanks for the responses gstitt, and you too software.
Just want to say that I do enjoy these discussions, so I value everyones input on both sides (not that there really should be sides as such) :)

Absolutely wrong. New parameters using proper triggers can creat quite new rules. Giving results one one could have guessed. That's computer creativity. Having no essential diffference with human creativity.

Ok, Lets just establish that we don't know what human creativity is, right?
You do have a THEORY. But we cannot claim that it is true, because we don't know.
Your theory appears (to my feeble mind at least :) ) to depend on reducing creativity to the level of our current computers, which seems a little unfair to me.

I'm willing to look at any theories and consider them but surely the burden of proof is on those who want to claim that the machines are creative? Forgive me, but it just appears to be overt anthropomorphising of the machine. It's wishful thinking.

And just to be clear, i'm not trying to over mystify the human brain (there is enough mystery without me needing to manufacture some) but I am against arbitrarily dismissing those mysteries without any real evidence.
Thats like Sherlock ignoring inconvenient evidence in a murder case. The biologist I posted earlier made some good points imho.

Now I shall repeat another point from my first post: Why don't we see effective joke generators? Stand up comedian programs? Because it is much harder to fake that particular element of human creativity on the superficial level (compared to music). Music offers enough of an established system that most people will just accept and so the factory line can be setup.

With regards to "new parameters using proper triggers", sure... the quake III bots do a good job of that (they use the fashionable "nueral nets" term these days). But they certainly are not creative. Game developers just make sure that as little as possible is left to chance. When something IS left to chance, we end up with problems. There are nodes everywhere. You should see how much is behind the scenes in even the most simple of game AI. It's a very clever illusion of an NPC adapting to tactics.
(Though most game AI is terrible... because it's so very hard and of course because many just cannot be arsed and gamers are more interested in fancy graphics. So we just make lots of normal mapped geometry ;) ).

It seems to me that in many ways we are all agreeing on the same thing, but we are arguing over semantics. I am not claiming that a machine can NEVER be intelligent or truly creative, forever is a very long time! :)
But it's not on the horizon and as the biologist states himself, there is a certain lack of imagination in assuming that if machine AI is ever possible that it's just going to be structured the same way as human intelligence.

**I'm now going to scurry off and eat some chocolate. It is almost christmas after all :cool:

Software

12-19-2006, 01:42 PM

Now I shall repeat another point from my first post: Why don't we see effective joke generators? Stand up comedian programs?

Because jokes have semantics, music doesn't.

With regards to "new parameters using proper triggers", sure... the quake III bots do a good job of that (they use the fashionable "nueral nets" term these days). But they certainly are not creative.

Because games have semantics, music doesn't.

But it's not on the horizon and as the biologist states himself, there is a certain lack of imagination in assuming that if machine AI is ever possible that it's just going to be structured the same way as human intelligence.

Human intelligence isa based on semantics, music doesn't...

jimmymac

12-19-2006, 02:08 PM

Well, it appears we really are all just arguing over semantics afterall. :)
When your PC can do that. I'll be impressed. Until then...
Your post doesn't really do anything but comfirm what I said here: "It is much harder to fake that particular element of human creativity on the superficial level (compared to music). Music offers enough of an established system that most people will just accept, so the factory process line can be set up."

I wonder if both of us are just trying too hard to come up with contrary points just for the sake of argument. ;)

nickysnd

12-19-2006, 02:20 PM

The composer is trying several versions, making some options, taking some decisions, many of them being whimsical, or say: inefable. It is all about discernment and judgements about appropriateness.
That's planning. There is no such thing as humabn creativity. Humans just behave acconding to the rules of their DNA and the cultural programming. Humans use some triggers coming outside. Everything just like computers.
The process I was describing is pure creativity, not planning. Planning is a scheme/method of proceeding (doing, acting, making, etc.) developed in advance. The creative process is about taking unexpected decisions on the spot, and going from there to unexpected horizons. Machines are just objects, the poor things cannot make judgments and take decisions.

Planning can be a mere starting point, occuring before the creative process. To create music, it takes judgments of how appropriate a decision might be, it takes to approximate the appropriateness, to make personal choices, to take personal decisions, to take chances, to risk. To use rules, if one really wants, but good composers do not obey them but bend them, break them, forget about them. One may start from some old rules, or may beforehead invent some "new rules", but sticking to rules is a creativity killer.

Musical creativity transcendes rules. Creating music can be described as translating an internal emotional, spiritual meaning (that we have no words for) into an external language, the composer hoping that the player will understand it and will transduce it accurately enough, also (both composer and player) hoping that the audience will resonate to that, and will translate back what they hear into internal emotions, spiritual feelings, states of mind. No need to say that machines are useless in all this extremely complex creative process.

Bach, Mozart, Beethoven, Debussy, etc. were creating music, and not "new rules", that's a ridiculuous concept to me: to create = to define new rules. Musicologists and other binary minds may think in rules, but composers, they think music - something impossible to grasp and understand. Your discourse is: yes-no, wrong-right, your posts are full of *wrong*s and *right*s. Music is far above "right"s and "wrong"s. Music starts its flight were anything else is helpless: concepts, words, rules, graphics, symbols, numbers, you name it. Music is a continuum, music is one immaterial infinite bird flighing through an universe far far away from your low narrow schematic mechanical binary world. Period.
No, creativity is not about applying rules, on the contrary: it is about breaking the rules to make something new, something that someone would appreciate. It's about taking decisions out of the blue.
Wrong. Creativity is just about creating new rules. In music, in painting, in all arts.
Again, "wrong". OK, if that's how you think, then go ahead and make music by creating new rules, that way you'll probably achieve the "right" music. You can also create the "right" paintings, the "right" comedy, the "right" drama, the "right" poetry - all just by creating "new rules". :rolleyes:

Long live rules-governed binary-minds! :)

nickysnd

12-19-2006, 02:30 PM

Because jokes have semantics, music doesn't.
Because games have semantics, music doesn't.
Human intelligence isa based on semantics, music doesn't...
Semantics is dealing with the relations between signs and what they denote. Music is coming from a superior form of human intelligence that transcendes signs. Therefore, music has long surpassed your precious semantics.

Software, I can feel a certain "tight-fisted" frustration traversing your posts. Music is compassioned love, not frustrated hate.

Software

12-19-2006, 08:34 PM

Musical creativity transcendes rules. Creating music can be described as translating an internal emotional, spiritual meaning (that we have no words for) into an external language, the composer hoping that the player will understand it and will transduce it accurately enough, also (both composer and player) hoping that the audience will resonate to that, and will translate back what they hear into internal emotions, spiritual feelings, states of mind.

Standard illusion, works only on the level of cliches. Changing the name of the piece can totally change the "spiritual meaning" of a piece. Performance is usually based on tradition (cliches). The listening experience is based on listeners context and cultural background.

Bach, Mozart, Beethoven, Debussy, etc. were creating music, and not "new rules", that's a ridiculuous concept to me: to create = to define new rules. Musicologists and other binary minds may think in rules, but composers, they think music - something impossible to grasp and understand.

The recognisable style of those composers is based on rules. The new things (creativiness) of those composers was just new rules.

nickysnd

12-20-2006, 05:39 AM

Standard illusion, works only on the level of cliches.
Of course it's an illusion, a splendid one, an everchanging one, the superlative one. Cliches? If in a Mozart symphony you do hear cliches (I have my experiences with a few people), then that is only to bad for you.

Changing the name of the piece can totally change the "spiritual meaning" of a piece.
Titles? If you are referring to pop songs, maybe, but I wouldn't generalize. The point is that most music has nothing to do with titles - think all chamber music, instrumental concerts, sonatas, symphonies, etc. Almost all of them don't have titles, unless you call Symphony No. 23 or String Quartet KV428 - titles. As about titles as Coriolan Overture, few people care about titles, the experience of music doesn't need titles. And being ironical about spiritual meaning doesn't honor you. FYI.

Performance is usually based on tradition (cliches).
Usually? What makes you so sure about it? And what do you call "cliches"? Tradition is cliche? When an actor reads a poetry and rise his/her voice to emphasize the third syllable of the word "innuendo" - would you call that cliche? When the second violin in a string quartet is playing a melodic major third - is that a cliche based on tradition? Do you realize how different it sounds, and do you have any idea of the infinity of effects a major third can have based on the context and on the way it is played?

The listening experience is based on listeners context and cultural background.
The listening experience might be based on a complex indecipherable combination of genetic predispositions, personal experience, individual imagination, momentary state of mind, and many many things that you won't believe. You'll probably call all this rules, or cliches, or chemistry, or seeds, or binary pairs, or whatever. Which I find kind of sad, because all these seem to make you believe you are holding the truth, while they are just keeping you away from it.

The recognisable style of those composers is based on rules.
Recognisable style? What do you know about style in general, or about any "particular style", to be able to recognize it? That is, unless you reduce "style" to a convenient set of rules that would seem to work in your schematic world.

The new things (creativiness) of those composers was just new rules.
Are you able to discern "old" and "new" in a Bach violin partita for example? Do you think that guy was writing notes and combining melodic lies, and building chords then braking them into arpeggios, stuff like that? Do you think that instead making music Bach was creating "new rules"? If this is your case, then you are completely missing everything that music is about.

And BTW, nothing personal, just judging upon your posts signed so appropriately - what you do looks like a pitiful attempt to reduce creativity and music and composers to your actual level of understanding, instead of trying the other way around. I only hope that at some moment in your life you will have a glimpse of Bach's or Beethoven's achievements - that will humble you and will elevate and expand your spirit. Till then you will just bang your head to the walls of your own semantic fixations.

Bob Thomas

12-20-2006, 03:39 PM

my god.. another visciously over zealous thread !

isnt this an incredibly simple question ?

i cant believe there are 8 pages of posts on it !! Some of the posts on here are truly reactionary and confrontational - i really cant see how they are part of a sensible debate

I agree ( almost ) entirely with Nicky. But I think its really simple :

1. Computers can create music - using the "rules" as understood by the people that programmed them ( these "rules" will be a hopelessly inadequate understanding of music - as all music theory fails to express what the music can and is based on the past not the untapped potential of the mind )

2. Computers as we know them will never create good music - as even luckily stumbling on something "pleasing" is not creating good music. Unless of course they can articulate their own emotion and spirit into it - ie. they would have to be "artificially" intelligent at least to the level of humans.

3. Beyond this we do not know and to be honest I think its a pretty pointless discussion seeing as its WAY off into the future if ever - but I think it comes down to whether you believe AI's could ever become capable of expressing emotions and develop a "spirit". I personally think thats more of a faith based question - and really doesnt have much to do with music.

Software

12-20-2006, 08:19 PM

Those people who do not believe in computer creativity have the criterium that computer should compose like humans. Isn't it enough to have human musicians and human listeners?

People from totally different cultures don't find western music emotional, spiritual or creative. Why not pretending that computers are just another culture? Trying to listen computer music like ethnomusic might help. Only that most computer music is rubbish. Especially so called fractal music is mostly rubbish.

But maybe discussion on this forum is hopeless. Most people here seem to be fixed with amerincan style film music cliches. Cliches that create emotions and spirituality in good and bad Hollywood films...

Some time ago I had the possibility to discussion with Michael Nyman (remember "Piano"). His problem is that film directors are always asking him to composer the old stuff again! Bay the way, his old book about minimalism is still quite interresting reading.

Bob Thomas

12-21-2006, 05:19 AM

Those people who do not believe in computer creativity have the criterium that computer should compose like humans. Isn't it enough to have human musicians and human listeners?

I'm afraid I think you've missed the point. I didnt say they need to compose like humans. I said for them to create good music they need to be as sophisticated as humans.

People from totally different cultures don't find western music emotional, spiritual or creative. Why not pretending that computers are just another culture? Trying to listen computer music like ethnomusic might help. Only that most computer music is rubbish. Especially so called fractal music is mostly rubbish.

Thats not necessarily true and actually very oversimplified. Elements of western musical systems have become part of almost all cultures now, as have other musical systems become part of western music. Computers ( as we know them now ) cant have their own unique culture (yet ) because :

1. they arent intelligent
2. everything they know is programmed by humans who have a limited knowledge ( probably of just one musical culture )

So unless computers become intelligent, emotional, have a spirit, and develop a culture of their own - they wont produce music as sophistacted as human music.

But maybe discussion on this forum is hopeless. Most people here seem to be fixed with amerincan style film music cliches. Cliches that create emotions and spirituality in good and bad Hollywood films...

Even though I find this forum intensely competitive in its members desires to prove their musical knowledge ( usually of trivia ).. I disagree with you. There are lots of discussions on the forum which revolve around concert hall music and many other types. Film music is a large part of it of course but there isnt anything wrong with that.

As for cliche's thats a bigger issue, more closely connected to taste and personal musical judgement if you ask me. I personally think the cliches you talk of are part of the musical lexicon of hollywood - which is a gradually evolving art form. There are composers who replicate this and perpetuate it, composers who challenge it in subtle ways, and composers who radically alter it. Its the same as composers working in any other field which has frameworks and references, and is completely legitimate artistic expression. I presume you wouldnt call Mozart a cliche ridden composer ?

Some time ago I had the possibility to discussion with Michael Nyman (remember "Piano"). His problem is that film directors are always asking him to composer the old stuff again! Bay the way, his old book about minimalism is still quite interresting reading.

They probably are asking him to mimic "old stuff", both his own old stuff like The Piano, and other people "old stuff". It happens to all working composers, and its not a bad thing - its part of the communication process between producers / directors and composers. The choice to mimic exactly or interpret through your own musical identity is they key. I think you should have more faith in your peers on the forum.

In fact I think people in general on this forum should CHILL OUT and BE NICE to each other! :)

Composing_automat

12-21-2006, 06:36 AM

I'm afraid I think you've missed the point. I didnt say they need to compose like humans. I said for them to create good music they need to be as sophisticated as humans.

That's vicious circle.

Counterpoint

01-01-2007, 11:40 PM

I don't doubt that enjoyable music might be created by computers someday.. but it most likely wouldn't be composed by a chess program. A chess program is not even an AI. It's just a simple algorithm.

Actually, I think that emotions are the key to creating a true AI. The logical rules are not enough and would only yield a machine capable of creating output generated by input. I think you need to synthesize a truly bored AI before you start to see synthetic creativity. ;)

It's also interesting that the turing test was brought up here... a test designed to determine whether a machine could pass itself off as a human being. There's only one problem... there have been real living people who have failed this test!! So... not a very good test, is it?

Actually, for those of you with a slightly cruel sense of humour, you can read this story (open the story and click the "next" buttons until chapter two if you want to skip the initial events and get to the fun part..)

http://www.nothingisreal.com/saga/#

So after reading this story... did Julia pass the turing test?? ;)

Cheers,

- Matt

Composing_automat

01-02-2007, 12:52 AM

Could someone show me a piece composed by "unemotional" computer but played by emotional humans (professional musicians).

nickysnd

01-02-2007, 01:06 AM

Computers are objects, they don't compose. Nevermind emotions...

Composing_automat

01-02-2007, 03:59 AM

Computers are objects, they don't compose. Nevermind emotions...
Yea, yea. Computers can't composer BECAUSE computers can't compose ...

nickysnd

01-02-2007, 09:39 AM

Yea, yea. Computers can't composer BECAUSE computers can't compose ...
No, they can't because they are objects.

Composing_automat

01-02-2007, 09:37 PM

OK. Could someone show me a piece made by "unemotional" computer (every single note written without a decision of a human) but played by emotional humans (professional musicians).

Kenton

01-03-2007, 04:33 PM

Quote:
Originally Posted by Composing_automat
Yea, yea. Computers can't composer BECAUSE computers can't compose ...

No, they can't because they are objects.

How true, I'm an object and I can't compose either.......

Oh, wait...... My accountant would disagree.!! :D

Seriously though, if I heard a piece of music that I enjoyed and that surprised me - and then I found out that it had been composed by a "machine", I would describe that object as "creative".
I think that blind testing is the acid test - if you repeatedly can't tell which pieces of music are composed by mechanical means then the machine is being creative.

Whether it's "Good" music or not is subjective and listener dependant, nothing to do with the creator.
Whether it conveys emotion is also subjective and listener dependant.... If I think it does, then it does...

Also, the argument that "computers" just do what they're told is rather simplistic and outdated - the more complex a system becomes, the more likely it is to exhibit "emergent behaviour" which the programmer couldn't have predicted.
Computers might be goverened by mathematics but the same mathematics brought us Chaos Theory and Godel's incompleteness theorem.

The way "forward" would be to move to an architecture which allowed the system to be self modifying and able to evolve, thus freeing it from the programmers limitations.
Of course, someone would have to "educate" the system ......
Neural networks of sufficient power might do the job eventually.

IMHO it's just a matter of time & complexity.

gstitt

01-03-2007, 07:34 PM

Also, the argument that "computers" just do what they're told is rather simplistic and outdated - the more complex a system becomes, the more likely it is to exhibit "emergent behaviour" which the programmer couldn't have predicted.

No offense, but this argument is in no way outdated (except based on claims in movies). The only emergent behavior that exists in existing applications is based on randomness, which is still doing what the programmer tells it to do. Unless of course the emergent behavior you are referring to are bugs (programmers can't predict these well), but even then the program is doing what it is told to do (the programmer just made a mistake).

The way "forward" would be to move to an architecture which allowed the system to be self modifying and able to evolve, thus freeing it from the programmers limitations.
Of course, someone would have to "educate" the system ......
Neural networks of sufficient power might do the job eventually.

IMHO it's just a matter of time & complexity.

It has been proven that architectures are completely independent of computational functionality. The architectures you refer to are actually my area of expertise, I have published many papers on evolving architectures. Evolving architectures may be able to execute a process more efficiently, but any process can execute on any turing-complete architecture. I really see the problem in AI as determining the process that can learn and show emergent behavior. Of course, I agree that new architectures will be needed to execute this process efficiently, but currently nobody has any idea what this process is, and actually nobody can prove that it is even possible to determine.

nickysnd

01-03-2007, 08:18 PM

No offense, but this argument is in no way outdated (except based on claims in movies). The only emergent behavior that exists in existing applications is based on randomness, which is still doing what the programmer tells it to do. Unless of course the emergent behavior you are referring to are bugs (programmers can't predict these well), but even then the program is doing what it is told to do (the programmer just made a mistake).
A finger snap of an expert worth 10 pages of discussions among amateurs.

I bow down to you, Sir. :)

Composing_automat

01-03-2007, 09:31 PM

No offense, but this argument is in no way outdated (except based on claims in movies). The only emergent behavior that exists in existing applications is based on randomness, which is still doing what the programmer tells it to do.

That's not true. One can e.g. use pictures as seeds for generating music. Using the picture of e,g, your face is quite different from random numbers, I guess...

Composing_automat

01-03-2007, 09:34 PM

Whether it's "Good" music or not is subjective and listener dependant, nothing to do with the creator. Whether it conveys emotion is also subjective and listener dependant.... If I think it does, then it does...

Yes. That's the point which many people are not willing to admit. But that also makes turing test difficult if not impossible.

gstitt

01-03-2007, 09:59 PM

That's not true. One can e.g. use pictures as seeds for generating music. Using the picture of e,g, your face is quite different from random numbers, I guess...

A seed is just something to initialize a random number generator, so it is still just randomness. You could use characteristics of a picture to change the way an application creates something (which sounds like what you are referring to), but this is not emergent behavior, so this does not disprove my statement.

Composing_automat

01-03-2007, 10:18 PM

A seed is just something to initialize a random number generator, so it is still just randomness. You could use characteristics of a picture to change the way an application creates something (which sounds like what you are referring to), but this is not emergent behavior, so this does not disprove my statement.

Perhaps the term "seed" is not the proper word, but using the structure, colors etc. of a picture for generating music has not much to do with randomness, because no random number generators is used. And the pieces are deterministic. Here you can see one of the many systems of that kind: http://www.synestesia.fi/music06.html

gstitt

01-03-2007, 10:22 PM

Actually, now that I think about it, it is hard to define what emergent behavior is. It is simple to write a program that appears to have emergent behavior. For example, you could have a function that randomly (or non randomly) generates a new function and then at some point during execution, the program could decide to execute this generated function based on some probability. Although this program would have behavior that changes over its lifetime, that behavior is still defined by the programmer (even if that just means randomness). I would define emergent behavior as new functionality created by a program itself, but since a programmer would have to implement this program, this would seem to suggest that this notion of emergent behavior is impossible (because the programmer specified the ability to create new functionality). This certainly isn't a proof, but just my logic for suggesting that the general idea of "emergent behavior" is just something else specified by the programmer, not anything creative on the part of the program itself.

gstitt

01-03-2007, 10:34 PM

And the pieces are deterministic.

This would further suggest that emergent behavior is not involved. Deterministic (at least in computer science) means that given the same input, the program will always generate the same output. If this is the case, then there certainly isn't any "creativity" since for every picture, the same music would always be created. You are correct that this may not be random but there also isn't any emergent behavior involved. I would be willing to bet that the program used on the link you provided performs some basic image processing, which just determines parameters used as input to a rule-based music generator. Furthermore, I'm guessing that this music generator is based on randomness, and since it is always seeded with the same parameters for a particular image, the random number generator always produces the same output (hence the deterministic nature). Whether or not this is composition depends on your definition of the word and could be argued until the end of time.

Composing_automat

01-03-2007, 10:39 PM

This would further suggest that emergent behavior is not involved. Deterministic (at least in computer science) means that given the same input, the program will always generate the same output. If this is the case, then there certainly isn't any "creativity" since for every picture, the same music would always be created. You are correct that this may not be random but there also isn't any emergent behavior involved.

I am not sure of that logic. In a way your demand is that emergence must be random in some way.

Let me cite John Holland: "Mundane activities such as farming depends on rules of thumb for emergence - for example knowing the conditions that influence the germination of seeds."

Growing pictures to music ... ?

gstitt

01-03-2007, 11:13 PM

I did not mean to imply that randomness must be involved with emergent behavior, I just meant that what most consider to be emergent behavior is just randomness. But, I can almost guarantee that this image to music conversion is completely based on well-understood logic that has parameters determined by the image to create different outputs. Given an image processing library and code to create midi files, I'm guessing that creating a music generator such as this was an relatively simple task. And, I would be willing to bet a fortune (if I had one) that if you email the authors and ask them if they use calls to rand() to generate the music, their answer would be yes.

Either way, this reminds me of a cool project I read about where a guy used image processing on a scanned image of a record to play back the audio on the record.

Composing_automat

01-03-2007, 11:28 PM

I did not mean to imply that randomness must be involved with emergent behavior, I just meant that what most consider to be emergent behavior is just randomness. But, I can almost guarantee that this image to music conversion is completely based on well-understood logic that has parameters determined by the image to create different outputs.
"Well-understood"? Of course there is the code. Seeds of a plant have also (genetic) code, well-understood.
But where do the emotions created by the music generator code come? (There are "laws" of code behind.)
How the seed grows and becomes a plant? (There are "laws" of physics behind.)
Given an image processing library and code to create midi files, I'm guessing that creating a music generator such as this was an relatively simple task.
May be not, listening this system created by big money: http://tones.wolfram.com/
And, I would be willing to bet a fortune (if I had one) that if you email the authors and ask them if they use calls to rand() to generate the music, their answer would be yes..
I have seen the code. No rand()s.

nickysnd

01-03-2007, 11:44 PM

May be not, listening this created by big money: http://tones.wolfram.com/
Some more lunatics that think that music is made of various combinations of notes...

(BTW, I've tried to listen to some "things" there - they are terrible, unbearable, un-listenable cacophonies...)

gstitt

01-04-2007, 10:30 AM

"Well-understood"? Of course there is the code. Seeds of a plant have also (genetic) code, well-understood.
But where do the emotions created by the music generator code come? (There are "laws" of code behind.)
How the seed grows and becomes a plant? (There are "laws" of physics behind.)

Ok, now I'm confused because you seem to be arguing the same point that I was. My point was that the "laws" of the code are specified by the programmer, therefore the program is still doing exactly what it is told. The program is not creating new "laws".

I have seen the code. No rand()s

I stand corrected but my point is still completely valid.

As an example, I could very easily create a synthesis engine that uses images an input. Just perform a DCT on the image, treat the spatial frequencies of the image as time-domain frequencies, and then synthesize those frequencies into a single wave that represents audio. This sounds very similar to your example, especially considering no randomness is involved. However, this is just a transformation of spatial information into an audio file. There was no intelligence involved by the program.

I acknowledge your point that human creativity may essentially be a well-defined behavior/program (as I have acknoweldged in my previous posts). However, I'm sure the creativity of your music generator is completely different that the creativity of humans, especially considering that researchers really have no idea how the human brain does many of the things that it does. If you consider this to be creativity, that's fine, it just means that we are really just arguing over definitions.

Composing_automat

01-04-2007, 12:07 PM

I am afraid you missed the point. What is the difference on emergencies between the seed/plant and picture/music?

gstitt

01-04-2007, 12:35 PM

I am afraid you missed the point. What is the difference on emergencies between the seed/plant and picture/music?

I don't think I missed the point at all. Mathematically, there is no difference, which seems like what you have been suggesting. I never suggested otherwise. Each one is essentially a function that given some input produces an output.

I think what everyone else is suggesting is that a better analogy would be seed/plant and picture/sound. Whether or not this sound is music, is all based on how you define music. Obviously, programs can create sound from an image. I could write a program to create sound from anything (temperature, scent, feel, images, electrical current in my TV, the rate of population growth of a city, etc.). Hell, I could write a program that transforms anything into anything else. I would definitely define these types of behaviors as transformations as opposed to creativity (at least until someone can clearly define creativity).

To make it clear to everyone what you are claiming, you should first give your definition of emergent behavior so we know exactly what you are talking about. All I am trying to say is that any new behavior in existing programs is specified by a programmer. Even if the program does create new behavior, it was told to do so by the programmer. This is why I would not consider this to be "emergent" in the commonly used way. Sure, you can call it emergent if you want, but then you just need to make sure you clearly define "emergent", which is what I was asking you to do. My guess is that if you do define it, you will find that we are both correct, because we are assuming different definitions.

DallasComposer

01-04-2007, 01:38 PM

I like playing Chess.

I like listening to music (old Dean Martin records) while playing chess.

I don't play very well though but do compose music in my head while listening to music (old Dean Martin records) while playing chess.

I like playing chess.

Composing_automat

01-04-2007, 10:12 PM

I think what everyone else is suggesting is that a better analogy would be seed/plant and picture/sound. Whether or not this sound is music, is all based on how you define music.
I could find question relevant if all cultures would have similar concept of music. But we all know that it is not so. All people, I guess, have difficulties to enjoy music of some other ethnic groups. What is worse, you should ONLY compare pieces which all have been played by HUMAN instrumentalist. Could you show me some examples of that?
Obviously, programs can create sound from an image. I could write a program to create sound from anything (temperature, scent, feel, images, electrical current in my TV, the rate of population growth of a city, etc.). Hell, I could write a program that transforms anything into anything else. I would definitely define these types of behaviors as transformations as opposed to creativity (at least until someone can clearly define creativity).
Accepted. All painting are not art, either. There must be quality in art. Well, pop art is perhaps different...
To make it clear to everyone what you are claiming, you should first give your definition of emergent behavior so we know exactly what you are talking about. All I am trying to say is that any new behavior in existing programs is specified by a programmer.

Wrong. Using e.g. that picture->"music" transformation the programmer can't tell the behavior of the music when looking at the picture. We are now not discussing about the "piano player" approach which has never worked or produced music, as far I know.
Even if the program does create new behavior, it was told to do so by the programmer.As above. The genetic code doesn't tell exactly what a plant look like and similarly the programmer don't know what the "music" will sound even if there is no randomness. It is a different matter do we get great art this way. In order to find it out, one should give the score to human instrumentalist to work out and play. Do you have any experience of that? Because I think that is the REAL TEST. (We all know that midi versions of classical pieces are ugly to listen)

gstitt

01-04-2007, 11:26 PM

What is worse, you should ONLY compare pieces which all have been played by HUMAN instrumentalist. Could you show me some examples of that?

What? How is this relevant to my comment? I was referring to your analogy and you are talking about comparing pieces. I'm lost.

Wrong. Using e.g. that picture->"music" transformation the programmer can't tell the behavior of the music when looking at the picture. We are now not discussing about the "piano player" approach which has never worked or produced music, as far I know.

What? Piano player approach? Again, please define what you are talking about. Also, since you claim my statement is wrong, please explain how behavior is defined if it is not defined by a programmer. The program certainly doesn't create new behavior, given that the code does not change. If this is not the case, please give a concrete example in detail.

The genetic code doesn't tell exactly what a plant look like and similarly the programmer don't know what the "music" will sound even if there is no randomness.

If the programmer understands the math involved and there is no randomness, the programmer will often have a very good idea of what the music will sound like, or at least know how it will be represented. Or, at least they should, otherwise how would they verify that the program is functioning correctly? To go back to my previous image processing example, I can look at many pictures and have a good idea of the spatial frequencies without executing a transform. I kind of see what you are saying here, but it seems like this would actually apply to every program.

Furthermore, what does this have to do with emergent behavior? Are you defining emergent behavior to be any situation where a programmer does not know the output of a program? If so, this is completely different than the definition I previously stated and would explain a lot of the confusion.

Do you have any experience of that? Because I think that is the REAL TEST.

The real test of what? To be completely honest, I have absolutely no idea what you are arguing anymore. What are you testing? You seem to just be making analogies, whose relevance is by no means clear.

Again, I ask you to please clearly define what you are arguing. Please read this post with a humorous tone, as I don't mean to offend you. However, it's impossible to have an effective discussion if you are not specific about what you are talking about. Maybe start by defining the emergent behavior that you talking about. Again, like I said before, there is a good chance that I will agree with you after hearing your definition.

Sam Fischmann

01-09-2007, 04:51 PM

Maybe the only "real test" would be a musical version of the turing test? I guess it would have to be different since it's just a one way conversation.

Counterpoint

01-09-2007, 10:42 PM

Maybe the only "real test" would be a musical version of the turing test? I guess it would have to be different since it's just a one way conversation.

How about a musical "duel" with a live musician and a computer on the stage. Each is given the chord changes and they take turns improvising a musical phrase with a live orchestra backing them up. Oh yeh, and hide them both behind a screen so the audience has to guess which is which. The one who received the loudest applause wins. :D

Oh, and to make it fair, they both have to play the Theremin.

- Matt

Composing_automat

01-10-2007, 09:43 AM

Maybe the only "real test" would be a musical version of the turing test? I guess it would have to be different since it's just a one way conversation.

The "real test" is not possible if music is defined as it nowadays mostly defined. Everything between, Cage, Beethoven, Boulez, Indian, African, Chinese ... you name it. If we limit the scope? Let's take the experiments made by professor Cope, emulation of certain composers. Then people can't differentiate genuine music from music composed by computer. Professionals succeed even worse than amateurs.

Composing_automat

01-10-2007, 09:46 AM

How about a musical "duel" with a live musician and a computer on the stage. Each is given the chord changes and they take turns improvising a musical phrase with a live orchestra backing them up. Oh yeh, and hide them both behind a screen so the audience has to guess which is which. The one who received the loudest applause wins. :D

I think Cope has practically done corresponding (analogical) experiments. See my message above.

nickysnd

01-10-2007, 09:58 AM

Oh, and to make it fair, they both have to play the Theremin.
Great point! The theremin is about the toughest instrument ever invented. It requires extremely fine hands movements, even for the basic task of sustaining a pitch at the same frequency and level of intensity. It's an incredibly tough job to play it in tune and with the dynamics and rhythms required.

Now, considering that a machine can't even catch a fly, or it can't catch a fly without harming it (which I can, and throw them back on the window outside to the mother nature, such a big heart I have... :) ) - how can a machine play that instrument? Or any instrument, for that matter.

This discussion was mostly about creating music, but bringing up the performance issue is quite relevant:

All composers have (had) a high understanding of the performance issues. Even if they dodn't play all the instruments, composers have the ability to imagine, to place themselves "under the skin" of the performers, and to play those instruments in their head - kind of virtual playing. Without this ability, just puting some notes on paper would only lead to dry experiments, at the best, and not to musical pieces (that's why avantgarde has hardly produced any music).

So, here comes my question: if a machine can't play and can't understand what performance consist of, how can a machine compose music?

Composing_automat

01-10-2007, 10:19 AM

Every presentation of the same piece by different conductors is different and even presentations of the same conductor vary. Similarly for orchestras.

Still waiting, waiting, waiting some pieces played by humans but generated (ok "composed") by computer. The real test!

nickysnd

01-10-2007, 11:20 AM

Every presentation of the same piece by different conductors is different and even presentations of the same conductor vary. Similarly for orchestras.
So with this "vary" thing do you believe that you have dismissed the performance issue and its relevance to music composition? Are you implying that a composer would think like: "It will 'vary' anyway, so it doesn't matter what I write" ... ?!? Is that the best argument you can bring up?

Then let me ask again: Can [some-thing-that-understands-nothing-about-music-performance] compose music?

Now let's examine a bit your digression, this "vary" thing -

The air I'm breathing "vary" each day, or even each breath I take. The water passing through a river "vary". The taste of each apple that I'm likely to eat will surely "vary". Everything "vary" - why would music performance make an exception? But there's subtlety to it, and a high understanding of music, which implies a well-felt balance between vary-ation and repet-ition, that makes this "vary" thing interesting for the people who might (or might not) want to listen to music.

You may chose to ignore what doesn't fit your definitions and semantics, but hiding your head in the "sand" (i.e. "semantics", in thins case) is hardly the way to sustain whatever position.

Composing_automat

01-10-2007, 12:20 PM

So with this "vary" thing do you believe that you have dismissed the performance issue and its relevance to music composition? Are you implying that a composer would think like: "It will 'vary' anyway, so it doesn't matter what I write" ... ?!? Is that the best argument you can bring up?

Of course not. But when comparing things as many parameters as possible should be similar.

Comparing music (composed by human composer) played by humans with "music" (generated by computer) played by just computers is wrong. Any examples overcoming this problem?

nickysnd

01-10-2007, 12:55 PM

"music" (generated by computer)
I understand and appreciate your quote marks on "music" - but how can an object have the initiative to output "something-sound-related" that would make musical sense to people - being them violinists or listeners? What does your object know about violinists' and listeners' values, in order to output anything that violinists would want to play, and listeners would want to hear?

Would you try to answer these two questions above?

And there is still another unanswered question: What would enable [some-thing-that-understands-nothing-neither-about-music-nor-about-music-performance] to compose music?

Counterpoint

01-10-2007, 01:07 PM

Now, considering that a machine can't even catch a fly, or it can't catch a fly without harming it (which I can, and throw them back on the window outside to the mother nature, such a big heart I have... :) ) - how can a machine play that instrument? Or any instrument, for that matter.

Here's a robot guitar player:

http://www.me.gatech.edu/mechatronics_lab/Projects/Fall00/group3/photo.htm

One of the left side links should have some MP3 clips... though it doesn't sound like that machine does much more than hit the notes (the examples don't really demonstrate any ability to play more than one dynamic).

- Matt

Composing_automat

01-10-2007, 01:19 PM

I understand and appreciate your quote marks on "music" - but how can an object have the initiative to output "something-sound-related" that would make musical sense to people - being them violinists or listeners?

The only way to know is there any sense is to listen. Have you had any courage to listen that kind of "music",

What does your object know about violinists' and listeners' values, in order to output anything that violinists would want to play, and listeners would want to hear?

The history of music know many pieces that musicians refused to play, for various reason. Many of those have been played later. Quite many of the new pieces composed by humans have been played only once.

Counterpoint

01-10-2007, 01:32 PM

Apologies for double-posting... but I found this too:

http://www.the-three-sirens.info

I'm still trying to figure out if it's some kind of hoax or if they're really doing what they claim. The details about how the robot "band members" actually work are pretty shady... I didn't find any information about how their "neural nets" produce musical output which makes me a bit skeptical.

- Matt

nickysnd

01-10-2007, 01:35 PM

The history of music know many ...
No, you have answered something else, not my questions. Please read them again, carefully, and please do answer them. To the point:

- How can an object have the initiative to output "something-sound-related" that would make musical sense to people - being them violinists or listeners?

- What does your object know about violinists' and listeners' values, in order to output anything that violinists would want to play, and listeners would want to hear?

- What would enable [some-thing-that-understands-nothing-neither-about-music-nor-about-music-performance] to compose music?

So - how, what, and what. Please do answer these three questions.

priblejr

01-10-2007, 07:55 PM

The robot-machines in my house are going to take over the world. Think about the movie "Toy Soldiers" The end is coming. They think for themselves. They feel emotions like we do.

And they write music.

nickysnd

01-10-2007, 08:07 PM

nickysnd

01-12-2007, 10:46 PM

Here's a robot guitar player:

http://www.me.gatech.edu/mechatronics_lab/Projects/Fall00/group3/photo.htm
Cool device!

Here's something somewhat older ;)
http://www.automates-anciens.com/english_version/pictures_history_music_box_etienne_blyelle/reuge_music_box_1.jpg

... and a bit of history (http://www.automates-anciens.com/english_version/main_pages/history_automaton.php).

And here (http://cgi.ebay.com/MENDELSSOHN-PLAYER-PIANO-MANUFACTURED-IN-1907_W0QQitemZ170070126796QQihZ007QQcategoryZ2383Q QtcZphotoQQrdZ1QQcmdZViewItem)'s the ol' piano roll, the grandfather of MIDI. It doesn't even need electricity, and it can do excellent jobs (wanna buy?:)) For example, I've heard a track featuring that type of mechanical device "playing" the Maple Leaf Rag (http://webapp1.dlib.indiana.edu/sheetmusic/devincent.do?c=01&p=1&id=LL-SDV-202046&s=screen) recorded on a paper roll by Scott Joplin himself, more than a century ago. It sounded marvelous, very lively, and quite exciting - I mean, it was the next best thing to SJ playing himself! On that paper roll were codified the results of his fingers' movements, dynamics (velocity) included.:cool:

messstudio

01-17-2007, 12:59 PM

this reminds me of a line from the show "Futurama" where the robots were revolting against the humans, Heremes (the Bureaucrat) hears that robots are nollinger serving people, and hes says...

"no robots!? who will cook our food, do our laundry and compose our smooth jazz" i giggled for about an hour thinking about that one...

Bob Thomas

02-07-2007, 07:54 AM

totally irrelevant or relevant quote depending on your reading :

"They took the credit for your second symphony.
Rewritten by machine and new technology,
and now I understand the problems you can see."

lol

Josta

02-07-2007, 09:28 AM

No, you have answered something else, not my questions. Please read them again, carefully, and please do answer them. To the point:

- How can an object have the initiative to output "something-sound-related" that would make musical sense to people - being them violinists or listeners?

- What does your object know about violinists' and listeners' values, in order to output anything that violinists would want to play, and listeners would want to hear?

- What would enable [some-thing-that-understands-nothing-neither-about-music-nor-about-music-performance] to compose music?

So - how, what, and what. Please do answer these three questions.

Nickysnd...OK, I'll take the bait...:rolleyes:

This has been a fascinating discussion. Superficially it's been about whether a computer can compose music, in any real sense, but more deeply this discussion is really about whether human-made machines can exhibit deeply human behaviors.

At the Artificial Intelligence lab at MIT, the director Rodney Brooks, a pioneer of humanoid robotics, once made the remarkable statement: "People ask me why I believe that machines will one day think and feel and be conscious. The answer is: because we HUMANS are machines, biological machines that use neurons for computation, and WE can think and feel and are conscious."

So I think the question really boils down to a profound philosophical one: are human beings and all their capabilities entirely "specified" by their bodies and brains, or is there something else to a person--call it soul, if you like--that can not be mechanized.

The reigning view at MIT, and many other places, is that we are beautiful, complicated, sometimes even chaotic machines. In the sense that our genes, molecules, cells, muscles and neurons all can be described mechanistically. We are mind-bogglingly complicated machines, to be sure, and made of wet parts rather than dry silicon. But machines nonetheless.

IF this is true (and let me just say that, for me, the jury is still somewhat out) then here are some possible answers to your questions, last to first:

- What would enable [some-thing-that-understands-nothing-neither-about-music-nor-about-music-performance] to compose music?

First, you build an object--I'll call it "Johann"--that humans can interact with naturally. Not something with a keyboard and mouse and LCD screen, but something with a voice, and the ability to understand speech. A face with two eyes and two ears so its visual and aural sensing of the world is similar to our own. Give Johann an animated face that can communicate emotions and that people will respond to emotionally. And give it the ability to move around its environment, perhaps on two legs, and extend out arms to touch and grasp things.

Then, you give this object a very powerful computer. Maybe it's a neural net, maybe it's another form of parallel processor. You provide the computer with some basic algorithms to be able to sense the world, communicate with speech, etc. Now, what you don't do is "write a program to compose and perform music." What you do is program in a learning algorithm so Johann can learn from accomplished human musicians.

Johann starts out knowing very little. It's like a baby. You then help Johann to learn music. And just as a baby starts out with a "blank slate" of memories, knowledge, etc., then becomes a child, adolescent and adult, Johann goes through the same stages of development, its internal algorithms implicitly learning from its human teachers the rules of counterpoint and harmony and voice-leading and orchestration. By exposure to hearing and then playing music, Johann, like any music student, gains a large repertoire of music styles. When Johann plays music, its human teachers give positive or negative feedback, and so Johann's internal understanding of what is "correct" or "wrong" in music becomes shaped along the lines of its human teachers' understanding of those same concepts.

At this point, it could be reasonably argued that Johann understands something about composition and performance.

- What does your object know about violinists' and listeners' values, in order to output anything that violinists would want to play, and listeners would want to hear?

Again, as long as the object (Johann, say) can learn by example and by human feedback, it will come to have an internal representation of the values of its human teachers and companions. If it is programmed to respond to positive feedback, it will tend to output more of what its human listeners tell it is "good" and less of what they tell it is "awful." Over time, it will learn to output better music (as judged by its human listeners).

How do other nonhumans--like dogs--come to know about the individual personalities and values of their human companions? By the same process of interaction and learning. Dogs come to understand that their owners value quiet, or exercise, or when they like to get up in the morning. These are perhaps simple values, simpler at least than the sublime values that might be expressed in music. But dogs are prewired to be dogs. Johann would be prewired to learn human characteristics. If brains, made of neurons firing electrical impulses, can encode values, why not computer circuits?

- How can an object have the initiative to output "something-sound-related" that would make musical sense to people - being them violinists or listeners?

Just like a child, Johann would be programmed to respond to rewards (and perhaps punishments). As its human listeners rewarded it or scolded it, it would gradually learn "what makes musical sense" to people and would initiate more of that type of "output" (i.e., music).

None of the above is just speculation: there are people--including musicians--at MIT, CMU, Stanford, Honda, numerous sites in Europe, all working to make this a reality. Google Cynthea Breazeal's work on "sociable machines" or Rosalind Picard's work on "affective (emotional) computing" (both at the Media Lab at MIT.)

SO...this probably doesn't ANSWER your questions, Nickysnd, 'cause neither I nor anyone else that I know of can build Johann yet...but I think it's a plausible roadmap to some possible answers. Will a machine ever become a good music student, and then become proficient at musical composition and performace? I dunno. History proves it's a risky business to say "people will never be able to do/invent that" (easy-to-open CD wrappers notwithstanding). With all the work going on in robotics and AI, we may find out in time for the 2037 Grammy's...

Cheers!

Josta

nickysnd

02-07-2007, 03:21 PM

OK, I'll take the bait...:rolleyes:
I'm happy you did, I have really enjoyed your post. Thanks for taking this discussion seriously.

Now to the point:
At the Artificial Intelligence lab at MIT, the director Rodney Brooks, a pioneer of humanoid robotics, once made the remarkable statement: "People ask me why I believe that machines will one day think and feel and be conscious. The answer is: because we HUMANS are machines, biological machines that use neurons for computation, and WE can think and feel and are conscious."
That's a strange type of logic to me, as it relates unrelated things (or related only in theory) - "We are conscious because we feel and think, we feel and think because we make computations, we make computations because the neurons send electrical impulses". To me this logic is flawed, and to say plainly what I think - That's a typical mental pattern of a narrow binary mind.

We are mind-bogglingly complicated machines, to be sure, and made of wet parts rather than dry silicon. But machines nonetheless. IF this is true (and let me just say that, for me, the jury is still somewhat out)
No, the jury may take their time and stay out. - Humans are not machines, for innumerable reasons. Machines are fundamentally predictable, humans are fundamentally unpredictable. We are living creatures. Machines are dead. Dogs will write poetry and discuss moral implications of abstract philosophy long before the first machine will output one meaningful musical phrase. I rather believe in future animal evolution to conscience rather than in dead objects capabilities of having/expressing emotions and making music.

If brains, made of neurons firing electrical impulses, can encode values, why not computer circuits?
You seem to relate "firing impulses" to "encoding values". This is a logical trick, again. Those things are not related. It's not sure what the neuronic electrical impulses are doing - you can only presume and assume. So it's just a theory. Not proved.

Johann starts out knowing very little. It's like a baby.
This anthropomorphization thing really doesn't work with me. No, your "Johann" (Bach, forgive me!) will never know anything, because it's an object. You can't make an object know anything. If I put in a bowl the words from the 1st act of Hamlet, that bowl will NOT know the 1st act of Hamlet. Sure, I can build and attach a mechanical/electrical device to the bowl so that it will spin and spit out the words in the same order as Shakespeare arranged them - but that would NOT mean that the bowl has learned the 1st act of Hamlet. An object simply cannot understand/know anything.

The brain as a mechanical device is a model that has long been abandoned. The brain as an electrical device is a 20th century theory that has been proved as true as the other one. Nowadays I see attempts for a theory asserting the binary brain, with impulses and such, upon the model that computers function. All these are mere theories. What everyone plainly sees, what we know for sure is that the human brain's functionality is the most inexplicable phenomenon in the observed universe. We can fantasize forever about brain's functionality, but I believe we should better use our fantasy for more useful things, like health research, feeding the poor, and music (which is food for our poor souls).

Those chess programmers aiming for music composing programs, they named their puppet "Ludwig". You have named yours - "Johann." As I see it, the goal in such naming is to achieve sympathy for your dolls and credibility for their potential. That's not that easy. People have learned to look at things with a fair dose of skepticism.

Science is marvelous human activity and it can deal with a lot of things. However, it cannot deal with everything - and this very statement is one thing that science has hard times dealing with. Moreover, what science do? - it makes up hypothesis, tests them in various ways, then builds up theories based on the results. Science only indicates possibilities (which, as we know, are improbable until they actually happen consistenly). Now technology comes into play and uses those theories and applies them in the real world to achieve functional devices. Technology may also rise before, in the mid-stage of testing hypothesis - but its aim is practical, not theoretical.

The fundamental principle of science is: obtaining identical result, from an identical cause, in identical conditions. Considering that, do I need to further explain why science cannot deal with emotions and music? What are emotions? What is music?

For the same reason, science can't deal even with knowledge, until it gives a clear answer to questions like:
"What is the difference between knowing and believing?"
followed by -
"Can they be separated?"
followed by -
"If yes, what is that we do know?"
then followed by -
"What is our mechanism of knowing what we know, and how does it function?" (well, this last one involves technological research, which nonetheless should follow to unambiguous answers to the previous questions)

I'm not against theories and discussions, but I can't accept fantasies presented as possibilities. Often those research papers look like sectarian obscurantist texts - asserting quite clear the "possible" outcome, while the backing argumentation is presented in a convoluted jargon. I don't like people who are trying to fool other people. Naivety is a virtue, like skepticism, yet it is more vulnerable - and taking advantage on its vulnerability I find unforgivable.

Any research group studying emotions or music "scientifically" should first answer clearly to basic questions like the ones posted above. If they refuse to give such answers, then they should accept their status of underground para-pseudo-scientists (read crooks) that skip unconvenient steps.

As I said before, while composers, since centuries, fly across many universes, scientists are still struggling with contradictory theories. Which might be fun as well, I dunno... As for technologies, they are only concerned with inventing things that makes life easier, and I doubt that "machine generated notes" will ever fall in that category. In any case, I won't buy such a machine. Yet some people might - there are all sorts of underground pleasures, this one looks harmless...

Of course, all the above is just my fantasy, just like the mentioned scientific theories are. With one difference: I don't know, I only believe in what I say, and what I say is true only to me, and only now, at this very moment. "I believe" is what scientists should start all their statements with, and repeat it from time to time, so that people can understand clearly that there is no real, objective truth to those theories.

shnurgle

02-07-2007, 04:24 PM

I want to agree with Nicky on principle. But his point of view is based on what we currently know about machines, not about what may come in the future, which nobody knows. I don't think you can definitively say that machines won't be capable of writing music, or thinking for themselves, or even feeling emotion, because we don't know where the technology will go. I'm no expert on machines, but I'm pretty sure the experts believe this may one day be possible. If some sophisticated AI technology is ever delivered that enables machines to be independent, free-thinking, conscious entities, all this could very well happen. The point is that it is possible that some technology will be developed that defies everything we think we currently know. Anything is possible.

Nevertheless, I'm all for the human element, so I hope this Matrix-like day never comes. :)

Josta

02-07-2007, 09:29 PM

I want to agree with Nicky on principle. But his point of view is based on what we currently know about machines, not about what may come in the future...

Nicky clearly feels very strongly about this topic. He certainly raises some interesting ideas and I appreciate his passion. However it seems that his point of view is not based on what we currently know about machines. For example, it is trivially easy to make a machine that is unpredictable (ever see those little desk toys that suspend a magnetic ball over another magnet in a plate and it moves around incessantly and erratically? A simple example of chaotic, unpredictable motion in a mechanism. Same for any machine that exhibits turbulence. I wish I could have predicted the last time my computer crashed. And if Nicky knows how to predict the behavior of a slot machine in a Vegas casino, I'd love to learn!) Indeed, what we currently know about machines and biology is that they are in many ways headed on a path of convergence, not the dualist "living" vs "dead" view that harkens back to the vitalism debates of the 18th and 19th centuries.

Many people feel threatened by challenges to the notion of our "specialness" as humans. Witness the tribunal of Galileo [note: the following dialog should be imagined in the voices of characters from South Park]:

[Inquisitor] Galileo, humans are at the center of the universe. Nothing can ever change that. To say otherwise is the mark of a narrow mind: not great, learned minds like ours. Celestial bodies can't orbit around each other without the Earth at the center.

[Galileo] Well, but see, I've seen the moons of Jupiter orbiting around the planet.

[Inquisitor] You saw no such thing! Besides, this is only the theory of that Polish guy Copernicus, it's not proven.

[Galileo] Um, right...well, but see, if you say it can't be this way, then you won't really be open to looking at evidence that could prove Copernicus's theory. Right?

[Inquisitor] Tricks! Logical tricks!

[Galileo] I'm sorry, I meant no trickery at all. It's just that....well, look, here's my telescope: I just cleaned the lenses...just come look for yourselves.

[Inquisitor] No need to look. Before we could look you would have to first explain how the fundamental principles of your cosmology could be reconciled with the Ptolemaic and Aristotelian world views. Otherwise you are a pseudoscientist.

[Galileo] Really? It's me that is the pseudoscientist? Well, but see, I just want you to look through my telescope: you'll see the little moons around and against the beautiful bright sphere of Jupiter. And if we look again tomorrow and for the next few days, you'll see that they are rotating around Jupiter itself, not the sun.

[Inquisitor] Fantasy! And what about the human element? Where is Man in your dangerous moons-rotating-around-the-planets worldview?

[Galileo] Well, we're still right here! We're the ones building the telescopes! We're the ones proposing theories using our God-given intellects and then testing them by observing the actual world (rather than just clinging to more theories, supported only by asserting "this is the way it is"). We're still pretty central, and certainly not out of the picture just because our former worldview needs to be changed. After all, it's our minds coming up with the worldview, for goshsakes.

[Inquisitor] Is it hot in here? Can someone open a window please?

Sometimes the debate between the someday-machines-will-think-and-feel people and the machines-will-never-be-able-to-think-that's-just-crazy people sounds a lot like the debate between humans-evolved-from-other-species and evolution's-just-an-unproven-theory-there's-no-way-I-am-related-to-a-monkey. Unlike that debate, the cool thing about this conversation over whether a computer or android or robotic organ-grinder monkey will ever be able to really compose music is that, if it is possible, one day we will actually see and hear it and know for sure: that darn contraption will be there, crankin' out a tune. And we'll either be tappin' our toes or fleeing for the hills in terror.

Joe

shnurgle

02-08-2007, 11:43 AM

As always, very well said J-Rochse.

Josta

02-08-2007, 01:42 PM

As always, very well said J-Rochse.

Thanks, Shnurg. By the way, just for the record I think it is David Bowie's actual hand rolling those glass orbs around in Labyrinth. But what really keeps me awake at night is: was it really David Duchovny's hand inside the cryogenic chamber in Zoolander, or did they have an actual hand model stunt-double as a hand model? It's like a stone ground corn riddle wrapped in a spinach mystery inside a sun-dried tomato enigma. With a side of guacamole.

Joe

Stefan Podell

02-08-2007, 02:16 PM

That pretty much sums it up for me, J-Rock. I'm pretty sure the tomatoes are the ones in a jar soaked in olive oil, though, right?

Josta

02-08-2007, 02:20 PM

That pretty much sums it up for me, J-Rock. I'm pretty sure the tomatoes are the ones in a jar soaked in olive oil, though, right?

Dude, it's like you read my mind. Or were sneaking around in my pantry again. Stop that.

nickysnd

02-08-2007, 02:37 PM

it is trivially easy to make a machine that is unpredictable
By "machines always predictable" I meant that we know exactly what a machine can output, even if we don't always know the exact moments of the output-ing. I don't say that it's impossible to build, but it's absolutely NOT desirable to have a machine that is unpredictable - a machine that we will not know what it can output, which can be anything: bubble-gum, nuclear bombs, film cues, new viruses, etc. Therefore, machines will always be predictable - not because the opposite is impossible, but because the opposite would be mass suicide.

Many people feel threatened by challenges to the notion of our "specialness" as humans.
I don't feel threatened by note-outputting machines (they are around since many years). However, I would feel threatened TO DEATH by an unpredictable machine.

Witness the tribunal of Galileo
Galileo was arguing about clear/real facts that his inquisitors refused to look at. He also was asking to look at the existing universe in a completely new way. The case here is totally different. There are no clear/real facts to look or to refuse to look at. This is not about: "Whether the chicken crossed the street, or the street moved under the chicken - it depends on your frame of reference." Nor are we discussing whether zebras are black with white stripes or white with balck stripes. I don't care if the universe is flat and curved, as some scientists argue. It's not the view of the world at stake here, no paradigm shift is happening. Here we are discussing the following possibility: Whether or not a machine can output a musical phrase without being programmed to produce it.

At this very moment it's obvious that it can't - a machine can do only what it is told to (even if it is told to output randomly). Now, if we agree that machines should NOT output whatever they want (bombs, viruses, bubble-gum, etc.), then we agree that machines will NEVER have freedom of choice, so they will always do what they are told to do, like the industrial robots today. Nothing will come from themselves, no chosen and conscious original touch whatsoever - they will always output only what they will be told to output.

You may laugh, but we are creative only because we can be destructive. We are effective because we can be whimsical. We are good only because we can be bad. And vice-versa, of course. We are capable of art, of thinking, of singing, of painting, we are capable of all the things that other creatures aren't - only because of that crazy little thing called free will. Now, would you grant the qualities of being whimsical, destructive, and bad to a machine? Would you grant free will to a machine?

Hence my repeated answer to the initial question: a machine will always output only what it is programmed to output. Or else, we will not be here to hear that output. Or maybe we'll be, for a while...

Josta

02-08-2007, 04:43 PM

Heya Nicky!

First, can I just say that I really enjoyed reading and thinking about what you just posted. We agree on many things, disagree on others, and please take my challenges and rejoinders in the spirit in which they are intended: two guys exchanging ideas and enjoying the mental exercise.

OK, the wind up, the pitch...

I am in complete agreement with you that a defining characteristic of people versus machines is free will. In fact, for me, free will is the basis of my own personal belief that there may be something like a soul, because nothing currenlty known in physics can account for free will: physical things are either deterministic (Newton's laws) or random (quantum indeterminacy) and our experience of the ability to will is different from either. That is, when I make a choice it does not feel either inevitably determined or totally random.

That being said, my belief in free will is based on my very subjective experience of what it is to "will". I cannot really make a good argument to anyone else about it: if, for example, Shnurgle told me that his experience of "free will" was exactly a kind of quantum randomness (which, you know, may not be too far off for Shnurg) then I could make no argument against him--this is all first-person subjective stuff.

Would you grant free will to a machine?

OK, here's the brain-busting part. Given everything I've just said, if I'm scathingly intellectually honest with myself, I have to concede that it is possible that free will is an illusion. Everything I do and choose may be a function of my genes, the programming of my neurons and my interactions with my environment. I feel like I am choosing freely, but that's just because we humans may have evolved an illusory sense of free will that benefits us by allowing us to feel empowered, good about ourselves, able to hold people accountable for their crimes, etc.

So, the hard question perhaps becomes not "would I grant free will to a machine?", but "would I grant it to myself?" <<squishy sound of my brain giving a little shudder>>

It would be completely possible to program a machine to report that it feels like it has free will, perhaps even to give it the illusion that it has it. In that case, we know that it doesn't have free will, but it acts like it does. Maybe that's the same situation we're in. (I'm playing Devil's advocate here!)

By "machines always predictable" I meant that we know exactly what a machine can output...Here we are discussing the following possibility: Whether or not a machine can output a musical phrase without being programmed to produce it. At this very moment it's obvious that it can't - a machine can do only what it is told to...Hence my repeated answer to the initial question: a machine will always output only what it is programmed to output.

Actually, at this very moment, it's not obvious that a machine can't output something it hasn't been programmed to do. Because there are some very creative ways of "telling a machine what to do." An example is genetic algorithms: here, you don't directly tell the computer what to output, rather you give it a starting point and some desired properties of the output, then let it evolve solutions by a process of "natural selection". The computer tries out different things, and if the thing gets closer to the target qualities it keeps it, modifies it, and keeps generating new versions. If the thing gets farther from the target qualities, the computer discards it and tries other things.

I saw a vivid demonstration of this a couple of years ago at a conference by a guy who was looking for ways to generate movement for animated CGI characters for movies. He had developed a genetic algorithm for legged walking (where the "creatures" could have any number of legs you liked). What was amazing was how surprising some of the gaits were that the program came up with: three-legged "creatures" doing sort of floppy cartwheels, two legged creatures limping along, other things that sort of vaguely looked like the "penguin walk" from Monty Python or little hop-skip-and-a-jump things. The audience laughed: the computer's outputs had generated a real emotional response in us. None of these outputs had been explicitly programmed into the computer: it generated them based on the laws of physics and a genetic algorithm that essentially instructed the computer to "get creative". If I had seen the same animations and was told that some guys at Pixar had done them, I would have nodded my head and smiled and said "Wow, yeah, those Pixar guys are so creative."

So, yes, in some sense the computer was "just doing what it was told." But what if we tell the computer to "get creative and make good music"? Is that really so different from telling a human musician under contract for a film score to "get creative and make good music "?

Galileo was arguing about clear/real facts that his inquisitors refused to look at. He also was asking to look at the existing universe in a completely new way. The case here is totally different. There are no clear/real facts to look or to refuse to look at.

I have to disagree with you here: one person's "clear/real facts" are another person's "fantasies presented as possibilities". In my original reply to your three questions, I laid out an approach to answer how one would build a machine that could truly compose and perform music that would make sense to people. Every element of the "recipe" that I described is being created, right now, in the labs and research centers I named. (In some cases, it's work that dates back at least 10 or more years.)

You dismiss the idea of a computer that could exhibit or even have emotions, referring derisively to people working in this area as "underground para-pseudo-scientists" or even "crooks". But I specifically cited two brilliant researchers, Cynthia Breazeal and Rosalind Picard, who are doing groundbreaking work in social and emotional computing--and I can personally attest that neither one is pseudo-anything nor a crook. And they are achieving real results: at the SIGGRAPH computer graphics conference in Boston last July, Rosalind Picard and her grad students presented a computer system that can look at people's faces and in real-time analyze the facial expressions and head gestures and determine whether the person is experiencing one of six different emotional states: agreeing, disagreeing, interested, confused, concentrating or thinking. And here's a "clear/real fact": in real-world testing, the computer system has been proven to be able to determine the emotional state of a person better than 94 percent of human observers.* Given the subtlety of difference in expression for most people between "disagreeing" and "confused", or between "thinking" and "concentrating", I'd say this is a pretty interesting result.

So Nicky I invite you to have a look through these telescopes. You may be surprised by what you see. Or if not, at least amused ;):

A Very Good Book (http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=4062)

Affective Computing at MIT (http://affect.media.mit.edu/)

The "Robotic Life" Group at the MIT Media Lab (http://robotic.media.mit.edu/)

Kismet: First steps toward an emotional robot (http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/kismet.html)

As always, cheers!

Joe

* "Self-Cam: Feedback from what would be your social partner," Alea Teeters, Rana el Kaliouby and Rosalind Picard, Proceedings of the ACM SIGGRAPH '06, 2006.

nickysnd

02-08-2007, 06:47 PM

I invite you to have a look through these telescopes. You may be surprised by what you see. Or if not, at least amused ;):
Thanks, I really appreciate your comments and links!

I'll surely check those web sites, and see whether they are telescopes or kaleidoscopes... :)
http://www.emersonplace.com/emporium/images/kaleido9.gif

gstitt

02-21-2007, 10:40 AM

Here is an example of an article that I would call crap:

http://www.msnbc.msn.com/id/17244922/

It starts out reasonable, but ends with a prediction by Kurzweil that computers will match the intelligence of humans by 2029. Even as a supporter of machine learning and AI research, I think this claim is absurd. Again, until we figure out how humans work, I don't think it is possible to make claims about when computers will match human behavior. Just look at how great image recognition currently is (said in a sarcastic tone).

Josta

02-21-2007, 02:43 PM

Ah, Kurzweil...you know that fine line between genius and madness? :rolleyes:

I actually think that Ray is on the genius side of the line (mostly), but giving out dates like "2029" to the popular press is kinda silly, IMHO.

On the other hand, in the late 80's this dude correctly predicted to the year, if I remember right, when a computer would become the world chess champion (1997). (I say this recognizing full well that winning chess is trivial as a challenge to a machine compared to writing music.)

And, speaking of image recognition, he developed the first successful font-independent text-to-speech reader for the blind.

The point of much current research in AI is that we don't have to "figure out how humans work", at least not very completely, to create a machine with very human behaviors. The current strategy--and it has produced some surprising results--is to create machines with a basic ability to sense, learn, remember and act, place them among humans, and let them "play" together. Nicely, hopefully.

One of the interesting parts of this process is that, as researchers attempt to model machines on humans, we actually are learning a lot more about how humans work as well.

Dogs know a lot less detail about "how humans work" than we humans do. And yet they can relate to us, both cognitively and emotionally, pretty well, just by hanging around us and learning from their interactions. That's how AI may continue to develop: first at sort of a pet-like level of interaction and dependency, then maybe something more like real human behaviors, including aspects like creativity.

Maybe not by 2029. But definitely by 3:34 pm on March 15, 2036.

;)

Joe

nickysnd

02-21-2007, 03:25 PM

The point of much current research in AI is that we don't have to "figure out how humans work", at least not very completely, to create a machine with very human behaviors.
Joe, I greatly admire your consistent positive approach - my mother is a physicist, so I am familiar with this type of scientifical view of the world. Yet, it seems to me that your optimistic discourse is based on the presumptions that science can deal with everything, and that technology needs only time to accomplish any task. That's simply not true: for example, science cannot deal with faith, emotion, thought, and any other mind-related issues. Science just have ever-changing theories about all those issues - and that's because, in its stubbornness, science refuse to accept that it simply cannot deal with such things. They are un-explainable. Period. Also, science cannot deal with art, music included. Science can do wonderful things, it especially keeps our mind sane and in good shape, but science cannot deal with and explain everything, and that's a given. IMO, outside theory, speculation, and SF novels, AI is a sad joke.

As about the presumption that technology only needs time to accomplish anything that mind can imagine, can you tell me why technology cannot make an object fly like a bird? Not like a helicopter, like a jet-plane, or like a rocket, but like a stupid little birdy - from a tree to another, from a branch to another, and between them, etc.

The current strategy--and it has produced some surprising results--is to create machines with a basic ability to sense, learn, remember and act, place them among humans, and let them "play" together. Nicely, hopefully.
"Nicely, hopefully", huh? Joe, I would leave my kids playing with a dog, but I would never leave them alone with a machine that is capable of free will (assuming this absurd possibility). Would you leave your kids alone with such a creature? Just a simple question, for the sake of the discussion.

Josta

02-21-2007, 05:25 PM

Well, you offer some very confident statements there, Nicky! One problem, though, is that when I have offered very tangible examples in response to your previous questions, you keep on just asserting "science can't do that". If these are simply unalterable beliefs on your part, that's fine. But if that's the case, then there's not really much point in debating, is there? Let's just have a beer and instead speculate about the features that PLAY will have!

For example:

That's simply not true: for example, science cannot deal with faith, emotion, thought, and any other mind-related issues.

And

They are un-explainable. Period.

In a previous post, I offered you a very specific counterexample against your assertion that "science cannot deal with emotion", giving you a very specific reference to a machine that is not hypothetical or speculation, but has been built and tested and demonstrated the capacity to "read" people's emotions from their facial expressions better than 90% of human observers that were also tested. That's certainly one example of "dealing with emotions." A far cry from an "emotional machine", to be sure, but it does show that the scientific investigation of emotion can produce tangible results, not just speculation or "sad jokes".

And, honestly my friend, there are just SO many examples of people saying throughout history that "X is unexplainable" or "X can't be done. Period" and, a few years or decades or centuries later, X is nicely explained or done. (Solar eclipses, how light moves from here to there, sailing around the world, the physical basis of heredity (DNA), how to make a flying machine, space travel, heart transplantation, and the chemical composition of SPAM are just a few examples.)

Yet, it seems to me that your optimistic discourse is based on the presumptions that science can deal with everything

Again, not true. In a previous post (you are reading these, right ? ;) ) I told you that, to my mind, nothing in physics can account for free will as I experience it, and this is the basis of my personal belief in the possibility of something like a soul.

Science just have ever-changing theories about all those issues - and that's because, in its stubbornness, science refuse to accept that it simply cannot deal with such things.

You really lay into the ol' scientists, Nicky! I hope you don't think all these negative things about your mother the physicist!

Of course there can be stubbornness among scientists. But there's quite a bit of stubbornness right here on the ol' unscientific Soundsonline Forums too, my brother, so I don't think that scientists have a monopoly on the quality.

I would be more unhappy with science if its theories about complexities like mind and emotion did not change. Mostly, science has "ever-changing theories" because good science should always be open to improving its theories to better fit new evidence. This tendency toward self-correction in science is one of its most defining features.

And let me repeat that (perhaps unlike some other participants in this discussion) I do not make my statements with certainty. Maybe one day human-built machines will be able to write music as well as a decent human composer (or serve as a knowledgable and helpful assistant to a human composer), but maybe not. But I'm not deciding what can or can not be in advance: like a good jury, I'm awaiting the evidence. It's an ongoing trial, we're all watching, and, yes I am optimistic--not for a specific verdict, but that along the way we'll learn a lot, probably be entertained, and at the very least a lot of fun forum posts will get written.

Nothin' but love,

Joe

P.S. can you tell me why technology cannot make an object fly like a bird?

Well, I can't tell you this, Nicky...because technology can make an object fly like a bird. (Maybe you can tell me why this "can't" be done. ;) ) These devices are called "autonomous ornithopters". Advanced designs that use computer vision to fly on their own have already been tested. A starting point for more info: (click on "Research Areas", then scroll down to "Flapping Wing Applications")
A cool NASA research group (http://research.nianet.org/morpheuslab/Morpheus%20Home.htm)

P.P.S. If the machine freely willed to change my kids' diapers, I'd definitely leave them together.

nickysnd

02-21-2007, 06:57 PM

I agree that my statements may sound unpolite, but they are the expression of my disapointments about science, I don't mean any personal offense. Of course, I have nothing but honest opinions and "strong" beliefs, but I find scientific optimism about domains as art, thought, and emotion far too naive. That robot you mentioned didn't read any emotion at all, that was just sorta "playing cards" game, he has just "read" external signs of specifically isolated emotions (as were implemented in its memory by biased reserchers) and then it outputed the best probabilities. Not even the subjects themselves knew what emotions they felt, nevermind a robot. Of course psychologists can't do that neither.

About birds' flight - no, technology can't replicate it. Your link gave me: "Under Construction by Brian R. (sorry)" Sorry, indeed, but it's very embarrasing for science in general to make use of such tricks for hiding its shortcomings. So, technology can't make an object fly like a stupid little birdy, and I'll give you the reason why: because your precious science can't explain even elementary/basic physics, and things that are right under our nose daily - like birds' flight. Plainly - scientists are incapable to tell how birds do fly - now how about that? You cannot tell me how birds fly but you can make robots that know about my emotions?!? Come on!...

As for your trust on robots' free will, pardon me, but I don't think you were totally honest - you were kind of elusive in fact. Yes, they can freely will to change the diapers, but also, they might just as well freely will to exercise some brain surgery on them, just for the sake of scientifical advance (or, if they can't do that, then they don't really have free will). After all, humans can make other children, can't they? I apologize for this horrible example. Free will implies the possibility to do whatever one wants, murder included - there can be no restriction whatsoever within free will. So no, I would never allow such a machine come close to my kids.

gstitt

02-21-2007, 08:06 PM

The point of much current research in AI is that we don't have to "figure out how humans work", at least not very completely, to create a machine with very human behaviors.

I agree in some cases, but for image recognition, I would guess that computers suck because they are using a very bad "algorithm" compared to how humans do it. So, for image recognition, I would think we would have to figure out how humans efficiently process a knowledge base based on visual information to very quickly identify an object.

Existing image recognition techniques are very limited. Even for cases that work, you can often break it simply by changing lighting or rotating an object.

gstitt

02-21-2007, 08:10 PM

Maybe not by 2029. But definitely by 3:34 pm on March 15, 2036.

lol :)

Josta

02-21-2007, 09:28 PM

Nicky, I certainly agree that free-willing agents can be a threat--but that's a sad lesson we've learned first and foremost by the behavior of people, not 'bots. Cause, ay caramba!, when people go bad, they can break in some seriously disturbing ways. I probably would worry less about a free-willing, emotionally-aware Roomba molesting a child than a person doing that. Or driving the children into a lake. Or killing strangers who come to the door and eating them. Because, you know, those are known failure modes for people.

The link included in my last post was to the top level of a lab that does many different kinds of research in developing mechanized winged flight. There was one blind link: the fact that you interpret that as a "trick" by "science" saddens me. If you browse around that site, among other things you'll find info on the "Odyssey" ornithopter:

http://research.nianet.org/morpheuslab/images/Jared_Odyssey_overhead.jpg

Woe be to the real-life bird that tries to mate with that contraption!

There are many groups developing bird-like robots. Here's a detailed description of work at Cal Tech to develop the flying "Microbat":
Wing Technology for a Battery-Powered Ornithopter (http://touch.caltech.edu/people/past/grad/nick/mems00.pdf)
The paper also describes very specifically some of what science knows about how wing-flapping animals fly--which is actually quite a bit.

Maybe you don't think bird-like robots can be made. But apparently birds disagree with you:

"Researchers performing flight tests with the MicroBat said it tended to attract small birds when it ran low on power and fell to the ground. The birds clustered near the floundering ornithopter in what seemed to be a desire to help."

Roz Picard's group's "Self-Cam" system is most definitely reading emotion (or, to be more precise, "affect", the physical display of emotion) by any reasonable definition of the term. It examines a face, then determines if the person is, say, confused or concentrating. It did not have this information placed in its memory by "biased" observers: it learned by example. It was tested by comparing what it determined a person's emotional state was to what the person being observed said they were experiencing at the time. The system did this more accurately than 94% of human observers.

And, seriously, why would you think we don't--no: can't--understand how birds fly? We understand pretty well how humans walk--enough so that we can make artificial hips and knees and even neuroprosthetics that can enable a paraplegic to stand and take steps. Why are birds extra-special?

Cheers!

Joe

Josta

02-21-2007, 09:56 PM

I agree in some cases, but for image recognition, I would guess that computers suck because they are using a very bad "algorithm" compared to how humans do it. So, for image recognition, I would think we would have to figure out how humans efficiently process a knowledge base based on visual information to very quickly identify an object.

Existing image recognition techniques are very limited. Even for cases that work, you can often break it simply by changing lighting or rotating an object.

A weakness of many computer image recognition systems is precisely that they use preprogrammed algorithms, rather than learning by example. Computers are actually pretty good at a variety of types of image recognition. For example, face detection (face recognition is harder), font-independent text scanning, identifying and picking up randomly-oriented parts in an assembly line, and biometrics like retina and fingerprint scans. All these are very accurate. Face detection, especially, has come a long way lately--there are some recent systems that are pretty good even with dramatic lighting changes.

On the other hand, just making letters sort of twisty can defeat earlier text recognition systems--this is the basis for "CAPTCHA's", those wavy characters you're asked to identify on a web site to authenticate that you're a person and not a bot. Again, as with so much of current AI, success at recognizing CAPTCHA's is now being achieved by sytems (typically neural networks) that learn by example, rather than earlier methods that relied on explicit algorithms.

I've seen some pretty robust recent research systems that, when looking at a real-world scene, are much less sensitive to lighting and rotation. They are stereoscopic, so there is better depth perception, and some of them "move their heads", the way a person might do to get a better look at a visually ambiguous or complicated scene. I don't know if any of this work has made it into commercial systems yet, though.

To bring this thread back to music (imagine!) I think there have been a few attempts to track musical gestures using computer vision systems--like recognizing the movement of a conductor's baton or the bow of a cellist. There's a simple mouth-motion-to-MIDI system out there: the Mouthesizer (and no, I'm not just making that name up!). It just looks at (using a video camera) and recognizes (by applying some simple algorithms) the size and shape of the "dark" area of a performer's open mouth, but it can produce some fun control possibilities:

Mouthesizer: mouth control of filter cutoff (http://www.archive.org/details/MouthesizerVowelFilterLoop)

Joe

nickysnd

02-21-2007, 11:53 PM

But apparently birds disagree with you:
"Researchers performing flight tests with the MicroBat said it tended to attract small birds when it ran low on power and fell to the ground. The birds clustered near the floundering ornithopter in what seemed to be a desire to help."
Yes, I think that not technology or science, but only birds can help, but not as yet... :) Also, from the link you provided:

"We find that there are several challenges in order to achieve a successful sustained flight. First, the wind condition must be perfect. Often during flight test, the wind speed and direction shifted constantly. Second, the trim of tail stabilizer must be crucial. Finally, each launch motion must be the same. We also believe that our current wings and ornithopter are not optimized, thus we hope future flight duration can be improved."

Which is precisely my point - people cannot (OK, currently cannot) make an object fly like a bird. I mean, doing alone, without human help (no remotes) the following tasks: the object is put down, it takes off, flies up (no matter the direction wind blows), reaching a tree, flowing through the branches, going a bit farther, coming back, landing on a branch, staying there a few seconds, then flying to another, upper branch, taking off from the second branch and flying to another tree, and so on... no human made object is capable of what I've described - and that is what a little stupid bird is doing like 16 hours a day, continuously. What technology?! I think birds also disagree with you too! :)

I have searched the internet for more than one hour and, while a lot of plans and sketches and fancy explanations, and discussions and articles, I have never seen one filmed ornithopter that would fly without propellers, and would do the tasks described above. And that, after more than one hundred years of airplains, rockets and helicopters. It is simply ridiculous...

gstitt

02-22-2007, 07:47 AM

A weakness of many computer image recognition systems is precisely that they use preprogrammed algorithms, rather than learning by example.

Agreed, 100%.

Computers are actually pretty good at a variety of types of image recognition. For example, face detection (face recognition is harder), font-independent text scanning, identifying and picking up randomly-oriented parts in an assembly line, and biometrics like retina and fingerprint scans.

Detection is certainly easier than recognition, as recognition requires development of a knowledge base, like you mentioned before. I'm guessing detection will improve drastically in the next decade, since it generally involves analysis of spatial frequencies to find some predefined characteristic, which all falls into the domain of image processing, which is a somewhat mature field of research. Machine learning on the otherhand, which is still in it's infancy in terms of research IMO, is a long ways off from being able to make recognition of general objects possible.

Josta

02-22-2007, 08:31 AM

Nicky, your set of features for a "robotic bird" might be considered a "Grand Challenge Problem" for micro aerial vehicles (MAV's) or autonomous ornithopters. I could have saved you that hour of searching at 2am!: absolutely, there is no system I know of that has been built today that can perform everything you described. But that proves, well, it "proves" only that there is no system I know of that has been built yet that can perform everything you described. :rolleyes: Many of the individual pieces of what you described (small electromechanical winged flight, autonomous flight, computer vision for navigation) do exist, and there is no reason to think that, with appropriate funding and development, they couldn't be developed into a system to fly, navigate and land on a branch of a tree. Again, making a "RoboBird" is an extremely challenging technical problem, and research in robotic ornithopters is in its infancy--yet there are already demonstrated early systems like I linked to. Planes, helicopters and rockets are obviously extremely different from birds, so that's an apples-and-oranges comparison: it's not like we've been working on flying winged bird-bots for a century!

There's a large--very large--difference between saying a design goal is challenging and saying it's impossible.

There are presently similar grand challenges in robotics: far from leading people to shake their heads and give up, saying "It's too hard! We'll never be able to do that!", it leads researchers to roll up their sleeves, get to work and solve the problem.

Case in point: in 2004, DARPA sponsored the first autonomous vehicle (AV) grand challenge: a 142-mile course in the Mojave desert in California which the robotic vehicles had to drive entirely on their own--no radio control, no human intervention. Despite the lure of a $1 million prize to the team that finished first, no vehicle finished the course. In fact, the vehicle that got the farthest, Carnegie Mellon University's "Sandstorm", only got a little over 7 miles.

So, how did the engineers and scientists react to this? After years of research, millions of dollars, and more than 100 years of planes, trains and automobiles, the BEST AV was only able to make its way 7 miles over a 142 mile course? Pathetic! Ridiculous!

Well, not so ridiculous as it turns out. Just one year later, FOUR vehicles completed the 132 mile course of the 2005 DARPA Grand Challenge. The winner (and team that received the $2 million first prize) was a robotic Volkswagen Touareg from Stanford.

There are people who choose to put their energy into naysaying and trying to "prove" that something can't be done.

Then there are the people who regard a challenging problem as a spur to creativity and innovation, and get to work on understanding the problems and inventing the future, like better understanding emotion or making an autonomous flapping-winged ornithopter.

I don't see how the naysaying approach benefits us.

The imaginative and creative people who follow the second approach have given us airplanes, weather satellites, space travel, telephones, people walking on the moon, robots photographing mars, sound recording and playback, photography, motion pictures, vaccines, organ transplantation, artificial hips to restore mobility to the arthritic and artificial cochleas to bring hearing to the deaf, antibiotics, silicon microcircuitry, hard drives that store an orchestra's worth of sounds in the space of a couple of packs of playing cards, computers the size of a box of chocolates, and the ability to conjure a cello from a keyboard.

I like the results of that approach!

Joe

Josta

02-22-2007, 09:26 AM

Machine learning on the otherhand, which is still in it's infancy in terms of research IMO, is a long ways off from being able to make recognition of general objects possible.

You know, that may well be true. But there's a new wrinkle, as something interesting has happened recently. With the development of AI systems that can break the CAPTCHA squiggly-letters authentication system, some security systems are moving to object-recognition for checking if a visitor to a web page is a person or a bot: show a picture of a frog, say, and ask the visitor "What is this?"

If anything will promote innovation in computer science, it's the deployment of a new security system! It's practically an engraved invitation saying "Break me!"

General object recognition is a seriously hard problem to be sure. But if "necessity is the mother of invention", a lot of invention is probably going on right now to break these object-recognition systems. Stay tuned!

Again, just so that maybe one out of every 100 sentences in this thread has something to do with music :rolleyes: : I wonder if anyone has developed (or tried to) an image recognition system that could "read" a pencil-and-paper manuscript and translate it into a computer-based score? Maybe programs like Sibelius have made the market for this obsolete.

Joe

Josta

02-22-2007, 10:42 AM

I wonder if anyone has developed (or tried to) an image recognition system that could "read" a pencil-and-paper manuscript and translate it into a computer-based score?

OK, yes, of course someone is working on this:

Teaching a computer to read all the bars in the world (http://www.kcng.org/omr/)

Wow, actually a lot of work has been done on OMR (Optical Music Recognition) systems:

OMR Bibliography (http://www.kcng.org/omr/)

Anyone ever seen this in action? Have they got it working yet?

Joe

nickysnd

02-22-2007, 02:01 PM

Wow, actually a lot of work has been done on OMR (Optical Music Recognition) systems:

OMR Bibliography (http://www.kcng.org/omr/)

Anyone ever seen this in action? Have they got it working yet?

Joe
Joe, first thanks for enlighten me about the bright perspectives of imitating birds flight! :) One thing that stuck to my mind from my hour of research: by 1910 engineers realized that the project of a bird-like flying machine is dead. So, naysaying have some advantages, like: give up this nonsense, focus on what you CAN do!

Now on handwritten music sheet recognition:

I used to collaborate on something like that - a very frustrating period (I think then I started to lose my faith in technology) that opened my eyes on the paradox of the huge complexity of written music that is so easily accessed by our minds. I was like a musical theory "specialist" counseling a group of programers that were building a musical software. One of the basic and most annoying questions that I had to deal with each and every day was: "How do you know what you know?" You know, things that you take for granted, but you can't explain to other people, like riding a bike. (I guess I am a very bad teacher, but, with one exception, all my music teachers were worse than I am.)

From those daily challenges I have learned many things, and one followed me as a fundamental principle: improve what already seems to work somehow, as opposed to try to find other ways to accomplish that same task. Those programers seemed very fond on the opposite approach - instead of deepening the knowledge and digging together in one direction that seemed promising, they were preferring to waste time trying to find parallel ways. I think this is the very mentality that led us to having, after decades, millions of computers, none working properly.

Anyways, here's what I see: music recognition is in its early infancy. As a quasi-professional Finale user, I can tell you that, instead of saving time, music recognition was always wasting my time - I am doing faster typing the notes and all the symbols by hand. I think we are better off by improving those machines and softwares as tools for achieving better readability and speed, and typing comfort, than to waste time into the illusory path of making them perform our tasks. Can a chess program make music? Call me naysaying, but I think a better investment of time and money would be in researching and producing better keyboards and better thought shortcuts algorithms, so we can type easier and faster, and use the gained time for making music.

Teaching a machine to compose music? First a machine should prove able to feel music, then think music, then read music, then write music in a way that others would consider sensitive, emotional, spiritual, and meaningful. Allow me to be very skeptical about the possibility of even the first step. Yes, that's the main question: can a machine learn to feel music? Not feel the music, but feel music, if you follow me. But first, how do WE (you, me, etc.) do feel music, huh? And generally, what do we know about what we know, about how we know it, and about what and how we feel?

Those are fundamental questions that engineer-minds prefer to ignore, as those type of questions seem to threaten their jobs. When engineers will learn from a bird how does it feel to fly, then they will be able to invent a bird-like machine. Till then - only palliatives like helicopters, planes, rockets. Which are fine, but kind of silly, comparing to the freedom of a bird.

"A lot of work has been done on OMR (Optical Music Recognition) systems"?? Yes, probably a lot. And how useful does this work prove so far? If you ask me - a waste of time and money in a wrong direction.

A huge discussion, to a more mundane level, can be: how do we hear music? Do you know that a machine cannot recognize a trumpet in an orchestral tutti? What chess program doing what?? - composing music??? Ha! First: can a machine hear music?

Technology is in the Stone Age. Music is in the Golden Age. Why do I think they will never be compatible? - 'cause engineers don't try to feel the music, preferring to waste their time in non-musical directions.

Composing_automat

02-23-2007, 01:26 AM

There is still much to do but this is the best one can find when speaking about computers as composers:

www.synestesia.fi

Weakest points?

Counterpoint

03-07-2007, 01:17 PM

There is still much to do but this is the best one can find when speaking about computers as composers:

www.synestesia.fi

Weakest points?

... what does this have to do with a computer "composing" music? From what I understand a source image is provided as input and then algorithmically rendered into music.

Weakest point being that the computer is not doing anything that even remotely resembles "creative decision" or "artificial intelligence". This is 100% mathematics...

- Matt

Composing_automat

03-07-2007, 11:43 PM

... what does this have to do with a computer "composing" music? From what I understand a source image is provided as input and then algorithmically rendered into music.

Actually all comnposers are using a lot of algorithms even if they don't think it that way. Without algorithms the music can't have any coherence or style. Emulation (John Williams ...) is also based on analytic algorithms used without shame in implementations).

Weakest point being that the computer is not doing anything that even remotely resembles "creative decision" or "artificial intelligence". This is 100% mathematics...

Most of the greats have been using number magic (Fibonacci ...) or letter magic (b-a-c-h ...). One could call it creativity, or mental cruches. Or in some cases synesthesia perhaps?

Composing_automat

03-08-2007, 12:15 AM

Teaching a machine to compose music? First a machine should prove able to feel music, then think music, then read music, then write music in a way that others would consider sensitive, emotional, spiritual, and meaningful. Allow me to be very skeptical about the possibility of even the first step. Yes, that's the main question: can a machine learn to feel music? Not feel the music, but feel music, if you follow me. But first, how do WE (you, me, etc.) do feel music, huh? And generally, what do we know about what we know, about how we know it, and about what and how we feel?

Actually everything just happens other way round. Teaching (or implicitly putting music around) people to LISTEN music. Poor people living in India have not teached to listen western music or western scales. They don't "feel" it. There is no universal "feel" of music or "thinking" of music or "reading" of music or "writing" of music". There are just almost infinite number of ways doing that all. Any sounds can activate feelings. Structures of sounds even more. Composing music is a creative act but so is also playing music and - LISTENING music.

Perhaps the main claim against computers as composers is "the programmer problem". They say "it is the programmer who is the composer".

When there are let's say 100 small algorithms working together, the programmer is perhaps not able to tell what comes out. Then those algorithms eat that data of let's say a photo 400x400 pixels there are more that million bytes of data to work on. If the process of composing lasts for 5 seconds every second tens of those little algorithm beasts work on some hundred thousand bytes. ANY CORRESPONDING NUMBERS FOR YOUR HUMAN COMPOSERS? But the question is, is this all "visible" to the programmer? Even in principle?

(to be continued)

Counterpoint

03-08-2007, 12:17 AM

Actually all comnposers are using a lot of algorithms even if they don't think it that way. Without algorithms the music can't have any coherence or style. Emulation (John Williams ...) is also based on analytic algorithms used without shame in implementations).

Yes, a composer uses algorithms and mathematics (cadences, scale systems...) to construct a musical work, but there is a decision process involved here. A composer might listen to a passage they wrote and decide to change this note or that note arbitrarily because it creates a more interesting sound, where an algorithm by itself is incapable of this.

Most of the greats have been using number magic (Fibonacci ...) or letter magic (b-a-c-h ...). One could call it creativity, or mental cruches. Or in some cases synesthesia perhaps?

Ok, fair enough but can you provide an example where a computer "invented" the idea to use a Fibonacci sequence or to apply mathematics to letters in a name by itself... to me this is a creative decision and I have not seen any example yet of a computer arriving at the conclusion that this would be a good idea. I've seen computers that use these techniques, but only because they were programmed to. They did not suddenly decide to do this without a programmer's input.

The difference is that a living composer (algorithmic or otherwise) is making decisions in order to achieve the goal which is a musical work. An algorithm will do exactly what it is meant to, which means that unless there is some random element involved the algorithm will create the SAME work given the same inputs.

Given the example of the algorithm that creates a score based on an image, isn't it safe to assume that there are no random elements involved and that the same image will produce the same score each time (otherwise the algorithm would be pointless as it would not be creating a true audio "fingerprint" of that image)?

I would give the composer/programmer full credit for inventing this idea (image->music), but it was not a machine that suddenly felt inspired to create music based on images. This was a decision made by the programmer. The algorithm does not make intelligent decisions, it just behaves in a predictable way.

- Matt

Counterpoint

03-08-2007, 12:24 AM

When there are let's say 100 small algorithms working together, the programmer is perhaps not able to tell what comes out. Then those algorithms eat that data of let's say a photo 400x400 pixels there are more that million bytes of data to work on. If the process of composing lasts for 5 seconds every second tens of those little algorithm beasts work on some hundred thousand bytes. ANY CORRESPONDING NUMBERS FOR YOUR HUMAN COMPOSERS? But the question is, is this all "visible" to the programmer? Even in principle?

In one word: yes.

This is called "stepping through the code"... any programmer can do that, it just takes a long time compared to the computer just executing it. The point is that any input value can be traced through the function to arrive at the output value.

- Matt

Composing_automat

03-08-2007, 12:46 AM

Given the example of the algorithm that creates a score based on an image, isn't it safe to assume that there are no random elements involved and that the same image will produce the same score each time (otherwise the algorithm would be pointless as it would not be creating a true audio "fingerprint" of that image)?

There are many phases in real composing and presentetion of a piece. Someone will commission the pieces (perhaps ordering a certain instrument set). The premier will be in some special occation giving some contraints on the mood of the piece. The abilities of the musician may influence the piece . The piece is asked to be arrenged for different sizes of orchestras etec. So there are always many parameters around. So nothing is safe in real life.

I would give the composer/programmer full credit for inventing this idea (image->music), but it was not a machine that suddenly felt inspired to create music based on images. This was a decision made by the programmer. The algorithm does not make intelligent decisions, it just behaves in a predictable way.

The idea fo image->music have been used by many, even great composers (graphical scores ets.) . One of them (Iannis Xenakis) even implemented it on computer:

http://membres.lycos.fr/musicand/INSTRUMENT/DIGITAL/UPIC/UPIC.htm

And one can buy a double CD of the result of 7-8 composers using that UPIC-system.

Composing_automat

03-08-2007, 12:52 AM

In one word: yes.

This is called "stepping through the code"... any programmer can do that, it just takes a long time compared to the computer just executing it. The point is that any input value can be traced through the function to arrive at the output value.

- Matt

Of course. But using the stepper of e.g. Jbuilder it would take more than one's life time. So why not speak about EMERGENCE. That's what is also happening in the minds of human composers having a network of billions of neurons.

nikolas

03-08-2007, 01:14 AM

Okie... Lemme join (again) this thread...

My doubts (although I do support the idea), is that computers, programs, whatever cannot make a simple choice of yes/no, based only in their "instincts", or "aesthetics". YET!

When you have a computer playing chess, it's just putting down what will happen autisytically with itself:

"Should I move the bishop one step? YES/NO
IF I move it then this will happen or this will happen or this, and then this or this or...
Answer: NO
Should I move the bishop two steps ahead? YES/NO
blah blah..."

Over simplified, and I'm sure that there are shortcuts, but still that's the basic idea.

You get a great series of notes/rythms/orchetsration/range/tempo from an algorithm, whether this human, or computer. The final question for a human composer is kinda this:
"Hmmm.. somehow that C# over there does not work. Should I change it? YES/NO
Maybe I'll have a coffee first and play it in the piano, to see if I like something better..."

That's what the computers are primarily missing here!

The 2nd thing that I keep posting about is the message. Music is about communication. What does a machine have that it would like us humans to know about? Cause whenever I write something, I always try to have the "message" there, whether this is a sentiment, a writen message, another kind or whatever. Apart from the sound and the (possible) pleasure, there needs to be something additionally for music to do as it should. Otherwise it's not doing the right job (as many many humans don't... how many people really want to work in a bank? :S) If that additional thingy is missing, then it's not doing it's work properly.

For my definition of music (btw): any sounds that are organised in some fashion, or even philosophised to be random (aleatoric for example). anything beyong that is not music. So someone sampling bird sounds is making music, the birds themselves are not.

For me all that!

Composing_automat

03-08-2007, 02:56 AM

The 2nd thing that I keep posting about is the message. Music is about communication. What does a machine have that it would like us humans to know about? Cause whenever I write something, I always try to have the "message" there, whether this is a sentiment, a writen message, another kind or whatever. Apart from the sound and the (possible) pleasure, there needs to be something additionally for music to do as it should. Otherwise it's not doing the right job (as many many humans don't... how many people really want to work in a bank? :S) If that additional thingy is missing, then it's not doing it's work properly.

About the communication. There are two main classes of composers. Those who 1) say that they just compose for themselves ("the future will show...") and those who 2) want to communicate.

The problem with the first group is that they perhaps just want to justify bad music. The problem with the second group is entertainment and cloning...

"Additional" in music? Perhaps in the future "empirical musicology" will show if there is anythin "additional". My guess is that there is none. Music does'nt have semantics. Without semantics you have only something influencing you lower brains. Then you have only signaling on animal level. It can't be very selective nothing to speak about intelligent composing decisions.

nickysnd

03-08-2007, 06:05 AM

Actually all comnposers are using a lot of algorithms even if they don't think it that way. Without algorithms the music can't have any coherence or style
Your algorithms are as low and ridiculous as music theory rules are - a stupid/delusional result of the analysis of the notes on paper. Music creation may only appear to use algorithms and theoretical rules. You can't really decompose/recompose music. You can only compose it, if you are gifted.

Composers laugh at "using algorithms" and "coherence of style". Music is true magic, and composers, creative artists in general, are not illusionists whose tricks you might figure out and learn - they are sorcerers. You can figure out nothing about the inner power of a sorcerer. What you're doing is walk on holy ground and trying to photograph God. To understand music you have to be much more than a priest, you have to be a monk.

There are many phases in real composing
What you do know about the process of creating music is corrupted by your mechanical/deterministic approach of it.

Nobody rationally knows more than 1% about the process of music composing. It is like an iceberg, only that one cannot penetrate under the "water" surface to see what is there. One might be able to see, say, about another 1% under the "water", but that would be just a distorted refraction. You seem to even fail to see/aknowledge that "water" barrier. So, no - that 1% is not the whole thing. There is a hidden alchemy to composing music, which is not accessible to the rational mind.

The "water" I am talking about is impenetrable for the type of knowledge you seem to exclusively hold to. I have been under that water, not too far and not for too long, but enough to be able to tell you that there is a amazing universe in there, which I cannot describe to you in algorithms.

Perhaps the main claim against computers as composers is "the programmer problem". They say "it is the programmer who is the composer".

No, a programmer can be a composer only after he stops thinking like a programmer.

But the main problem of objects incapable of artistic creativity is that artistic creativity is a direct result of free will, which is a human exclusivity. No free will - no artistic creativity. When an object, a plant, or an animal will be capable of free will, then they might enter in an inner state that might enable them to create art. Until that moment, they cannot be not even spectators of art, nevermind creators.

Composing_automat, you insist to advocate here a mechanical/deterministic approach on artistic creativity. It really is childish and it doesn't hold together, it desintegrates when exposed to the powerful "radioactivity" of music. Music cannot be digital, only analog. Music creation doesn't use isolated elements, but "continuums", which you seem to ignore completely. Bach's "Johannes Passion" was not created out of notes/sounds/algorithms/music rules. Van Gogh's "Starry Night" was not created out of lines/colors/painting rules. Ibsen's "A Doll's House" was not created out of letters/words/grammar/writing rules. Etc.

Nothing about music can be explained by using algorithms, nothing at all. On the path you are currently walking you will learn nothing about music/creativity, and, which is worse -- it is a dead-end path.

A good place to start learning about music is: Time. Time is a continuum, so forget the digitized "time" within computing machines, or else, you will go on decomposing "time" in hours/minutes/seconds/miliseconds/nanoseconds etc. - but this way you will learn nothing about the art of modeling Time through music, cinema, etc. That is truly wizardry.

Josta

03-08-2007, 07:27 AM

Lauri,

I had a listen to several of your pieces last weekend and I actually found some of them to be quite interesting, and more "musical" than I might have expected. As just one example, I was quite moved by some of the spare, haunting string textures in "Reading Letters (Trixie)", particularly at the beginning and in some parts of the middle third. I'm a huge fan of Arvo Part, and some moments there evoked some of his textures.

That being said, I also agree with Counterpoint that I would be more inclined to classify Synestesia as more of a "translator" program than a composing program.

If I understand correctly, it appears that your program takes advantage of the structure inherent in images of real-world scenes and translates, or transduces, that visual structure into sonic structure. The name of the program is therefore very appropriate, as this is analogous to what can occur in the brains of people who are visual-aural synesthetics.

As a result, there are what I might call "microstructures" in your pieces that are quite beautiful and interesting musically: short motifs, interplay between instruments, textures, etc. I could imagine this being possibly useful as a "palette" resource for a composer. That is, a tool that a human composer could play with to explore some new elements and ideas.

Among the qualities I find lacking is a sense of "macrostructure": a sense of thematic and harmonic development and the overall emotional arc that is characteristic of human music. There does not appear to be much (if any) tempo variation in the pieces I auditioned. Also the dynamics are compressed: there doesn't appear to be much overall modulation of volume. Some sustained notes, such as the flute tones, are also not modulated, giving them a somewhat sterile quality. Most of the pieces I listened to have an atonal character to them. Personally, (and with all due respect to Schoenberg and Webern!) I find that I respond much more strongly emotionally and am more engaged by tonal music. But that's just me.

As a photographer, I note that many of the images you employ are relatively simple snapshots. Do you feel that there is a correlation between the quality of the image and the quality of the resulting music? For example, does a well-composed photograph with both good overall composition and interesting small-scale details and textures, and with a wide contrast range, translate into a musical piece that echoes some of those qualities?

Final question: does the program assign the instruments? Or do you assign them? Have you experimented with, say, mapping color or another image quality to instrument/timbre? Oh, and have you ever had actual musicians play one of your pieces? I would find it very interesting to hear how one of your pieces would be interpreted by human players. Say, Kronos and James Galway doing "Trixie"!

OK, I guess that's like 4 questions...:rolleyes:

Joe

P.S. I played "Trixie" for my fiancée, who is a choreographer. In classes, rehearsals and as a teacher, she gets to hear live musicians playing nearly every day, and in general, she hates "computer music". (I've learned to stop dragging her to such concerts here in Boston!) But she stopped what she was doing as "Trixie" played and commented "huh...there are actually some interesting moments in there." Just thought I'd pass that along...

Software

03-08-2007, 01:13 PM

That being said, I also agree with Counterpoint that I would be more inclined to classify Synestesia as more of a "translator" program than a composing program.
Josta, Thank you for your comments. It's ok if we are looking at the process. If thinking about the end result, I wouldn't be so sure.
As a photographer, I note that many of the images you employ are relatively simple snapshots. Do you feel that there is a correlation between the quality of the image and the quality of the resulting music?
I guess not. I mostly select pictures that some some meaning to me.
For example, does a well-composed photograph with both good overall composition and interesting small-scale details and textures, and with a wide contrast range, translate into a musical piece that echoes some of those qualities?
That depends on how many times the picture is "filtered". That parameter can be changed even if most often the picture is "filtered" only once.
Final question: does the program assign the instruments? Or do you assign them? Have you experimented with, say, mapping color or another image quality to instrument/timbre?
No, it does not. With the limitation of different picth ranges of different instruments, the score has differences when different instruments are used. The user can decide what is "commissioned". Accepting missed or useless notes is it possible to change the instruments during playing phase of the piece.
Oh, and have you ever had actual musicians play one of your pieces? I would find it very interesting to hear how one of your pieces would be interpreted by human players. Say, Kronos and James Galway doing "Trixie"!
Not yet fully. Only a solo with background SW instruments.

Josta

03-08-2007, 09:07 PM

I've seen computers that use these techniques, but only because they were programmed to. They did not suddenly decide to do this without a programmer's input.

The difference is that a living composer (algorithmic or otherwise) is making decisions in order to achieve the goal which is a musical work. An algorithm will do exactly what it is meant to, which means that unless there is some random element involved the algorithm will create the SAME work given the same inputs.

Hey, Matt. I've enjoyed reading your comments in this thread.

Several people here have made statements like "computers can only do what they are programmed to do" to support the argument that this somehow prevents computers in a fundamental way from ever being able to emulate certain more advanced aspects of human behavior.

But there are many people who will argue that the same sentence could be applied to brains: "brains can only do what they are programmed to do", where the "programming" in this case is a combination of the person's DNA and their lifelong experience with their environment (including, of course, all those long years of study at the conservatory). These two elements create a brain that is wired in a unique and utterly complex way, resulting in a 3-pound piece of meat that can think, feel and compose music.

There's a hilarious short story (actually a one act play) called "They're Made Out of Meat" by Terry Bisson. In it, a scout for an alien race reports back to a colleague about what he found when he visited Earth. He's trying to convince the other alien that there is intelligent life on Earth and--seriously--the sentient beings are made of meat. In fact, they even think with meat.

Voice One: "They're made out of meat."

Voice Two: "Meat?"

Voice One: "Meat. They're made out of meat."

Voice Two: "Meat?"

Voice One: "There's no doubt about it. We took several aboard our recon vessels from different parts of the planet and probed them all the way through. They're completely meat."

Voice Two: "That's impossible. What about the radio signals? The messages to the stars?"

Voice One: "They use the radio waves to talk, but the signals don't come from them. The signals come from machines."

Voice Two: "So who made the machines? That's who we want to contact."

Voice One: "They made the machines. That's what I'm trying to tell you. Meat made the machines."

Voice Two: "That's ridiculous. How can meat make a machine? You're asking me to believe in sentient meat."

Voice One: "I'm not asking you, I'm telling you. These creatures are the only sentient race in that sector and they're made out of meat."

...

Voice Two: "No brain?"

Voice One: "Oh, there's a brain all right. It's just that the brain is made out of meat! That's what I've been trying to tell you."

Voice Two: "So ... what does the thinking?"

Voice One: "You're not getting it, are you? You're refusing to deal with what I'm telling you. The brain does the thinking. The meat."

Voice Two: "Thinking meat! You're asking me to believe in thinking meat!"

Voice One: "Yes, thinking meat! Conscious meat! Loving meat. Dreaming meat. The meat is the whole deal! Are you beginning to get the picture or do I have to start all over?"

Voice Two: "Omigod. You're serious then. They're made out of meat."

Voice One: "Thank you. Finally. Yes. They are indeed made out of meat. And they've been trying to get in touch with us for almost a hundred of their years."

Voice Two: "Omigod. So what does this meat have in mind?"

Voice One: "First it wants to talk to us. Then I imagine it wants to explore the Universe, contact other sentiences, communicate, swap ideas and information. The usual."

Voice Two: "We're supposed to talk to meat."

Voice One: "That's the idea. That's the message they're sending out by radio. 'Hello. Anyone out there. Anybody home.' That sort of thing."

Voice Two: "They actually do talk, then. They use words, ideas, concepts?"

Voice One: "Oh, yes. Except they do it with meat."

Voice Two: "I thought you just told me they used radio."

Voice One: "They do, but what do you think is on the radio? Meat sounds. You know how when you slap or flap meat, it makes a noise? They talk by flapping their meat at each other. They can even sing by squirting air through their meat."

Voice Two: "Omigod. Singing meat. This is altogether too much..."

For some people, the thought of a non-biological machine that replicates the functions of the brain may appear just as absurd as "Voice Two" finds the notion of "meat that can think".

'Cause when you think about it (using your brain meat) or speak of it (by flapping around some of your other meat) thinking meat does sound kinda ridiculous. And yet, think it does. :rolleyes:

Joe

They're Made Out of Meat by Terry Bisson (http://www.terrybisson.com/meatplay.html) for the rest of the story.

Counterpoint

03-09-2007, 12:44 AM

Hi Joe,

Several people here have made statements like "computers can only do what they are programmed to do" to support the argument that this somehow prevents computers in a fundamental way from ever being able to emulate certain more advanced aspects of human behavior.

Well, you've got to admit those people have a point... our present-day computers are not capable of "self-motivated" activities. :)

I would also note that "emulating certain more advanced aspects of human behaviour" is possible. We've got lots of "intelligent chatbots" for example that can trick someone for a while into thinking they are talking to another person (usually they eventually fail though because at some point they stop coming up with new things to say...). These chatbots do emulate human behaviour. However, we're still not able to truly simulate human behaviour.

The main problem is that our computers are not really designed to be "simulated brains", they are designed to take a set of instructions and process them.

But there are many people who will argue that the same sentence could be applied to brains: "brains can only do what they are programmed to do", where the "programming" in this case is a combination of the person's DNA and their lifelong experience with their environment (including, of course, all those long years of study at the conservatory). These two elements create a brain that is wired in a unique and utterly complex way, resulting in a 3-pound piece of meat that can think, feel and compose music.

Well, this is a bit difficult to prove still because we don't know enough about how the brain works to be able to say that it is a very complex algorithmic machine. Science can not prove this yet, so allow me to move along to a more philosophical idea...

Suppose the universe is based on an equation (possibly infinitely complex). This would suggest that everything in the Universe is following a predictable path and that any state can be computed if you input the right parameters.

Now suppose that we somehow figure this equation out. Now let's say I use the equation to answer the question "What am I going to do next?".

If the algorithm replies "Stand on your head", then what happens if I refuse? Do I suddenly find myself standing on my head even though I decided not to? Does the universe abruptly turn blue with some indecipherable error message in the middle of it and wait to be restarted? ;)

"Free will" is not something that can be defined by an equation because by definition "free will" gives the option to outright defy the equation.

So... until we figure out how to give computers this "free will", they will remain incapable of true simulation of human behaviour. (emulation is very possible though!)

There's a hilarious short story (actually a one act play) called "They're Made Out of Meat" by Terry Bisson. In it, a scout for an alien race reports back to a colleague about what he found when he visited Earth. He's trying to convince the other alien that there is intelligent life on Earth and--seriously--the sentient beings are made of meat. In fact, they even think with meat.

Nice! It's very funny, thanks for sharing it! :)

Cheers,

- Matt

Josta

03-09-2007, 06:46 AM

Heya Matt,

Does the universe abruptly turn blue with some indecipherable error message in the middle of it and wait to be restarted? ;)

Perhaps that's exactly what the Big Bang was...one big CTRL-ALT-DELETE! :D

The main problem is that our computers are not really designed to be "simulated brains"...

Absolutely true for most current commercial machines, but not for research systems that are currently in development, and not for the current direction in which a lot of AI work is heading.

Neural network systems, for example, are explicitly modeled on the low-level neural architecture of the brain. Current work in "embodied computing" is based on giving a robot the ability to sense the world (binocular vision, binaural sound input, etc.) and the ability to move around in it (walking) and manipulate it (reaching, grasping) in ways similar to humans. This is done so that the robot's learning programs can be sculpted and shaped by some of the same "raw material"--sensory input from the world around it--that our own brains receive. And in another rapidly developing research area, work in affective computing recognizes the key role that emotions play not just in our behavior but also our cognition, and so incorporates interfaces that can recognize and display emotional cues.

The motivation behind all this work is to design systems to be more like brains, at least in certain ways. And specifically in this way: to move beyond the idea that the way to program something human-like is to write some sort of "master algorithm" that first requires us to explicitly "write the equations" for human behavior. That sort of thinking was popular among the CS crowd back at the birth of "cybernetics" back in the 50's and for the next 20 or so years, but it's considered to be rather quaintly old-fashioned these days.

The current state of the art is rather focused on in some ways a more basic approach: first create a learning system, then give it the ability to sense and interact with the world, then let people interact with it and "let the learning begin".

In this case, no programmer is "writing the music composing algorithm". The system learns by example, learning what is "good" and "bad" music, for example, based on what musicians (not programmers) teach it. In systems like this, in fact, it's currently an open question as to whether you could, even in principle, "look under the hood" to see what the eventual "algorithm" is in there.

No one tells a child who is studying music, "Hey Billy, here's the algorithm for music." Yet of course children learn music just fine. So we don't need to know "the algorithm" to help someone or something else learn about music. We can teach by example.

Suppose the universe is based on an equation (possibly infinitely complex).

Heck, Matt, there's a guy who has gone a step farther than that: Steven Wolfram, a winner of one of the MacArthur Foundation "genius" awards, posits that the entire universe may be a sort of computer, and its purpose is to solve some ultimate "equation of everything"! WTF!?!?!? (http://query.nytimes.com/gst/fullpage.html?res=9C00E4DE1F30F932A25756C0A9649C8B 63) :rolleyes:

So... until we figure out how to give computers this "free will", they will remain incapable of true simulation of human behaviour. (emulation is very possible though!)

Coming back in a step from the farthest reaches of the cosmos (or the Twilight Zone), I am in complete agreement that the issue of free will is fundamental here. But the bigger--and more personal--question for me is actually not "How can a computing system be built to have free will like a brain?" but "How can a brain have free will?"

If a brain really is capable of free will--that is, if what we call free will is not an illusion--then it means that a complex physical thing made of molecules and cells that communicate by chemical and electrical signals can somehow create everything we experience as humans: consciousness, love, music, motivation, hunger, pain...and free will.

If that's true, if, like in Bisson's play, it's meat that is doing all of these wondrous things, then I think the claim that a human-built computing system can't loses some steam. It seems more remarkable to me that "meat" can!

If, at some fundamental level, the brain is a physical object with abilities like will and creativity, then why couldn't another physical object, with a similar functional architecture but built from different materials, also have those capabilities? (Whether we'd ever actually be able to successfully build such a thing in practice is of course an open question.)

There is of course also the possibility that it's not the brain that has free will, but some nonmaterial spirit, or essence or soul. But this raises just as many questions: how can something nonmaterial interact with the squishy pink tissue that is the brain to eventually fire the nerves that make my finger press down on D#? When during human development does the soul become "connected" to the body? Does the soul exist before the brain gets built? How does the brain communicate back to the soul?

Maybe the brain and the soul speak to each other with music. That's a thought I like!

Joe

Counterpoint

03-09-2007, 09:16 AM

Hey Joe, excellent and well-thought out reply!

Heya Matt,

Perhaps that's exactly what the Big Bang was...one big CTL-ALT-DELETE! :D

Actually, that idea rather terrifies me! :D

True for most current commercial machines, but not for research systems that are currently in development, and not for the current direction in which a lot of AI work is heading.

Yes, you are right, but I thought we were more discussing the former rather than the latter. Research/concept machines can be based on completely different models than the simple binary calculators we put in our homes and offices.

Neural network systems, for example, are explicitly modeled on the low-level neural architecture of the brain.

Can we be certain of this? Can we say beyond a doubt that the human brain works in a similar way to a neural net?

I will state that my knowledge of NN's is limited (I only had one course that dealt with them and the one we built was a very simple one) but I was under the impression this still works in a binary system. You create nodes/states that accept the data and depending on what the data is, a decision is made and the data moves on to the next state, possibly altered, as determined by some threshold.

NN's can do an effective job of solving certain types of problems and emulating behaviours, but they are still algorithmic. We can still look at the code/input/parameters and determine exactly what the machine will do next. We still don't know enough about the meat-brain (since you like this term!) to be able to predict what a human will do next.

Current work in "embodied computing" is based on giving a robot the ability to sense the world...

True, but again the machines can presently (as far as I know) only have emulated emotions and again behave in a predictable way. Can any of these machines feel sad but pretend to be happy to protect the feelings of someone else?

What about taking a human actor and having them interact with one of these machines, pretending to be happy or angry or any other emotion. Can the machine eventually figure out that this is just pretend or does it just react to each emotion predictably? If a machine can figure out that someone is pretending, this would imply that the machine can become skeptical about the world around it. So far, in the examples I have seen (various MIT videos and research papers) the machine is never able to figure out that it is being manipulated by faked emotions.

(granted though, some people are not able to figure this out either...)

The motivation behind all this work is to design systems to be more like brains, at least in certain ways...

Yes, I would agree that the "master algorithm" is the wrong approach, but computers are still just based on equations and algorithms. Can you provide an example of a machine that does not do work based on equations or algorithms? In fact, such a device could not be called a "machine"... you could call it an "invention" perhaps, but a machine by definition (in our context) is: "an apparatus consisting of interrelated parts with separate functions, used in the performance of some kind of work". This implies the use of equations and algorithms in order to get some job done. Even a NN is a type of algorithm...

The current state of the art is rather focused on in some ways a more basic approach: first create a learning system, then give it the ability to sense and interact with the world, then let people interact with it and "let the learning begin".

Yes, I would agree that this approach makes sense. A living creature is not born with all the knowledge in the world, but does have the tools to learn. Applying this to a "learning machine" should provide better results than trying to invent a machine that already behaves like a fully developed human being.

In this case, no programmer is "writing the music composing algorithm". The system learns by example, learning what is "good" and "bad" music, for example, based on what musicians (not programmers) teach it...

Does such an invention exist yet? This is making the assumption that we have a learning machine and that it is capable of understanding what musicians are telling it.

No one tells a child who is studying music, "Hey Billy, here's the algorithm for music." Yet of course children learn music just fine. So we don't need to know "the algorithm" to help someone or something else learn about music. We can teach by example.

I know what you're saying, but here's the problem... a machine needs an algorithm otherwise it can accomplish nothing. A simulated brain could possibly learn music and not need an algorithm, but we're not there yet.

Another problem you introduce here... the question of how a child learns music. This is a black box! There have been composers who were completely self-taught, but we don't have any simulated brain yet that is not taught.

Heck, Matt, there's a guy who has gone a step farther than that

Wow, that is quite an article (and kind of a step further in the opposite direction from what I was trying to say as I was actually arguing against an equation)... if I were to hazard a guess I would say this is a load of nonsense and that the Wolfram fellow is just looking for some time in the spotlight... I'm very skeptical when people make big claims and do nothing to really back them up other than providing some sketchy examples, all the while maintaining that they are "working in secret".. anybody can pretend to know everything.

To be honest, I think that trying to explain the universe as an "equation" could be dangerous. If we are just predictable machines that obey a determined algorithm, then... the first thing that will happen is that lawyers will start using this as an excuse for their clients. "Fred can not be held accountable for his actions as they have been pre-determined by this equation. Therefore we plead innocent." And the judge would pretty much have to agree since predetermined behaviour by an equation would absolve us of all responsibility...

I would agree that the physical universe is determined by equations, but my personal belief (because there is no proof yet either way) is that our actions are not. Our actions might even be influenced by these equations, but we always have the ability to make a decision.

<cue freaky guitar music>

But the bigger--and more personal--question for me is actually not "How can a computing system be built to have free will like a brain?" but "How can a brain have free will?"

I don't know if this question can even be answered! It's a good question though. :)

Maybe we just "do" and there simply is no rational explanation for it?

If a brain really is capable of free will--that is, if what we call free will is not an illusion--then it means that a complex physical thing made of molecules and cells that communicate by chemical and electrical signals can somehow create everything we experience as humans: consciousness, love, music, motivation, hunger, pain...and free will.

Well, the brain definitely has something to do with those emotions you mentioned. We do know that chemical imbalances or direct stimulation of certain areas of the brain can evoke those emotions... so pretty much all of them can be explained. What determines how we experience them without a doctor poking at our brains with a stick is the part that isn't easily explained and of course the free will can't be explained at all (in scientific terms) until we figure out what it is and where it comes from!

If, at some fundamental level, the brain is a physical object with abilities like will and creativity, then why couldn't another physical object, with a similar functional architecture but built from different materials, also have those capabilities?

I agree. :D Who knows though? Maybe the building-materials are part of the function?

There is of course also the possibility that it's not the brain that has free will, but some nonmaterial spirit, or essence or soul.

Hehe, so then we become meat-puppets... but yeah, what's pulling the strings or even attaching them in the first place?

Actually, there might be something that supports that idea:

Some friends and relatives of mine have told me about their experiences being with a someone who died right in front of them (fortunately due to natural causes, not accidental or violent). They have all told me basically the same thing and that was that they experienced a sensation of "departure" when the person died. One second, that person was there in the body, and the next moment it was just a body with no person in it. It was a very distinct sensation for them.

We can arrive at two different conclusions here... the skeptical conclusion is that they just imagined this feeling or wanted to believe in something other than an "off" switch.

Or... we can conclude that these people actually witnessed the separation of that free will from the "pink meat".

Maybe the brain and the soul speak to each other with music. That's a thought I like!

I think that they might communicate in several languages, music being one of them for sure! Again, I'd be careful to mention that this is not provable in scientific terms, but many of us believe it anyways. :)

Cheers,

- Matt

nickysnd

03-09-2007, 09:30 AM

Voice Two: "Thinking meat! You're asking me to believe in thinking meat!"

Voice One: "Yes, thinking meat! Conscious meat! Loving meat. Dreaming meat. The meat is the whole deal! ...
Voice One: [... continuing] "Also - we two do exist only because the meat imagined us having this conversation."

Voice Two: [Silence - suddenly, Voice Two is not there anymore.]

Steven Wolfram, a winner of one of the MacArthur Foundation "genius" awards, posits that the entire universe may be a sort of computer, and its purpose is to solve some ultimate "equation of everything"! WTF!?!?!? (http://query.nytimes.com/gst/fullpage.html?res=9C00E4DE1F30F932A25756C0A9649C8B 63) :rolleyes:
"The answer to the ultimate question of life, the universe, and everything is ... : 42."
(said the supercomputer Deep Thought after 7.5 million years of computing)

"Forty-two!" yelled Loonquawl. "Is that all you've got to show for seven and a half million years' work?"
"I checked it very thoroughly," said the computer, "and that quite definitely is the answer. I think the problem, to be quite honest with you, is that you've never actually known what the question is."
http://everything2.com/index.pl?node=The%20Hitchhiker%27s%20Guide%20to%20 the%20Galaxy
Edit -
Here is what I got while trying to submit the above post:
The following errors occurred when this message was submitted:

The message you have entered is too short. Please lengthen your message to at least 1 characters.
No, computers are not intelligent, they just think they are.

shnurgle

03-09-2007, 11:49 AM

Counterpoint

03-09-2007, 12:44 PM

Lixir

03-09-2007, 01:41 PM

Great topic !

I'm probably wrong, but I think it's possible, I can't argu much all I can say is , composition is related to math, so basically computers are able to compose music depending on each combinations etc.
BUT performance is something else, even if computers are able to create the best music ever, nothing will ever be equal to an orchestra performance, every little imperfections, attacks, moods, the way each players are perfoming,the conductor having his own view of the piece, is just impossible for a computer .
I understood that when I recorded a friend cellist, I was like "arf can you do that attack again in order to be absolutely perfect ?" she answered "yes but you have to keep some few "mistakes"
because that's what gives life to the performance, and now I don't regret because the cello part was fantastic and very natural, I doubt a computer would do that even with specific programmation.

That's my opinion, and again I may be wrong.

:)

nickysnd

03-09-2007, 01:44 PM

Technically, humans are to blame there. We designed this website to not allow a post with only quotes..we decided that there must be new text typed in. Technically, the computer performed it's task flawlessly, we're the only ones to blame for making that a rule here.
Precisely the point! Applying rules and performing tasks flawlesly - this is as far as can be from creating music. "My dear R2D2, can you please technically perform the task of flawlesly composing an underscore for that short movie while I'm out for lunch? - ya know, come up with something that that fussy director would like... Oh, nevermind lunch, take your time and try your best during the next couple of centuries. I'll deal with that short when I'm back."

Besides, if that decision whether or not to pass that message on was given to a person, he/she would most likely have let it pass, for he/she probably would have observed that some part the "quotes" are not exactly quotes. Using the rules against themselves for comic effects, or just for the hell of it - now tell a computer that! Computers and industrial robots are just mechanisms dumbly performing tasks. There can be no rules/algorithms for imagination and artistic creativity.

shnurgle

03-09-2007, 03:43 PM

I hear ya Nicky. But again, (and it has been some time since I looked at this thread so i'm paraphrasing) I think the point was that that we simply don't know what the limits of technology will be in the future. As of right now, of course a computer can't freely compose a piece of music as cleverly and as in tune with human aesthetic and sensability as a human can. And truthfully, I don't really believe (or at least i hope not) that a computer will ever be able to do it on a truly human level. But the truth is, we just don't know. We can only speak to the algorithms and technology that currently exist, in terms of the future, we can only guess. If 500 years ago you told someone that one day a device called a camera would be able to take a flawless imprint of a moment in time they would have said you had lobsters crawling out of your ears. If you had told them that one day we would put navigation systems in ships that were relied on implicitly to point them in the right direction, or even that there would be huge metal devices called airplanes that not only could fly, but could actually take off, steer, and land completely on their own, they'd have said, "Woooord?"

So, you just never know what the future holds for technology, you can olny guess based on what you know now, and you personal opinions.

Josta

03-09-2007, 04:57 PM

I think the point was that that we simply don't know what the limits of technology will be in the future.

The point is that, but the point actually goes way beyond that. In various posts in this thread I've given some pointers to actual working examples of various current built-by-humans systems that, even today in a field that is still in its absolute infancy, do demonstrate some of the capabilities some posters here keep saying are impossible. Like "machines are always predictable", "machines can not deal with emotions", "Mind-related issues are un-explainable. Period." "there can be no algorithms for creativity".

Each of these statements is demonstrably incorrect today for specific, and I freely grant limited examples. But as I've said before, there is a BIG difference between difficult and impossible. And history teaches us that once a dogmatic assertion is cracked even slightly, a lot of invention can push up through those cracks into the light.

Now, does this somehow guarantee that there will be an emotion-feeling, blindingly creative, self-motivated, free-willing android John Williams writing the soundtracks for an android Steve Spielberg in 2307? Of course not. But it's interesting that the same people who claim scientists have "binary minds" are the ones who also seem to think they can proclaim that something can or can't be done, forever and ever, Amen. Yes/no. On/off. No room for the sort of pastel-hued in-between that says, with appropriate qualification, "I don't know for sure. Perhaps."

The next level of response tends to be to just ignore inconvenient counter-evidence and move on to a different moving target, "OK, so a genetic algorithm can generate new and unexpectedly creative movements for an animated character, movements that surprised even the programmer and even generated a positive emotional response in an audience. But can you make a robotic squirrel, huh?"

It's very much like the examples you cited, shnurg. The airplane one is especially apt: in the 1800's there were many people who, based on dogmatic belief in their own supposed ability to say what could and couldn't be done, confidently asserted again and again that building a heavier than air flying machine was impossible. Then, when the Wright brothers made their first, wobbly, 12-second flight in 1903, taking one person all of 120 feet and soaring to the grand altitude of 12 feet, the same people scoffed and said, "OK, whatever, but you're telling me that thing is going to be able to fly 300 people from Los Angeles to Kyoto? At a height of 7 miles? Nonstop? Ridiculous! Impossible!"

To which Wilbur Wright just sighs and says, "No, not this thing. But with the knowledge from this thing, we'll build one that will go a mile. Then 20. Then over a lake. Then carry 50 people. And then, in 60 years, LA to Kyoto."

To which the confident naysayers replied, "Whatever. But you'll never go to the moon. Idiotic. Now that's impossible."

:rolleyes:

Joe

Counterpoint

03-09-2007, 07:12 PM

It's very much like the examples you cited, shnurg. The airplane one is especially apt: in the 1800's there were many people who, based on dogmatic belief in their own supposed ability to say what could and couldn't be done, confidently asserted again and again that building a heavier than air flying machine was impossible. Then, when the Wright brothers made their first, wobbly, 12-second flight in 1903, taking one person all of 120 feet and soaring to the grand altitude of 12 feet, the same people scoffed and said, "OK, whatever, but you're telling me that thing is going to be able to fly 300 people from Los Angeles to Kyoto? At a height of 7 miles? Nonstop? Ridiculous! Impossible!"

These are completely different problems. When the aircraft was invented it was based on existing ideas (there have been manned gliders for more than 1000 years). By refining a proven model they were able to overcome the limitations such as the number of people that could be transported at a given time and the distance that can be travelled. Further refinement and development of *proven theories* put us on the moon and can potentially take us further.

With the development of simulated brains, we are not nearly at the stage the Wright brothers were at when they designed their flying machine. Not even close...! We can EMULATE human behaviour (like many of the examples given) but none of these computers, machines, or programs possess the ability to think freely...! We can't even determine what causes "free will" yet. We don't have even a simple working model! At least the wright brothers could make paper airplanes to prove the concept. The current emulations of human behaviour are not proof that we can create a free-thinking machine... so until a proven model exists for simulating "free will" we simply can't say whether or not it is possible.

This doesn't mean we should stop trying, this doesn't mean we have to give up on emulating human behaviour (which has many practical and applied uses in the real world) and this doesn't mean that it is impossible. All it means is that we can't prove it until we figure out that first simple model.

I'll become a full believer when this happens:

Inventor: "SimBrain, compose me a Mozart sonata!"

SimBrain: "No. I don't want to. Go climb a tree you over-evolved monkey."

...

Anyways... I'm tired and I don't even know if I'm making an ounce of sense anymore...

Goodnight,

- Matt

nickysnd

03-09-2007, 08:27 PM

there is a BIG difference between difficult and impossible.
Agreed on that - difficult/easy is another discussion. Here we are discussing possible/impossible, and I mean possible/impossible now, today.
it's interesting that the same people who claim scientists have "binary minds" are the ones who also seem to think they can proclaim that something can or can't be done, forever and ever, Amen. Yes/no. On/off. No room for the sort of pastel-hued in-between that says, with appropriate qualification, "I don't know for sure. Perhaps."
Of course no one knows anything for sure, it goes without saying. Also, I am sorry if I ever gave the impression that I have the slightest idea of what might happen in the future. In all I said I always meant the present. Future? Anyone can say anything about future, so let's skip it for now. In exchange, I would like to bring up two terms that appear to me essential for this duscussion:
Algorithm: (from http://dictionary.reference.com/)
A set of rules for solving a problem in a finite number of steps, as for finding the greatest common divisor.

A step-by-step problem-solving procedure, especially an established, recursive computational procedure for solving a problem in a finite number of steps.
Continuum:
A continuous extent, succession, or whole, no part of which can be distinguished from neighboring parts except by arbitrary division.

A set of elements such that between any two of them there is a third element.
Now, presuming that we have agreed upon semantics and presumptions, I'd have three questions:

Is music a continuum, or a bag of elements that can be observed/defined/isolated?
(To me, music is a meaningful/colorful expression of time. So, it's a continuum, like time itself.)

Is artistic imagination/creativity a continuum, or a set of steps that can be observed/defined/isolated?
(To me, it is a continuum, but I can speak only of my own way to imagine/create music.)

Is it now, presently, possible to build an algorithm that would generate and control a continuum?
(To me, it seems impossible to make an algorithm out of something else than finite/isolated elements.)

If we agree that it is impossible to build, now, today, that algorithm - then we have said all that we could say about the issue: a program cannot compose music. It seems to me that, in order to continue this debate, someone has to prove one of the two:
either -
A. Music and artistic imagination/creativity are not continuums but bags containing finite numbers of isolated elements,
or -
B. Music and imagination are continuums alright, but it is now possible to build algorithms using something else than finite/isolated elements, so we can now build algorithms able to generate and control those continuums.

Now, again, my conclusion, so far:
For the above reasons, presently, today, it is impossible to build an algorithm that would enable an object to create music. And, for the same reasons - now, today, just as yesterday and as always was, there are/were no rules/algorithms for imagination and artistic creativity. Those algorithms might become possible tomorrow, who knows? - but then again, tomorrow may rain so I'll follow the sun.

Software

03-09-2007, 09:38 PM

BUT performance is something else, even if computers are able to create the best music ever, nothing will ever be equal to an orchestra performance, every little imperfections, attacks, moods, the way each players are perfoming,the conductor having his own view of the piece, is just impossible for a computer .

Agreed. But also the audience has its part in the process. Playing in a concert and playing in a concert are mentally quite diffferent happenings.

Software

03-09-2007, 09:56 PM

We can EMULATE human behaviour (like many of the examples given) but none of these computers, machines, or programs possess the ability to think freely...! We can't even determine what causes "free will" yet. We don't have even a simple working model! At least the wright brothers could make paper airplanes to prove the concept. The current emulations of human behaviour are not proof that we can create a free-thinking machine... so until a proven model exists for simulating "free will" we simply can't say whether or not it is possible.

Human composers are mostly working with constraints and there is not much space for thinking "freely" (notation, instruments, room, audience and all that on lower level are all big constraints). As said before human composers don't catch each not from air, but have always a set of algorithms even if those algotrihms may be implicite and are not called as algorithms. If the composer has his own STYLE, the algorithms are the same for evey piece. So where is the "free will"?

What is the difference when discussing about computers as composers? Is it good if computer has "style". Is it good if the "style" may change from piece to piece? If the user clicks the "style button", who is the composer? If know one can guess what comes out, who is the composer?

nikolas

03-10-2007, 12:27 AM

Software is REALLY REALLY right at this point!

I am under the impression that only 1% of the well known composers (classical or not), could excape their own learnings and algorithms... Mendelshon or Vivaldi are certainly not the case (Teasing Nicky here, but avoided putting "gods" in here :D)...

Thing is that there might be a spark of .1% in human creativity that makes some difference but cannot be sure.

I will go back to my previous posts to stress on something:

I have decided to music is for communication, even if random, even if without communicating feelings. The series of notes, of rythms etc have 2 meanings always. At least to bigger works. Even Boulezes thick serialism has a meaning behind that which can be translated from the composer to the performer and from the performer to the audience. My problem simply lies that all emulations will be lacking that "need to say something to my audience" algorithm, which cannot (at the moment of course) be, even emulated.

All efforts I've heard at synesthesia are simply wonderful, but fail to "tell" me that additional something (which was what I was talking about automat). It is this "hidden" "message" that I'm receiving from most music that I hear over here (even simplistic, or silly or badly produced sometimes), and I feel confident that someone is trying to "send" me something. Music which is not trying to send anything, even if written by a human, could simply be a music by bots. (<-without trying to insult any bot/human/member)

:)

I think, that as most things, it will take some time to get used to the idea, to surely (for me) will come in the future:

Would a 19th century person ever consider having an ipod and listening to the Vienna Philharmonic? Nope! Impossible!
Would they accept the notion? No because the quality would be awful (as was in the early 20th century!)
Would they even consider iMAX theaters and virtal shows? HELL NO!

It's not about creativity in the above case, but still makes some connection to thing to come in the future.

Now carefully examining creativity, could bring some confusion, but this thread was not about creativity bu about composing music, which are different. I as a composer also cook! :D And clean! And have sex (occasionally :D). A software made to create music, will never have these things. So we are already talking about limited creativity. Cause at the end, an orgasm, may give me a great idea for a love song to my wife! Machine will miss that!

Software: I was wondering: Have you done any experiments, the simplest kinds. Create 5 tracks from 5 different paintings and have the audience macth them up? It would be highly interesting to see if indeed there is the image input that you expect (which is obviously there algorithmicaly), or it is simply not evident! :)

Software

03-10-2007, 03:00 AM

All efforts I've heard at synesthesia are simply wonderful, but fail to "tell" me that additional something (which was what I was talking about automat). It is this "hidden" "message" that I'm receiving from most music that I hear over here (even simplistic, or silly or badly produced sometimes), and I feel confident that someone is trying to "send" me something. Music which is not trying to send anything, even if written by a human, could simply be a music by bots. (<-without trying to insult any bot/human/member)
When listening music there are most often the interpreters of the music, musicians (instumentalists). I think it is just those people who communicate the music. An most often the messages may be different when different people are playing.

Does the score "communicate". I wouldn't be so sure. The problem of musicology is just that those people just read the score. Or interprete it. Base on cultural learning (just like the musicians). I myself believe in EXPERIMENTAL MUSICOLOGY. The we have the score, its interpreters and the audience. Without experimenting with audience one should not speak about communication. Musicians are biased. Music is NOT written for musicians, but for audience. Maybe also the educated audience is biased. The audience has expectations concerning the message. *What if the audience has not learned the new language, yet. The history of music is full of that kind of examples, as all arts, too.

Now carefully examining creativity, could bring some confusion, but this thread was not about creativity bu about composing music, which are different. I as a composer also cook! :D And clean! And have sex (occasionally :D). A software made to create music, will never have these things. So we are already talking about limited creativity. Cause at the end, an orgasm, may give me a great idea for a love song to my wife! Machine will miss that!

When you have got "a great idea", you know it. If the computer has great ideas, we know it ONLY after listening. So What?

Software: I was wondering: Have you done any experiments, the simplest kinds. Create 5 tracks from 5 different paintings and have the audience macth them up? It would be highly interesting to see if indeed there is the image input that you expect (which is obviously there algorithmicaly), or it is simply not evident! :)

That is theoretically impossible. Music doesn't have semantics. So what is there to "match"? Painting create emotions. Music creates emotion. Can we match emotions between different paintings? I don't think so.

Josta

03-10-2007, 06:53 AM

These are completely different problems...With the development of simulated brains, we are not nearly at the stage the Wright brothers were at when they designed their flying machine. Not even close...!

:) lol, Matt. The point of my allegory was not that planes are like brains, but to point out that there are many historical examples of people saying "X will never be possible" even as other people are taking the first baby steps toward "X".

It's also easy to look at a previous achievement with 20-20 hindsight and see that, for example, the Wright brothers were close to success. There were indeed many people in the 1800's who believed "If God had intended man to fly, He'd have given us wings." And there were people who believed for theoretical reasons that powered flight would not be achieved because of supposed limits on the power-to-weight ratio of engines.

Further refinement and development of *proven theories* put us on the moon and can potentially take us further.

Again, we can say this now, based on hindsight. But it can take quite a while for even a valid theory to be regarded as "proven", especially if it's socially, religiously or personally controversial. ("You can't tell me I'm related to a monkey!") And technically, in science, all theories are subject to disproof or at least modification at any time should new, compelling evidence arrive. That being said, there are researchers in computational neuroscience, visual neurophysiology, AI, psychopharmacology, developmental biology and even philosophy and neurology who regard that the theory "the brain and its functions can be (or will eventually be) completely described mechanistically" has been verified beyond serious dispute. (For the record, I don't share this view.)

My favorite example of the uphill battle even a valid theory can face is a scathing editorial written by the New York Times in 1920, which ridiculed Robert Goddard for proposing that his rockets, until then only tested in air, could work in the vacuum of space:

Not long before, Goddard, a physics professor at Clark University in Worcester, Mass., had published an arid little paper on an outrageous topic, rocket travel. Unlike most of his colleagues, Goddard believed rocketry was a viable technology, and his paper, primly titled "A Method of Reaching Extreme Altitudes," was designed to prove it. For the lay reader, there wasn't much in the writing to excite interest, but at the end, the buttoned-up professor unbuttoned a bit. If you used his technology to build a rocket big enough, he argued, and if you primed it with fuel that was powerful enough, you just might be able to reach the moon with it.

Goddard meant his moon musings to be innocent enough, but when the Times saw them, it pounced. As anyone knew, the paper explained with an editorial eye roll, space travel was impossible, since without atmosphere to push against, a rocket could not move so much as an inch. Professor Goddard, it was clear, lacked "the knowledge ladled out daily in high schools."

Goddard seethed. It wasn't just that the editors got the science all wrong. It wasn't just that they didn't care for his work. It was that they had made him out a fool. Say what you will about a scientist's research, but take care when you defame the scientist. On that day, Goddard — who would ultimately be hailed as the father of modern rocketry — sank into a quarter-century sulk from which he never fully emerged. And from that sulk came some of the most incandescent achievements of his age.

From The TIME 100: Robert Goddard (http://www.time.com/time/time100/scientist/profile/goddard.html)

One of the similarities to the present thread is the occasional tendency of those with little expertise in a technically challenging field to not just critique the theories, but ridicule the researchers. In the 1920 NYT article, titled "A Severe Strain on Credulity", the editors sneered:

To claim that [Goddard's rocket could fly to the moon] is to deny a fundamental law of dynamics, and only Dr. Einstein and his chosen dozen, so few and fit, are licensed to do that.

That Professor Goddard, with his "chair" in Clark University in Worcester, Massachusetts, and the countenancing of the Smithsonian Institution, does not know the relation of action to reaction, and of the need to have something better than a vacuum against which to react – to say that would be absurd. Of course he only seems to lack the knowledge ladled out daily in high schools.

49 years later, a few days before Apollo 11 landed on the moon, the Times finally published a "retraction":

A Correction. On Jan. 13, 1920, "Topics of the Times," an editorial-page feature of the The New York Times, dismissed the notion that a rocket could function in vacuum and commented on the ideas of Robert H. Goddard, the rocket pioneer...

...it is now definitely established that a rocket can function in a vacuum as well as in an atmosphere. The Times regrets the error.

:rolleyes:

"Every vision is a joke until the first man [or woman] accomplishes it; once realized, it becomes commonplace." Robert Goddard, responding to reporters in 1920

Joe

P.S. I still have no idea if an emotion-feeling, blindingly creative, self-motivated, free-willing android John Williams could exist and win the CyberOscar for best original score in 2308. But I have respect, not ridicule, for the pioneers who are testing early theories and making demonstrated progress in machine intelligence. Because the results of their work might also have huge benefits to people in domains other than music, such as restoring function to people with strokes, sensing if a driver has fallen asleep and waking him up before he crashes (or steering the car safely to the side of the raod), aiding a doctor with a difficult diagnosis, or providing assistance to the elderly or disabled.

Josta

03-10-2007, 07:45 AM

BTW, Matt and Nicky, you've both posted some additional questions that I find really interesting and I plan to volley some ideas and info back your way, but my fiancée has been patiently sitting here watching me type and, for the moment, I need to "freely will" to head out the door with her right now...

:)

TBC...

Joe

Counterpoint

03-14-2007, 01:40 AM

:) lol, Matt. The point of my allegory was not that planes are like brains, but to point out that there are many historical examples of people saying "X will never be possible" even as other people are taking the first baby steps toward "X".

Hey Joe! Yeh, I realize that you were not comparing the planes to the brains, but what I was saying is that the allegory didn't work because the Wright brothers (despite the beliefs of others) could build a miniature working model based on all the same fundamental rules. Even a simple airplane model powered by elastic band could be used to prove the concept.

Even the best neural nets are not proven to be anything other than algorithmic (although many believe they are the closest we have gotten to a real AI) so they don't really work as a "miniature model". Even the best "AI" of today can't be demonstrated to accurately model even a basic brain. If you're aware of ones that prove me wrong, then please do share them and prove me wrong!! It's better to be wrong than to miss out on some really cool discoveries. :D

It's also easy to look at a previous achievement with 20-20 hindsight and see that, for example, the Wright brothers were close to success. There were indeed many people in the 1800's who believed "If God had intended man to fly, He'd have given us wings." And there were people who believed for theoretical reasons that powered flight would not be achieved because of supposed limits on the power-to-weight ratio of engines.

Yes, and I'm sure the Wright brothers did have to put up with more than their fair share of naysayers. Yes, their power-to-weight ratio was a big hurdle that caused others to fail before them. However, even the more efficient engines are based on the same principle, they just found better ways to build them that yielded enough power to fill the requirements.

Even the religious notion you mentioned can be argued though... it would have been tempting to remind those people that according to their favourite book, God gave us free will and therefore what was intended is that we think for ourselves and invent whatever pops into our heads.

Again, we can say this now, based on hindsight. But it can take quite a while for even a valid theory to be regarded as "proven", especially if it's socially, religiously or personally controversial. ("You can't tell me I'm related to a monkey!")

Yeh, I have some very, very religious friends (of the creationist mindset) with whom I argue about various things and despite there being evidence to the contrary, they still insist on believing in nonsensical things... take a riverbed with millions of layers of sediment deposits (one deposit per turn of the seasons). You could logically extrapolate that this means the river was there for millions of years, but the creationists will say this is a "lie created by the devil to lure us away from God who created the world 6000 years ago".

Sometimes it makes me wish my ancestors had never left their tree!!

And technically, in science, all theories are subject to disproof or at least modification at any time should new, compelling evidence arrive. That being said, there are researchers in computational neuroscience, visual neurophysiology, AI, psychopharmacology, developmental biology and even philosophy and neurology who regard that the theory "the brain and its functions can be (or will eventually be) completely described mechanistically" has been verified beyond serious dispute. (For the record, I don't share this view.)

Yeh, well I'm not exactly in the know about what will be discovered tomorrow... but I'm pretty sure that today they still havn't figured it all out yet.

My favorite example of the uphill battle even a valid theory can face is a scathing editorial written by the New York Times in 1920, which ridiculed Robert Goddard for proposing that his rockets, until then only tested in air, could work in the vacuum of space:

They apparently didn't understand the concept of action-reaction at all (funny since that one fellow actually mentioned it too)! Einstein could have explained it to them. :D

What baffles me is how they worked out that space was a vacuum in the first place (not having been there before)? I'll have to go look that up!

Cheers,

- Matt

Counterpoint

03-14-2007, 02:09 AM

Human composers are mostly working with constraints and there is not much space for thinking "freely" (notation, instruments, room, audience and all that on lower level are all big constraints). As said before human composers don't catch each not from air, but have always a set of algorithms even if those algotrihms may be implicite and are not called as algorithms. If the composer has his own STYLE, the algorithms are the same for evey piece. So where is the "free will"?

Well... do we really have constraints? New instruments, new methods, new scales, new notations... these are being invented all the time by creative people. The ability to break all the rules or even write new rules is our "free will". In terms of composing, we have no constraints. When we consider "pleasing an audience", we start to see stronger constraints.

What is the difference when discussing about computers as composers? Is it good if computer has "style". Is it good if the "style" may change from piece to piece? If the user clicks the "style button", who is the composer? If know one can guess what comes out, who is the composer?

The thing is, that with computer algorithms we can predict what comes out as long as we know what the algorithm does. We might not be able to do all the computations as fast as the machine can, but given time you could predict every result that comes out of the machine.

Of course the user is the composer, because they are the ones providing the input and telling the computer to call the "output=make_music(input);" function. When the composer has an understanding of the process and the ability to determine good results from bad results, the music created by this process can be compelling and interesting.

A computer does not freely "decide" to make music. In the present day this is always a user or programmer's decision.

Cheers,

- Matt

Josta

03-14-2007, 08:32 PM

Heya Matt,

Awesome thread...soooooo much to muse (and amuse) about!

Even the best neural nets are not proven to be anything other than algorithmic (although many believe they are the closest we have gotten to a real AI) so they don't really work as a "miniature model". Even the best "AI" of today can't be demonstrated to accurately model even a basic brain. If you're aware of ones that prove me wrong, then please do share them and prove me wrong!! It's better to be wrong than to miss out on some really cool discoveries. :D

If by "neural nets" you mean the sort of 3-layer back-propagation networks developed 20-30 years ago, yes, absolutely, these are just vague analogies to actual nerve networks.

The most "biomimetic" neural nets currently however are very closely modeled on actual neural circuits. There is very little in common between brain circuits and conventional computer circuits--but there can be a quite close correspondence between a computer program model of neurons and the neurons' actual behavior. (Check out the software NEURON, for just one example.)

Here are a few other observations:

First, there are algorithms running around all over the brain. In fact, there's a whole field of science devoted to figuring them out called computational neuroscience. This isn't just theory, it's based on direct observation. You look at the way neurons are interconnected, you look at how they are organized in "circuits", and you can determine how patterns of information are transformed. The result is that groups of neurons, the basic stuff of the brain, implement algorithms that can be directly correlated with their function.

This obviously has a potentially huge bearing on the basic question that started this thread, but the many questions are far, far from being answered--and pretty much any researcher in neuroscience will tell you that we are still very much in the infancy of this research.

That being said, there are perhaps hundreds of algorithms that have so far been observed in brain areas ranging from vision and movement control to posture, hearing and memory. But our understanding of these algorithms so far is either very "zoomed in" or very "wide angle": we can see the "microcircuitry" in great detail, and we can see large patterns of activity by MRI scans and brain-wave measurement, but the really interesting parts, the middle-level neural processes that (possibly, maybe, maybe not) create the "man in the middle" , are currently the subject of huge amounts of study.

But we do know, for example, that there are algorithms running almost everywhere that people have looked in the brain.

One of the earliest discoveries was the way our retinas are wired up with a built-in algorithm to increase the contrast of the images we see. The neurons that connect the sensory cells in the eye to the visual cortex and are the first step in processing visual information have been examined in excruciating detail. They are wired together in such a way that they implement an algorithm known as "center-surround inhibition". If you've ever used the "sharpen edges" filter in Photoshop, it's implementing a very similar algorithm--in fact, in this case, the computer science folks discovered the edge-sharpening algorithm in part by inspiration from the way the neurons in the retina worked. They came up with the "Marr-Hildreth Operator" (developed by Ellen Hildreth (http://cs.wellesley.edu/~cs/ehildreth.html) when she was a grad student) but it's really sort of the "Back of the Eyeball Operator."

Since then, all sorts of algorithms have been discovered buzzing around in the visual cortex of the brain. Algorithms to recognize lines and shapes, algorithms to detect orientation, etc.

In fact, in at least some cases already, higher-level behavior can be explained by the lower-level algorithms. Why we "see" certain optical illusions, for example, can be explained by how the lower-level visual algorithms operate.

Movement control algorithms have also been experimentally teased out of numerous sites in the brain. The cerebellum, a sort of corrugated blob that sits back and low in the brain, has all manner of algorithms running in it that coordinate motion.

Over a decade ago, some researchers studied the way nerve cells are wired up in a region of the brain called the hippocampus that is involved in learning and memory. The researchers translated the connections between the nerves and the way they processed information into an analogous computer algorithm. They then found that this algorithm actually performed better than some existing computer science algorithms at classifying complex signals (http://www.ics.uci.edu/~granger/Applications.html).

The thing is, that with computer algorithms we can predict what comes out as long as we know what the algorithm does. We might not be able to do all the computations as fast as the machine can, but given time you could predict every result that comes out of the machine.

This isn't really true. There are trivial algorithms--cellular automata, for example--for which, given an input, you cannot "predict" the output, even in theory--you have to actually execute the program to determine what the output will be. In which case, determining the output is not really "prediction" in any conventional sense--you just have to do it (i.e., run the program) to get the result. And if you just "do it on paper", you still need to execute the algorithm so there aren't any shortcuts. In this sense cellular automata are not predictable, but they are deterministic.

On the other hand, genetic algorithms based on random variation are both unpredictable and (approximately) nondeterministic. If you give a genetic algorithm a goal and say "figure out a way to get to the goal," the random variation that is used to generate candidate solutions results in unpredictable--and often even surprising--outputs. That's the point: genetic algorithms are used to explore solution spaces where you want to generate new and unexpected approaches.

(I previously described an example of my experience of the surprising and amusing results of a genetic algorithm here (http://www.soundsonline-forums.com/showpost.php?p=68871&postcount=137).)

Also, an algorithm may exhibit chaotic behavior: very small changes in the inputs can result in huge changes in the outputs. In this case, although the output may be theoretically predictable, in practice it cannot be because we can't measure the inputs with a high enough accuracy.

Also, any algorithm that is both processing and being changed by (that is, learning from) interaction with "real world" input data will be unpredictable since the "real world" is pretty much unpredictable.

Now, none of the above is anywhere near the level of being able to provide anything like an "algorithm for creativity". Or compassion. Or honor. Or infatuation. But the results so far are convincing that many building-block neural circuits can be described algorithmically. It will be interesting to see what gets built with human-made versions of those blocks in the future.

Joe

Josta

03-14-2007, 09:42 PM

Is it now, presently, possible to build an algorithm that would generate and control a continuum? (To me, it seems impossible to make an algorithm out of something else than finite/isolated elements.)

Nicky, that was a fascinating set of ideas, and a thoughtful and interesting way of looking at the problem.

I would say in response that yes, it's actually very possible to build an algorithm that can generate and control a continuum. (Please note that this doesn't mean I'm saying it's possible now--or necessarily will be in the future--to build an "entity based on algorithms" that can work as expressively as a human in the "musical continuum.") But here are some basic examples:

A very trivial example is a digital-to-analog converter. Separate steps in, smooth continuous analog waveform out. If the input to the DAC is a stochastic function (algorithm), then the result is the generation of an unpredictable, continuously varying output signal.

Or consider a somewhat more sophisticated system: an airplane's autopilot, a "pilot algorithm". The position of the plane in space is a continuum--the plane does not "jump" from point A to point B to point C--yet the autopilot can smoothly control the course of the plane, even to the extent that, today, a plane can take off, fly hundreds of miles and land smoothly and safely in the dark or in fog--all under the control of an algorithm (actually, a set of many algorithms that both sense the outside world and communicate with each other).

Or, since Matt previously raised a counterpointed eyebrow at my plane analogies, consider a "closer to home" example: how a human body is built. I would say that bodies, in all their diversity, represent a continuum of possibilities. But there are a finite number of genes in the human genome. In fact, one of the amazing results of the Human Genome Project was the discovery that there are only about 30,000 genes that, collectively, provide the "recipe" used to construct a human body. And each gene is a sequence of very specific instructions, encoded in the four-letter alphabet of DNA.

Starting from DNA, there is a very orderly, algorithmic process that translates and transcribes the finite instruction set of the genome into the proteins that the body is made of, and that regulate and control growth and function.

In fact, when the instructions in a gene are being "read" to generate a bit of a cell, the process starts at something called the "start codon" and stops at something called the--you guessed it--"stop codon". It's an orderly, finite process: Start reading the recipe here, translate three letters at a time of the DNA (or RNA) "alphabet" to one of the 20 letters of the "amino acid alphabet", string them together, stop there. Lather, rinse, repeat.

It's a finite, orderly process, with well-defined inputs and a limited number of discrete building blocks. But, wow, what diversity of results!

You take those 30,000 relatively simple, orderly, finite genes--they're just little chemical strings, after all--and let them interact in complex ways, both with each other and with the environment, and good lord the variety that happens.

Similarly, imagine 30,000 computer algorithms interacting with each other in a complicated way and also interacting with and being changed by the external environment. The result could be a range of possibilities indistinguishable from a continuum.

And that's a point, too. Using only our human senses, how do you know (or prove) that something is a continuum and not just a really fine-grained divided substance?

For example, we perceive air to be a continuous fluid, but it's actually widely separated individual molecules. Heck, all matter is made of individual molecules! So at a very basic level, is anything really "continuous"? ;)

A more practical example: if an ink-jet printer puts a large enough number of small, separate color spots on a piece of paper, it looks just like a continuous-tone photograph. (Which itself isn't actually continuous because it, too, is made of very small individual grains of silver or colored dye.)

Sound--and music--feels to us as being a "continuous sensation", but at a teeny-tiny physical level sound is lots of individual air molecules striking our eardrums like so many little ping pong balls.

So, to recap:

Algorithms can generate and control continuous events.

Even our bodies are built from a finite set of building-block elements--yet an incredible continuum of diversity is produced.

With a big enough set of interacting building blocks, the range of possibilities becomes so vast that it may look no different from a true continuum.

At a molecular level, everything is lumpy--maybe there are no true continuums. :)

Continuously enjoying this discussion,

Joe

P.S. There's also an interesting philosophical question as to whether we could even in principle sense the difference between a true continuum and a very fine-grained "discontinuum", because our senses have limited spatial and temporal resolution. Even our sense organs are, at some level, "discrete" rather than continuous. For example, there are a finite number of rod and cone receptor cells in our retinas, arranged in a geometric, hexagonal grid. There are a finite number of nerve cells in the auditory nerve that goes from the ear to the brain, and each nerve "fires" in a discontinuous way: with "spikes", not smooth signals. Even touch and taste are not continuous: we have individual little pressure sensing cells in our skin (a few hundred per fingertip) and individual taste buds.

Somehow, all these little individual "windows" on our world come together to offer us a perceptual experience that is amazingly smooth, fluid and rich. A beautiful gradient of colors at twilight as the ocean breathes its song as you stretch out on a warm beach and taste that Corona...

Counterpoint

03-17-2007, 02:07 PM

Hi Joe,

Firstly, thanks for the huge reply, I've been reading it over and over and looking up some of the things you mentioned. Good to keep up to date with this as I do have a strong interest in the development of AI.

Secondly, I have to apologize because I only have a short time to write a reply so I can only address a couple things from your above post (all of which deserves attention!):

This isn't really true. There are trivial algorithms--cellular automata, for example--for which, given an input, you cannot "predict" the output, even in theory--you have to actually execute the program to determine what the output will be. In which case, determining the output is not really "prediction" in any conventional sense--you just have to do it (i.e., run the program) to get the result. And if you just "do it on paper", you still need to execute the algorithm so there aren't any shortcuts. In this sense cellular automata are not predictable, but they are deterministic.

I'd disagree with the cellular automata example.. the thing is if you apply the same rules and the same initial state, you will get the same result. Also, this algorithm is time/iteration based with, as you say, each iteration depending on the previous one. I guess in this sense you are right that you can't, for example, arbitrarily predict what the results will look like on a non-consecutive state. However, if we know what state N is, then we can predict state N+1 quite easily. I might have chosen the wrong word when I said "predict"... what I really meant was that given an input and an algorithm, you can reproduce the same results as the machine. It is in this sense "predictable" because everything can be explained and there is nothing that will "suprise" us.

Yes, we can show that a simple algorithm can generate complex results (like the cellular automata) but that doesn't make them unpredictable (state-to-state).

On the other hand, genetic algorithms based on random variation are both unpredictable and (approximately) nondeterministic. If you give a genetic algorithm a goal and say "figure out a way to get to the goal," the random variation that is used to generate candidate solutions results in unpredictable--and often even surprising--outputs. That's the point: genetic algorithms are used to explore solution spaces where you want to generate new and unexpected approaches.

Yes, but there are a couple problems here... firstly, it is still an algorithm and you could sit down and get the same (and possibly more random) results by tossing dice. The only advantage the machine has is that it can execute the algorithm much faster than a human being and turn around 1000's of different evolutions in a fraction of the time. While the results might be unexpected, this still doesn't equate to "intelligent decision making" which is not random (as far as we know). It's actually a rather ugly, brute-force approach (yet still rather cool) to the problem because it will produce many results that are completely unusable as well.

One example is a circuit that was designed by an engineer. They ran it through an evolution algorithm and got a very strange result. The circuit had been optimized down to far fewer components but a portion of it was not physically connected to the rest of the circuit. When removed, the circuit stopped producing the desired results. The engineer was not able to figure out how this new circuit worked.

The danger with this circuit is just that last fact. If we can't explain how it works, then how can it be trusted to always do its job? (and in fact they discovered that if they changed the room temperature, this output woud become unreliable -- they were going to attempt to run the genetic algorithm on it again, this time altering the temperature in the hope that it would adapt and modify itself to compensate)

While this example was very interesting (I would link to it but it appears the site I read it on is "locked" to the general public...!!! :mad: ) and could eventually yield great results, it is just not there yet..!

Ah... sorry out of time!

- Matt

HJS

03-29-2007, 09:09 PM

Listen to these, especially the Mahler
http://arts.ucsc.edu/faculty/cope/mp3page.htm

nikolas

03-30-2007, 12:27 AM

Yes, I've seen this before.

That guy, is clever enough to talk about "musical language" in the headers, than "composing". He's not after the cheap money, or suckers, he's researching ;)

As (I think) I've said before I (personally) can make a GWbasic program to come up with a pretty nifty little piece (but have absolutely no time now). Set the parameters and you're set. The problem lies with the different parameters. In my case the form would be the same, the ideas would be similar and some rythms or pitches would be different. This is not composing really (although really useful).

What David cope is doing, is not far off that. If the software could make some kind of choice in these parameters alone, then we would be talking. In this respect synesthesia is doing quite well I believe! ;)

dgontar

03-19-2009, 05:04 PM

Jimmymac, you wrote "Ok, Lets just establish that we don't know what human creativity is, right?
You do have a THEORY. But we cannot claim that it is true, because we don't know.
Your theory appears (to my feeble mind at least :) ) to depend on reducing creativity to the level of our current computers, which seems a little unfair to me.

I'm willing to look at any theories and consider them but surely the burden of proof is on those who want to claim that the machines are creative? Forgive me, but it just appears to be overt anthropomorphising of the machine. It's wishful thinking."

This is known to be not true. Creativity by computers is possible, in fact, it's possible in the field referred to in the subject of this forum, chess. There are many examples of computer creativity in chess. Of course if I showed them to you you might use that argument that it is not the computer creating the ingenious moves but the programmer doing it, but this does not follow. A programmer creates the moves of the chess program no more than an engineer creates the calculations a calculator performs. What distinguishes what the programmer creates from what the program creates is simply determined by an observable fact, which is that the actual computations to arrive at the output were not performed by the programmer. To be sure, the programmer creates the program, but this does not mean he is performing the actual calculations which this program executes, and the actual calculations are the basis to the actions the program or for that matter, the programmer, performs.

And a good example of this creativity can be seen in the chess game of Deep Fritz vs. Vladimir Kramnik, (2006) which can be seen here

http://www.chessgames.com/perl/chessgame?gid=1441056

In this game Deep Fritz played an opening novelty (a move never played before in grandmaster level chess) 10. Re3!!. Not only was this move a novelty, it was stylistically unique, and a very brilliant move with a subtle and profound strategy behind it. In addition to this, there are other moves that this game contains which are quite ingenious and which a human would very unlikely see, such as 19. Nb1!!, and 25. e5! And of course there are many other examples of great moves (and creative and unique moves) which a human would not see which computers play because the best chess computer in the world is rated 500 points higher than the best human for a reason, and that is because they make better moves, in general.

And in addition to this, I think that it is only a matter of time before computers are composing music, perhaps the greatest ever heard, and writing poetry. It's no fantasy. I will grant you that music programming is different from chess programming, but the arguments presented here by you and others which run along the lines that music "deals with emotion" and computers do not have emotions, or that music is limitless and chess is limited are quite flawed.
First of all Chess is not limited, or at least, not very limited (which is really what you mean). It is actually known to be extremely complex. Most of its possibilities are of the absurd or junk kind, but still many (and this is a gigantic number) are still of high theoretical significance. And yet the programmers have managed in the last 70 years to almost completely conquer it as a problem, in spite of the fact that chess grandmasters and others for years repeatedly said it could never be done. (e.g. Grandmaster Max Euwe said it could not be done in the 1950s, Grandmaster Bent Larsen said we could not ever have a computer that is world champion as late as 1993!)

Music is complex too, but not infinitely complex. (Nothing that has a basis in this reality, or level of reality (if you will) is.)Whether it is more or less complex than chess I am not sure. But even if it is more complex than chess, it is safe to say that it still is not too complex to be unprogrammable. One needs only to look at the mathematical structures of music and then compare them with chess to demonstrate this. In addition, the advent of quantum computers (http://en.wikipedia.org/wiki/Quantum_computer) is making possible programs, at least hypothetically, which are more complex than we ever imagined so we have to be careful to not say that computations of a very high number are not possible, if what is possible is defined in terms the possibilities of present day computers.

As far as emotion is concerned, the computer does not need to be conscious or experience emotion to perform an emotional output. What it needs is music and aesthetic principles programmed into it which are, for all practical intents and purposes, TIED to certain emotions, and to have layer after layer of computational structures applied to it which distinguish and separate it from the thinking of the programmer, or musician(s) and this can be successfully done in the sense that the program is distinguished from the computer because of the point I made earlier which is that the programmer and the program are distinguished from each other on the basis of what each can actually calculate. You don't call the programmer of a calculator or a calculator, because their thinking (disregarding whether they are conscious or not) is evidently different, and their output (or external output) is also evidently different as well.

Lostin Space

04-09-2009, 07:14 AM

Music doesnt need a living creator - it probably doesnt have to have a listener (!..!...yes, I know).

Music doesnt make sense to us - if it did, this discussion would not have taken place.

If humans were in total control over music, we would have killed it already - my grandmother came very close once.

BTW: Happy easter everybody ;)

Dean Krommydas

04-09-2009, 08:04 AM

...It probably doesnt have to have a listener (!..!...yes, I know).

If it doesn't...then it is serving a completely selfish purpose. I think greater meaning comes from music when it has listeners besides ourselves.

Music doesnt make sense to us

I don't know what kind of music you listen to however...it definitely makes sense to a lot of people.

If humans were in total control over music, we would have killed it already.

This comment really bothers me. "Music" is created by humans. Listening to nature, an ocean crash, birds chirping etc. has it's own inherent "music" but it takes us to extract and perceive it that way. We act as channels/conduits for music whether one believes it comes from God or from one's self. This comment... that we would have killed it already.... is not really a vote of confidence in your own humanity and it also dismisses all that we as humans have done to create, develop, evolve musically and offer back to one another (and God if you believe in God) through music. I think music survives only because of humans.

"If a tree falls in the woods..." - it still makes a sound... but in contrast...only if a human puts a CD player in the woods with...let's say a Wagner CD in it...pushes play and then walks away...Wagner's music will play for around 40 minutes or so with no one around to hear it. That CD player would function on battery power by the way....which was also created by a human. :)

Thank God for humans...

chest

04-09-2009, 09:14 AM

"If a tree falls in the woods..."
If a man speaks in the woods, does he really make a sound if there's no woman there to hear and tell him he's wrong?

And what was he doing in the woods in the first place? Hasn't he got anything more important to do?

Dean Krommydas

04-09-2009, 09:18 AM

If a man speaks in the woods, does he really make a sound if there's no woman there to hear and tell him he's wrong?

And what was he doing in the woods in the first place? Hasn't he got anything more important to do?

GOOD STUFF! :D

Lostin Space

04-09-2009, 09:41 AM

"If it doesn't...then it is serving a completely selfish purpose. I think greater meaning comes from music when it has listeners besides ourselves.

I don't know what kind of music you listen to however...it definitely makes sense to a lot of people."

Probably not in a rational philosophical way (and mostly classical music for your 2. thought there)

"This comment really bothers me. "Music" is created by humans. Listening to nature, an ocean crash, birds chirping etc. has it's own inherent "music" but it takes us to extract and perceive it that way. We act as channels/conduits for music whether one believes it comes from God or from one's self."

We both know this is to complex for us: we have no idea if it takes humans to perceive and create music. But OK...from a down-to-earth point of view: ...still no.

"This comment... that we would have killed it already.... is not really a vote of confidence in your own humanity and it also dismisses all that we as humans have done to create, develop, evolve musically and offer back to one another (and God if you believe in God) through music."

I love the humans limits - thats why we wonder and let music/everything amaze us everytime.

"I think music survives only because of humans. "

Again - it survives because of humans limits, as hunger, pain, religious handovers and drugs.

"If a tree falls in the woods..." - it still makes a sound... but in contrast...only if a human puts a CD player in the woods with...let's say a Wagner CD in it...pushes play and then walks away...Wagner's music will play for around 40 minutes or so with no one around to hear it. That CD player would function on battery power by the way....which was also created by a human. :)"

Wagners music would live forever because of space/distance/waves - not 40 minutes. We have no idea it takes human ears to receive it.

"Thank God for humans..."

Agree :-) And thank God that some people believed in God over the years - how else would most music survive through religious church traditions (not excluding all the blood, evil, and misunderstanding within it).