When VAN asked me to do a review of an artificial-intelligence-created realization of Beethoven’s Tenth Symphony called “Beethoven X: The AI Project,” which is based on the skimpy sketches he left when he died, I more or less groaned in my reply. “Not for me,” I said. “I know pretty much what I’ll think about it, and my review could get snarky.” “If so, that would be all right with us,” VAN said. “Well, OK,” I groaned back. So here I am and here goes.
I’ll start by noting that a while ago I spent a dozen years writing a hefty biography of Beethoven. That doesn’t make me any sort of ultimate authority, but it does mean that I know a bit about the guy and his work.
In their publicity I read the rosy words of the project creators with their ambiguously impressive credentials. (The composer on the team is best known for an Intel TV jingle.) I took in the third and fourth movements the computer disgorged after programmers had primed it with a massive dose of Beethoven data. I kept my mind open and so forth. After all, when I expect little from some matter or other I’m always glad to be pleasantly surprised. At the end of the symphony I found myself more philosophical than annoyed. I’ll start with that.
One recalls some precedents in machine improvements on the human. In the realm of fiction, I thought of the machine-created popular tunes that manage to move Winston Smith in Orwell’s 1984. In the real world, I recalled that in 1997, IBM’s computer Deep Blue bested reigning chessmaster Garry Kasparov. When IBM’s Watson supercomputer beat a couple of “Jeopardy!” winners in 2011, one of the players noted that he found Watson had a lot in common with other top-ranked “Jeopardy!” players: “It’s very smart, very fast, speaks in an uneven monotone, and has never known the touch of a woman.” (Watson pocketed, or whatever computers do with their money, a million bucks for the win.) I thought inevitably of HAL, the talking computer in “2001: A Space Odyssey,” who is programmed to express empathy and turns out to have real emotions, such as panic.
I meanwhile thought of how thrilling it is when a baseball outfielder grabs the ball off the wall and throws it dead-on to home plate to catch a runner. Doesn’t happen very often. A computer with a throwing arm, of course, could do the same every single time, without fail. A simple machine can throw a baseball miles further than any human. I thought of what a composer teacher of mine told me about his first experience with electronic music. When he started working with an analog synthesizer and tape (that being what you did in the ‘70s) he was interested in creating complex rhythms beyond the capacity of human musicians. He did so. And he discovered that it didn’t sound like anything, was about as interesting as throwing a handful of gravel on a tin roof. Which is to say, it didn’t matter. “What I realized,” he told me, “was that I wanted the intensity of real musicians struggling to play complicated rhythms.”
I have a late friend who I think was a very fine painter, though probably few know her name by now: Francis Gillespie. She would spend a year or more on a painting of flowers, struggling to represent them with virtually photographic accuracy. In fact, as she knew perfectly well, she didn’t have the technique to do that. “I’m really sort of a primitive,” Fran would say grimly as she worked. But what makes her paintings hers is exactly the grand failure of her attempt. Her pictures are beautiful, close to photos, but always a little off, and the offness makes them singular. In the same way, the excitement of hitting in baseball comes from the fact that the best hitters fail to connect two-thirds of the time.
I’m midway in the philosophizing here, but my point so far is obvious enough: The ability of a machine to do or outdo something humans do is interesting once at most. Deep Blue isn’t playing chess anymore and Watson isn’t on “Jeopardy!” because nobody cares. It doesn’t matter. We humans need to see the human doing it: Willie Mays making the catch that doesn’t look possible. When it comes to art, we need to see a woman or a man struggling with the universal mediocrity that is the natural lot of all of us and somehow out of some mélange of talent, skill, and luck doing the impossible, making something happen that is splendid and moving—or funny, or frightening, or whatever the artist set out to do.
That’s half the answer to why I wasn’t expecting much in the way of revelation from DeepLudwig. The other has to do with a conviction about computers in general and artificial intelligence in particular, when it comes to their attempts to play ball on our turf in one way or another. Here’s my assertion: True intelligence is in a body. Intelligence outside a living body, as some sort of abstraction, is innately impossible, or should be given another name. When it comes to the significant things in life—love, loss, lust, yearning, rage, confusion, God, our abject awe at the spectacle of the cosmos, and so on—artificial intelligence is incompetent and in some ways meaningless, outside an ability to figure out the weather better and other useful and even critical practical matters. Nobody argues with AI’s ability to make a better mousetrap. If humans screw it all up just right, AI could save the world. (It could also take over the world, as Stephen Hawking feared, but that’s another matter.)
Artificial intelligence can mimic art, but it can’t be expressive at it because, other than the definition of the word, it doesn’t know what expressive is. It also doesn’t know what excitement is, because there’s a reason people call excitement “pulse-pounding,” and computers don’t have pulses. When computers set out to do art, they don’t fashion it in a whirl of creative trance inflected by a deadline; they can’t account for the heat or alarming lack of it in the room, sensations in the groin, the failure or success of drawing a foot that looks planted on the ground, the failure or success of creating rhythmic momentum on the page, the bit that’s bullshit and needs to be fixed and the bit that’s really good and you see where it wants to go, the woman or man you just met who excites you and whom you hope to excite, the thought of the idiots who think they can write as well as you, also the bastards who write better than you, what you’re having for dinner or what you had for dinner that’s not agreeing with you, the hairs falling out of your head onto the page, the expense of ink or paint or the rehearsal costs of a symphony orchestra, and so forth and so on, while in your trance you’re trying to conjure out of the air a portrait in words or tones or images of, say, that man or woman who maybe will appreciate you for the effort. Along with all that and maybe above all that, the gnawing and relentless drive to do something really good, this time, for all the above reasons and more, whether it’s trying to convince the woman to love you, or the public or God to love you, and/or to pay the rent, and to show yourself that you can damned well do something at least in the direction of really good in this possibly cursed endeavor that you believe you’ve been born to do, and without which your life is something in the direction of meaningless.
None of that can be programmed into a computer in any way that means anything. To repeat: The only true, meaningful intelligence is in a body, and likewise the only true and meaningful creativity. By “body” I’m including the mind, contained in a brain. And as has been said, the brain is the most complex thing in the universe. This has to do with one of the things particular to Beethoven: In contrast say to Mozart and Bach, with him it’s often as if you can hear the effort, the struggle, hear in the notes what it cost him to rise above the universal mediocrity—and the frequent mediocrity of his sketches—into something that he considered worthy of his name, worthy of an immortal place in the long human chronicle, worthy of the talent God’s nature gave him. That, and the need to pay the rent.
All this was, in embryo, in my mind when I groaned at the invitation to write a response to the AI realization of Beethoven’s Tenth Symphony. Now I’m going to take off the philosopher’s hat and put on the music critic’s.
To start, the turn-on-the-radio test: If I turned on the radio and heard this music, would I guess it was by Beethoven? No. The orchestra doesn’t sound like Beethoven. If anything it sounds rather like Robert Schumann’s handling of the orchestra, which is to say a bit clunky—though Schumann is rarely as clunky as this. The scoring is competent as far as it goes, you can hear everything fairly clearly, there’s some variety of color, but in the end it’s aimless and uninspired and sometimes tries too hard, such as some grandiose trombonic moments that Ludwig wouldn’t do that way.
Could this computer effort pass for bad Beethoven? Again, no. He wrote his share of cheesy stuff, but the difference is this: The computer is doing the best it can, whereas when Beethoven wrote junk it was usually because that’s what he intended to do. He was as good at crapola as everything else. The best example is the gloriously trashy “Wellington’s Victory,” an orchestra piece portraying the English victory over Napoleon (the score is complete with musket volleys) that was designed to rouse the crowd with patriotic twaddle and make him a pile of money—which it did. As he scrawled across a newspaper complaint about how tacky the piece is: “What I shit is better than anything you ever did!!!” He was not talking about his personal excrement, but rather the slickly crafted potboiler that accomplished the job he intended it to. (Some of the climaxes in AI’s Tenth are audibly based on “Wellington’s Victory.”)
What about our computer’s musical material, the notes as distinct from the orchestral garb? The creators input Beethoven’s sketches and programmed their computer to think in terms of the kind of motivic development that the composer carried on with simple material. An example is the dah-dah-dah-dum rhythm that begins the Fifth Symphony, which permeates the rest of the piece in constantly evolving guises. That’s the kind of thing an artificial composer can understand; it can take the rags and bones of Beethoven’s sketches and develop the ideas dutifully. There are some problems with this. First, a final-draft Beethoven theme does not just take up the ideas of the piece logically, he also insisted on it being a good theme, by his lights, which happens to be by the lights of most people. There’s a rejected sketch of the second theme of the Fifth Symphony’s second movement that absolutely uses the motifs and shapes and such that he’s been working with, but it’s a bland tune and he replaced it with one no more logical, but much more interesting and expressive. And again: Artificial intelligence can never understand what “interesting” and “expressive” amount to in regard to living, breathing, loving, lusting, aspiring, etc., human beings.
Many of Beethoven’s sketches, even for his greatest works, are bland and conventional. They’re also mostly jotted down in one line, while what he was hearing in his head was a string quartet or a full orchestra. He was invariably working toward an overall conception, like the “from darkness to light” progress of the Fifth Symphony, that may not be evident in the sketches at all. Often in the sketches you see how he refines and transforms ordinary ideas into something fresh and remarkable, and always apropos in the context of the whole work. Sometimes he can take a banal idea and magically turn it around on itself, revel in it, make its simplicity into something ravishing, like the silly little two-bar folktune that repeats over and over in the trio of the Ninth Symphony scherzo, which gradually grows into something like a definition of peace and joy. In the AI Tenth, the Beethoven ideas sit there uncooked, naked in their simplicity, developed but not turned around on themselves, not reveled in.
A final point with the music critic hat. DeepLudwig has a problem with rhythm. Skilled composers are masters of using rhythm to keep the energy and continuity of a piece going, coasting or intensifying as needed. One of Beethoven’s familiar devices is to compress a rhythm to goose the momentum. It’s something like this: dum dum didddlediddle, dum dum diddlediddle, dum diddlediddle, dum diddlediddle, diddlediddlediddlediddle… etc. Our computer didn’t know how to keep the momentum going. The rhythms just sit there, headed nowhere in particular. There are climaxes in the appropriate places in the form, as the computer has been taught, but they haven’t been earned."In the AI Tenth, the Beethoven ideas sit there uncooked, naked in their simplicity, developed but not turned around on themselves, not reveled in." Jan Swafford reviews Beethoven X—the composer's unfinished Tenth Symphony, as completed by AI.… Click To Tweet
In the end, what were the creators of the AI Tenth Symphony actually trying to do? Did they want to create a piece that would be added to the Beethoven repertoire, played by orchestras everywhere? I hope that they weren’t attempting something that futile. Were they more modestly trying, like the creators of Watson and Deep Blue, to demonstrate the potential of computers to think faster and more pointedly than humans in mimicking human endeavors? If so, I’ll concede that they more or less succeeded: After an impressive feat of programming by its human proprietors, AI produced something that sounds unquestionably like a piece of music, only a gangly and forgettable one.
So, sadly, we’ll have to continue to accept Beethoven’s Tenth as one more thing on which the hand of fate put the kibosh, like the dozens of unfinished drafts Mozart left behind. Near the end of his life Beethoven told a friend that in the Tenth Symphony he was attempting to discover “a new kind of gravity.” We’ll never know what he meant by that, but we can dream about it. Dreaming being, yet again, something a computer can never do. ¶
Subscribers keep VAN running!
VAN is proud to be an independent classical music magazine, and we’re able to do so in part thanks to our subscribers. For less than $0.11 a day, you can join our community of supporters, access over 550 articles in our archives, and get new ones delivered straight to your inbox each week.
(PS: Not ready to commit to a full year? You can test-drive VAN for a month for the price of a coffee.)