Can a computer program produce classical music compositions that rival the works of Bach or Mozart?
Emmy was once the world’s most advanced artificially intelligent composer, and because he’d managed to breathe a sort of life into her, he became a modern-day musical Dr. Frankenstein. She produced thousands of scores in the style of classical heavyweights, scores so impressive that classical music scholars failed to identify them as computer-created. Cope attracted praise from musicians and computer scientists, but his creation raised troubling questions: If a machine could write a Mozart sonata every bit as good as the originals, then what was so special about Mozart? And was there really any soul behind the great works, or were Beethoven and his ilk just clever mathematical manipulators of notes?
Cope’s answers — not much, and yes — made some people very angry. He was so often criticized for these views that colleagues nicknamed him “The Tin Man,” after the Wizard of Oz character without a heart.
[ . . . ]
Finally, Cope’s program could divine what made Bach sound like Bach and create music in that style. It broke rules just as Bach had broken them, and made the result sound musical. It was as if the software had somehow captured Bach’s spirit — and it performed just as well in producing new Mozart compositions and Shakespeare sonnets. One afternoon, a few years after he’d begun work on Emmy, Cope clicked a button and went out for a sandwich, and she spit out 5,000 beautiful, artificial Bach chorales, work that would’ve taken him several lifetimes to produce by hand.
Although the article is fascinating, I must confess that my primary reason for posting about it was to provoke a rant—and defense of Bach—from David Goldman.
(Via: Kottke)
March 10th, 2010 | 6:22 pm
Mozart did not use magic to create music, but a brain shaped by experiences. Composers who followed learned from his compositions by listening — that is, they discerned rules for “good” music (formally known as music theory.)
So computers, like humans, can analyze and learn from great composers by using the simple but powerful tools of comparison and contrast. By organizing compositions into subsets based on shared or similar properties, the logical deduction of rules is made possible. In this way a computer can start with a “random” melody and develop it accordingly. Each iteration is then judged by its likeness to the properties of known subsets, and shaped by further application of rules.
This should not be surprising in light of the fact that birds and whales also create music without the aid of a human soul. Humans, however, create music suited for our physiology and our psychology. But when a computer analyzes Bach or Mozart, it learns to create music specifically for humans sensibilities — computers can not create music to please itself. So while a computer may be able to produce Art, it has no means to enjoy it as such.
March 10th, 2010 | 8:36 pm
I’ve heard this sort of thing before, and I still don’t believe it. Cope advertises the creative powers of his own brainchild, Emmy, and accuses detractors of romantic bias, but, until Emmy’s music starts selling out concerts on a regular basis, I’m going to suspect that it is Cope who has fallen victim to the oldest bias in the world, the bias in favor of one’s own idea.
Personally, having sampled the work of Emmy on the linked website, I am not impressed. It sounded to me a lot like what it is: tones arranged in a mathematically logical manner, but without any organizing genius to provide dramatic coherence. And if this is an “incredibly romanticized” view of music, as Cope alleges, then I plead guilty.
I hope that I would be capable of distinguishes these sample pieces from the works of great composers. If I couldn’t, then my first conclusion would be that I am a horribly obtuse listener, not that the human mind is equivalent to a computer. The fact that many audience members couldn’t distinguish Emmy’s fugal pieces from Bach is probably due to the fugue’s more archaic and unfamiliar properties, relative to classical and romantic forms. If a musician steeped in Baroque music couldn’t tell the difference, then Emmy would perhaps deserve more respect.
March 10th, 2010 | 9:26 pm
It is truly a remarkable accomplishment of the rational human soul to have written a computer program that can capture the essence of a composer’s musical style so perfectly that it can successfully mimic that style.
This accomplishment is as remarkable as the original accomplishments of the musical geniuses who’s styles are being mimicked.
I somehow doubt that these “compositions” will make a lasting mark on our cultural history. The creative genius of that produced Emily, however, might.
March 10th, 2010 | 9:33 pm
We humans are much too prideful of our cognitive abilities, to the point where we conflate intelligence with divinity and the soul. Yet we should remember that Adam and Eve possessed souls – and a state of grace – before being cursed with (the pursuit of) knowledge.
From the Harvard Crimson, November 9, 1993
“In a widely publicized match between Deep Thought and Kasparov in 1989, the world’s best human player beat the computer twice. But Deep Thought did manage to beat a number of grandmasters on various occasions, including Robert Byrne, chess columnist for The New York Times … But will computers someday beat humans in all games? Most human chess masters doubt, or at least convince themselves to doubt, that computer players will ever be as good as the best humans.”
Garry Kasparov, February 11, 2010:
It was my luck (perhaps my bad luck) to be the world chess champion during the critical years in which computers challenged, then surpassed, human chess players. Before 1994 and after 2004 these duels held little interest. The computers quickly went from too weak to too strong…
Perhaps chess is the wrong game for the times. Poker is now everywhere, as amateurs dream of winning millions and being on television for playing a card game whose complexities can be detailed on a single piece of paper. But while chess is a 100% information game—both players are aware of all the data all the time—and therefore directly susceptible to computing power, poker has hidden cards and variable stakes, creating critical roles for chance, bluffing, and risk management.
These might seem to be aspects of poker based entirely on human psychology and therefore invulnerable to computer incursion. A machine can trivially calculate the odds of every hand, but what to make of an opponent with poor odds making a large bet? And yet the computers are advancing here as well. Jonathan Schaeffer, the inventor of the checkers-solving program, has moved on to poker and his digital players are performing better and better against strong humans—with obvious implications for online gambling sites.
March 10th, 2010 | 9:49 pm
Well said, Paul. I have to say I’m pretty unimpressed by the assertion that concert-goers couldn’t tell the difference between Bach and Emmy. Big deal. They probably couldn’t tell the difference between, say, early Mozart and Haydn. The point is, get back to me when Emmy writes No. 9.
March 10th, 2010 | 10:40 pm
Could a computer be an effective blogger?
March 10th, 2010 | 10:52 pm
Brian Eno has done similar work on what he calls generative music: http://www.inmotionmagazine.com/eno1.html
My view is somewhere between Hampton’s and Paul’s.
I am a composer who happens to have the gift of being able to listen to a musical style and quickly compose something that sounds similar. Each style, and each particular song, has its own patterns. They’re not too difficult to recognize if one knows what to look for. Though no AI system can match human logic, much less human creativity, I believe music theory and pattern recognition could be programmed well enough that software could create pleasing music.
But note the composers who Cope’s program is touted against. The music of Mozart and Bach is more straightforward and symmetrical than that of Romantics like Beethoven or Tchaikovsky. There is more reliance on modulation (repeating a pattern from a different base note). The music appeals more to intellect than emotion. It more often evokes general moods than takes one on emotional adventures.
And, as Paul suggests, has an AI yet created a masterpiece with mass popularity? Or can an AI start a new musical movement?
I can trust that Cope’s software is capable of creating pleasing music. I have experienced Eno’s generative music in a game called Spore, and it performs well as environmental music. But software will always have limits which human beings do not. I’m reminded of the film War Games.
Let’s not forget that music publishers have tried for generations to codify popular styles and churn out hit songs by formula. Undoubtedly, they have had much success… but not as much as they would like.
March 10th, 2010 | 10:52 pm
The best human chess players have ELO ratings around 2800, while the best chess engine is rated 3200. It doesn’t make chess any less enjoyable to play, and it hasn’t meant the end of chess. If anything, just the opposite.
I don’t think this is that big a deal. First, if there’s a program that can make more music like Beethoven, then please never turn it off. Second, the software could never make music like Bach until there actually had been a Bach, and I’m pretty sure that machines still can’t make new human beings, music prodigies or not.
March 10th, 2010 | 11:06 pm
I just listened to the samples on the webpage, and I agree with Paul, they weren’t that good.
March 11th, 2010 | 1:24 am
Social comments and analytics for this post…
This post was mentioned on Twitter by ROFTERS: Can a Computer Compose Better Music Than Mozart? http://bit.ly/asyjT9...
March 11th, 2010 | 4:12 am
[...] Can a Computer Compose Better Music Than Mozart? » First Thoughts … AKPC_IDS += "3832,";Popularity: unranked [?] Share and [...]
March 11th, 2010 | 11:44 am
If those two listening samples are the best Emmy can come up with, I’m not too worried. They are idiomatic but have little coherent form. There is little art in them. It’s the humanness of art, or the artfulness of being human, that a computer will never be able to replicate.
March 11th, 2010 | 11:57 am
It’s not just a computer composing, it’s a computer programmed by a human being — ergo, human creativity is still in play, albeit the creativity of programming a computerized composer.
March 13th, 2010 | 3:18 pm
Beware of attitudes evident in lines like these from the article, informing judgment about composition in general, as well as Cope’s projects,
“Algorithmic composers use a list of instructions — as opposed to sheer inspiration”;
“a level of planning that’s rare among free-spirited musicians”
The great composers were all “algorithmic composers”; no one works on “sheer inspiration”, not Robert Johnson, not Tchaikiovsky. Even when they claim to, it’s hyperbole and what they actually do can be as laborious as bricklaying. When composers need inspiration, they find many varied ways to create for themselves, as Cope did with his gargantuan effort.
The story of Cope’s development of the Emmy program is as much scientific as artistic. The process of refinement is a classic case of scientific investigation, trial and error.
The program is as good as Cope’s own powers of composition, analysis and intuition, no more no less, and they’re pretty darn good.
As for those here who say the sample don’t seem that great, read the whole article for cautions to your presumptions.
March 14th, 2010 | 7:24 pm
Mr. Cummins:
I did read the whole article. I never questioned Cope’s talents as a programmer, nor did I claim that musical composition is “sheer inspiration,” whatever that may mean. I simply found the programmed music to be dull and tried to explain why. I gave my opinion for what it’s worth. There’s nothing particularly presumptuous about that. On the other hand, if Cope (as opposed to his admirers) is claiming equality with Bach and other luminaries, then presumption is just the word for it.
March 16th, 2010 | 6:05 pm
Above R Hampton wrote an example of an essential confusion,
“These might seem to be aspects of poker based entirely on human psychology and therefore invulnerable to computer incursion. A machine can trivially calculate the odds of every hand, but what to make of an opponent with poor odds making a large bet? And yet the computers are advancing here as well. Jonathan Schaeffer, the inventor of the checkers-solving program, has moved on to poker and his digital players are performing better and better against strong humans—with obvious implications for online gambling sites.”
“Computer incursion” and “computers are advancing here as well” are incorrect. It is Jonathan Schaeffer who is advancing. Douglas Cope’s skills as a programmer were developed to serve his skills as a composer and analyst of music and both skills then developed together. The computer is nothing; these are all human products. Kasparov would have been beaten earlier if the computer hardware had been up to the task. This fact is emphasized in the Crimson article R Hampton refers to. Note the mathematical evaluation of the nature of a chess game, the crucial distinction between winning and solving the game, and the bit about Turing’s chess program, “he didn’t even have a computer! He processed the algorithm on pieces of paper and this “paper machine” played a competent game.”
Cope’s Emmy program hasn’t solved music but it is such that, and Bach’s and Mozart’s intellects and work ethics were such that, were Bach and Mozart living now, they would be capable of improving Emmy’s analysis engine to emulate their own perceptions and powers, to produce not just better copies of their music than Cope could, but to actually produce their own music.
It’s not evident in biographies for lay readers, but but most of the great composers in the Germanic tradition dealt all their waking lives with the mechanics of music in as objective and thorough a way as engineers or scientists deal with their projects.
Not all these composers had the level of insight into both music in general and their own music, and the raw ability to program, that Bach and Mozart had, but they were close enough to have been able to use something like Emmy successfully. Even one like Schubert could have, a proposal that may be surprising to laymen who are generally unaware of how thoroughly the supposed paragon of swooning emotionalism pursued theoretical knowledge. Though it’s a bit better known that Bach could accurately predict what another composer would do with their material after the first theme and that Chopin had a strong interest in mathematics and relentlessly analyzed his own works-in-progress to justify his inspired material against his standards of craft and logic.
What is the raw ability to program and why did Bach and Mozart likely possess it? It’s comprised of immediate problem-solving and minute, precise organization directed towards immediate goals, and the ability to repeat these two tasks on larger scales formed from aggregates of them directed towards larger goals. Musical composition, especially in the Germanic tradition, is the same thing.
Computers are nothing; it’s the human mind that is so great at providing tools, whether purely mental ones or electrified ones, to investigate, manipulate and render symbolic systems and patterns.
Yes, the few examples do seem a bit dull, but it might be the human playing. Plenty of humans are capable of dulling masterpieces. As far as substance, the simulated Chopin seemed the best.
Here’s the article’s chief caution against presumption,
‘At one Santa Cruz concert, the program notes neglected to mention that Emily Howell wasn’t a human being, and a chemistry professor and music aficionado in the audience described the performance of a Howell composition as one of the most moving experiences of his musical life. Six months later, when the same professor attended a lecture of Cope’s on Emily Howell and heard the same concert played from a recording, Cope remembers him saying, ‘You know, that’s pretty music, but I could tell absolutely, immediately that it was computer-composed. There’s no heart or soul or depth to the piece.’ ”
The admonition wasn’t meant pejoratively. Anyone who’s had the experience of comprehending and liking a piece of literature, art or music that they’d previously found entirely unenticing for any number of reasons based on maturity and/or experience and knowledge knows how mistaken perception can be.
March 16th, 2010 | 6:30 pm
sorry, correction:
“This fact is emphasized in the NYRB article R Hampton refers to.”
March 17th, 2010 | 1:23 pm
Paul, it was the article on Cope that used the quoted phrases that juxtaposed “list of instructions” with “sheer inspiration”, and “level of planning” with “free-spirited musicians”. This is dumbing-down glibness on the part of the author, though the rest of the article redeemed them by going on to refer to what really goes on in composing, analysis and programming.
March 29th, 2010 | 12:58 pm
John Cummins:
You make some interesting points, and I give credit to your dispassionate perseverence in this discussion. But I still think you are wrong. The indisputable fact that mathematics plays a central role in musical composition (with the related fact that many composers were math ethusiasts) does not mean that music is an essentially mechanical endeavor, if that is what you are saying.
You seem to claim on behalf of Cope that computer music is a product of human imagination just as much as music made the old-fashioned way; therefore, Bach and Mozart, for example, had they lived in a computer age, could have composed equally great works, using Emmy as a tool. This does not seem to jive with Cope’s position, for, at least in the article, he definitely pits Emmy against mere human composers. He is almost as gushing and immoderate when speaking of her as we incurable romantics are when speaking of Chopin.
Anyway, I don’t think your position, as far as I can decipher it, fits the facts. Haydn may have played musical dice, but that’s not the same as the dice taking over and writing the music. The random element in some of Haydn’s or Bach’s experiments was only the springboard for their creativity.
Your analogy to the chess computers is also revealing. There is clearly a creative component to chess that the computers did not have, else how could Kasparov have beaten the computer even once? It is curious how the computer programmers (in all too human a fashion) gloated over Kasparov’s defeat in a majority of matches. Yet, to me, the fact that Kasparov won ANY match, even with such unfair handicaps, is proof positive that there is more to chess than mathematical calculations.
I’ll conclude with a quote from a friend of mine relative to this thread: “A human being’s creativity goes beyond that, doesn’t it? I mean, you wouldn’t say, for example, that Einstein came up with his theory of relativity by inserting random elements into the established laws of Newtonian physics and then — voila! — the theory of relativity. In the same way, you wouldn’t say that Bach’s creativity amounted to random variations in the rules of counterpoint. In other words, there is something about human creativity that transcends a simple (or even incredibly complex) system of rule-following and random variation — which is all a computer can ever do. There is, for one thing, the intuition, the seeing.”
April 1st, 2010 | 9:02 pm
For anyone interested in an in depth look at this topic. Check out the website link. I started my CS degree with background in Cope’s work and wanted to dive into “emotion” and music…The link is my thesis on emotion IN algorithmic music.
Links
Blogs
Find Us
Contact