It’s been just a few months since I checked off one of the top items on my bucket list. This past May, my wife and I traveled to London and Amsterdam. We built our trip around attending several concerts at the third-ever Gustav Mahler Festival [1]. We’d awoken at 3:30 a.m. back in February to order tickets, which went online at 10 a.m. Amsterdam time.
We booked three concerts, but the centerpiece was a Sunday afternoon performance of Mahler’s 8th Symphony, The Symphony of a Thousand — nicknamed thus because it requires a huge orchestra, two (!) full choirs, a children’s choir, and eight solo vocalists. No surprise that it’s rarely performed. Up until this past May 18, I’d never heard it live.
I’ve loved this extravagant, lush, and intensely expressive piece ever since I got my hands on a three-LP set with Mahler’s 2nd and 8th Symphonies [2]. Each symphony took three record sides. Both performances were stunning. But I remember listening to the first movement of the 8th (which took up one full record side — Mahler’s music is not succinct!) repeatedly.
Now I sat in the beautiful Royal Concertgebouw concert hall, facing the immense Royal Concertgebouw Orchestra. In front stood the soloists, and behind the massed choirs. The organ and strings blasted out the thunderous opening chord, and the choirs lifted the hearts of all attending with the opening lyrics:
“Veni, veni, creator spiritus!”
“Come, come, creator spirit!”
The words are from a medieval hymn. But feeling the tidal wave of sound wash over me, I felt this spoke to something contemporary, something very much on my mind lately — what creativity means, and what its value is, in the context of generative AI.
This invokes a word in my mind. Counterpoint. In both the musical and general sense. I turn, as I often do, to the Oxford English Dictionary.
Counterpoint – Music: “The art of adding one or more melodies as accompaniment to a given melody or ‘plain-song’ according to certain fixed rules; the style of composition in which melodies are thus combined.” [3]
Mahler is a master of counterpoint. Though he wrote for mammoth musical resources, his contrapuntal textures are as transparent as J. S. Bach’s. In fact, Mahler often used Bach’s favorite form of counterpoint, the fugue. The tremendous double fugue Mahler composed for the first movement of his Eighth Symphony is a splendid example. I wrote about Bach’s fugues and data governance in Towards Data Governance’s 4th Era – Part II. You can find out about the basic structure of fugues in that article.
Counterpoint – General: “The opposite point; the exact opposite, antithesis.” [4]
I can’t think of a more extreme counterpoint to Mahler calling the creative spirit than AI-generated music. This article, 10 Top Text to Song AI Tools of 2024 from autogpt.net, illustrates the point. Among its top 10 music-extruding machines [5] is Suno. I wrote a LinkedIn post about Mikey Shulman, Suno’s CEO and founder. Mikey had said on a podcast that “It’s not really enjoyable to make music now.” It’s too much work without AI. As a musician myself who enjoys making music without AI, I had some fun at Mikey’s expense.
I can’t comment on the quality or capabilities of any of these products. I have not tried them. I’m not interested in doing so. But I found the article’s “Benefits of Text to Song AI” vs. “Challenges and Limitations” sections useful. Revealing and illustrative of counterpoint.
Both sections mention creativity:
“Text to song AI encourages creativity by allowing users to experiment with different styles, melodies, and instruments. It opens up new possibilities for songwriting and music composition.”
Fine. What’s wrong with allowing a person to experiment with different styles, melodies, and instruments? When I tried composing as a middle schooler, I had only two instruments at my disposal: cello (which I could play adequately) and piano (which I could barely play at all). In case you’re wondering, no, I didn’t find success as a composer.
But, for “Challenges and Limitations,” the article’s author, Onome [6] does admit:
“While AI can generate music, it may not always capture the creative nuances and unique styles of individual musicians. Human touch and creativity are still essential for producing truly original and expressive music.”
You think? Can AI ever “capture the creative nuances and unique styles of individual musicians”? With its output based on billions of bits of training data, assembled by algorithms from the fruits of human creativity? No, AI cannot do this.
Back to Counterpoint: thesis (Mahler’s 8th – Come, Creator Spirit) and antithesis (AI music generators like Suno).
Notably, Onome does include “Ethical Concerns” among the “Challenges and Limitations”:
“There are ethical concerns surrounding the use of AI in music production, particularly regarding copyright and intellectual property. It is important to ensure that AI-generated music respects the rights of original creators and composers.”
I’ve written about this aspect of generative AI. And I listened to an intriguing discussion on this theme on one of my favorite podcasts, Jacke Wilson’s The History of Literature Podcast. A recent episode featured Jacke’s interview with Mira T. Sundara Rajan. Mira is a law professor and an expert on copyright, as well as a writer and musician. They discussed her new book, The Moral Rights of Authors and Artists: From the Birth of Copyright to the Age of Artificial Intelligence.
I found the interview fascinating and informative, giving me a perspective on copyright’s birth and development. And the concept of moral rights is fascinating. But when the conversation turned to AI, Mira’s comments resonated deeply with me:
“You know, I think here we’re really facing something existential that is not existential in the sense of devaluing human creativity, not at all. But I think the problem is more that we as a society have kind of given away our creativity.
“And we don’t, as a society, look upon creativity as our greatest asset anymore. You know, I don’t think that this was the case in the past. I think that creativity was much more recognized and prioritized in the past.
“Even at the time of the censorship regimes, it’s not that the regimes didn’t recognize the value of creativity. Of course they did. They thought it was the most dangerous thing in the world.
“So now we’re in a situation where somebody seems to think that what machines can do is as valuable as human creativity and perhaps more valuable in many ways than human creativity. And to me, this is the fundamental problem. We haven’t put enough importance on fostering human creativity.
“And instead, we’re kind of thinking that we’re going to outsource this to machines, and somehow that’s going to make something better for someone.”
Yes! And fostering creativity doesn’t mean making composing music as easy as typing a prompt and letting generative AI do its thing. There’s nothing wrong with creating through sweat equity. Even if Mahler, well known for his titanic struggles composing music, wrote the 8th Symphony in the ridiculous span of 10 weeks, composing, according to his wife Alma, “as if in a fever.”
Christopher Johnson, linguist and writer, and one of my go-to LinkedIn posters, spoke eloquently about creativity and generative AI on a recent episode of Human Intelligence Podcast. When asked why creativity matters, Christopher answered simply, “Because it’s what makes life beautiful.” I can’t agree more.
Later, Christopher speaks to a salient characteristic of creativity that relates directly to what I felt at that performance back in May:
“Yes, so one important aspect of creativity in my mind is communication. I’m talking about creativity in the sort of artistic sense or literary sense…How do we interpret when we read a novel or read a nonfiction work? The process of interpretation is not just a matter of extracting bits of information from words, but it’s thinking about why did the author put it that way? Why did the author make the choices that the author made?
“And this is a general idea that applies to any sort of creative expression. Like you’re looking at a painting. You get interested in why the artist chose to pick this object in this way. That’s part of the meaning of the work … That is intrinsically a process of communication because we’re making inferences about what was going on in somebody’s mind … If you take that away, then it’s hard to even really talk about whatever product you’re contemplating as being created.”
The same is true with listening to a piece of music. Take the Symphony of a Thousand. Why did Mahler choose to pair an ancient Latin hymn in the first movement with the final scene of Goethe’s Faust for the last movement? What drove Mahler to assemble the most musicians ever for this salute to creativity?
But more than inferring what was in Mahler’s mind when he wrote this piece is how it’s framed in a particular moment of the composer’s life. We sandwiched hearing the Symphony of a Thousand between two other festival concerts. Mahler’s 9th the night before. And the 10th and the other worldly symphonic song cycle Das Lied von der Erde (The Song of the Earth) that same evening (both by the outstanding Berlin Philharmonic). Between the 8th and these works, life dealt Mahler three hammer blows: “The death of his eldest daughter, his condition of a weakened heart (which eventually caused his death), and his ouster from the Vienna Opera.” Knowing the biography adds a poignant contrast. The joyous affirmation of the 8th. The haunted atmosphere of the latter works, filled with tragic echoes and hints of impending death. Nothing extruded by AI could have brought these aspects to the performance.
But what about AI-generated music? What Christopher Johnson said about text and images holds true for music too:
“So if I get a bunch of information from the spit out of ChatGPT, a bunch of text, I can’t ask any of those questions about it. I can’t ask, why did the large language model choose to express it that way? Because the answer is always the same. It did because that’s how the number crunching turned out. There’s no conceptual motivation for anything that you’d observe in the output.
“So, it’s not creativity. When you are looking at text spit out, it’s not creative. When you are looking at generated images, it’s not creative. It’s just because there’s no communication.”
Mahler wrote his symphonies in the last decade of the 19th century and the first decade of the 20th. He died in 1911, 50 years old. Yet he communicated vividly to all those who attended those unforgettable concerts.
Like Mira T. Sundara Rajan, I remain optimistic about technology. I feel as she does when she says she “can think of many contexts in which AI could be an incredibly beneficial tool.” But I also share the belief of Joseph Weizenbaum. Weizenbaum, the AI pioneer who created the ELIZA chatbot back in the 1960s and wrote Computer Power and Human Reason. Weizenbaum wrote, “There are certain tasks which computers ought not be made to do, independent of whether computers can be made to do them.”[7] Delegating our creativity to AI is one of these.
