Will AI ever write a good song?
Recently, Nick Cave responded to a song that had been sent to him by a fan, calling it a “grotesque mockery of what it is to be human”. However, thankfully, for both the fan and the rest of us, the song had been written by an AI programme called ChatGPT, developed by the artificial intelligence lab OpenAI.
The fan, Mark from New Zealand, had asked ChatGPT to write a song “in the style of Nick Cave”. Admittedly, the result was rather Cave-esque in its approach to lyrical composition. For example, the chorus to the song produced the words: “I am the sinner, I am the saint, I am the darkness, I am the light, I am the hunter, I am the prey, I am the devil, I am the saviour”.
Cave was understandably outraged and sickened by the feeble attempt of an AI programme to replicate his work. He wrote in his The Red Hand Files newsletter: “The apocalypse is well on its way,” before explaining that “songs arise out of suffering, by which I mean they are predicated upon the complex, internal human struggle of creation and, well, as far as I know, algorithms don’t feel.”
“Data doesn’t suffer,” Cave continued, “ChatGPT has no inner being, it has been nowhere, it has endured nothing, it has not had the audacity to reach beyond its limitations, and hence it doesn’t have the capacity for a shared transcendent experience, as it has no limitations from which to transcend.”
Evidently, Cave was not amused nor impressed by ChatGPT’s efforts to write a song, regardless of whether it was in the style of his own work or not. Still, the fact remains that the AI programme managed to churn out (whether Cave will admit to it or otherwise) a half-decent song that several millions of living human beings would be unlikely to be able to muster.
Well, on the surface, at least. While the lyrics to the track may appear somewhat impressive to an unsuspecting onlooker, all their qualities are quickly drained away when we learn they are meaningless. Meaningless in that they have been generated from data instead from the primary motivators of art, the human soul and lived human experience.
Then again, our appreciation of art is often considered from the audience’s perspective rather than the artist’s intentions. Who is to say that words generated from an AI data bank cannot move a human being in some way, shape or form and guide and subsequently change their moral and emotional lives? The whole ordeal is rather tricky ground to tread upon.
To help clarify things, I recently (and cautiously) asked ChatGPT whether they thought it was possible “for AI to write a good song”. It responded, sending several shivers up my spine, that “it is certainly possible for AI to write a good song, and in fact, there have been several examples of AI-generated music that have been well-received by audiences”. However, ChatGPT admitted the fact that what is considered “good” is essentially “subjective” and depends on “individual preferences, cultural context, and historical trends”. Detailing further, it continued: “AI can be trained to analyse and understand patterns in music, including melody, harmony, rhythm, and lyrics. By using this understanding, AI algorithms can generate new compositions that are similar in style to existing music or that combine elements from different genres in innovative ways.”
While it may have appeared that ChatGPT was tooting its own French horn, it also noted the limitations of AI-generated music, appealing, no doubt, to the sentiments of Nick Cave. ChatGPT acknowledged that while AI can generate music with technical proficiency and speed beyond the capabilities of the average human, the critical factor in any work of music worth its salt is the undeniably human quality of emotion. I more than understand the great irony of asking an AI programme about its limitations, but I felt it would provide a good counterpoint to Cave’s vitriol. So, all things considered, will AI ever write a good song? Well, there are several key factors at play here, and the first is taste.
I have firmly believed that any work of art can be “good” so long as it is indeed just that. By “good”, I mean it in the classical understanding of it providing some sort of moral, spiritual and emotional growth. It should raise questions about whether the audience would have experienced that very impact had they not come into contact with the same work. In that sense, AI will likely be able to write a popular EDM track as it can analyse the rhythms and melodies that go into creating a “good” song in that genre. However, EDM is a genre that lacks emotional depth; its sole purpose is to make a good beat to dance to and nothing more. But again, the spectre of taste comes creeping in; for who am I to say that a David Guetta song cannot provide one person with the same emotional intensity and moral growth that a Balzac text can give to another?
Shelving that problematic notion for a moment, we understand that AI can indeed replicate a good song on the surface. But coming back to that classic notion of the “good”, AI is utterly unable to stand up to the task. Our favourite works of art, as Cave alludes to, come from a place of human experience. Listening to the lyrics from ‘x song’ accompanied by melody and movement greatly affects the listener. The audience understands the writer’s pain, and suddenly, they are not alone in their suffering.
This is something that AI can never attain; it will never escape the confines of the data bank, never love, never lose, and never hurt. They will never have the need to create, only the request to, ironically enough, by a living human being. As expressed earlier, whether or not AI will ever write a “good song” is challenging terrain to traverse. It largely depends on what one looks for in a “good” song, whether it be a catchy tune or lyrics with emotional depth. The former certainly can be written by an AI programme, while the latter surely cannot.
However, the fact remains that a certain degree of epistemological background knowledge is necessary to consider whether ‘x song’ in this light holds any particular level of artistic merit. For instance, should I come across ‘x song’ without knowing that it has indeed been written by an AI programme (mistakenly believing that it was humanly penned) and am moved by its lyrics and melody, then is it a “good song”? Well, it would be hard to argue that it is not, and I will likely cherish it and show my friends its wonders.
Then, one of my clued-up friends reveals the horrors of the world to me and tells me that an AI programme had written my new favourite song. How do I feel about the music now? Do I vow never to listen to it again and proclaim it now to be a “shit song”? Or do I consider how it has improved my life, brought me closer to my true self and provided comfort when the onslaught of the modern world becomes too much to bear?
To answer these ever-increasingly circular questions, I decree now that I denounce the song as the work of the devil, remove it from all my playlists, throw the physical copies at the bonfire, and take up arms against the evil conspirators that brought it into being. But somewhere in the world, when darkness falls, I lay moments from sleep in my bed, and that one-time sweet melody rings eternally in my head. A single tear falls from my eye, cupped by the curled corner of my lips, and I lament what could have been.
It does appear that the “apocalypse” is drawing ever closer, as Cave professes, at least in terms of the artistic merit of creators, when not just music, but visual art, poetry, literature, comedy, and (hell) even journalism is at the periled mercy of the ever-growing strength of AI programmes.
To remedy this, I would urge the tonic of considering the source of a given work of art prior to allowing it into one’s life. AI-generated musical works can provide good (as in pleasant to the senses) melodies we can dance to or talk over. But the real good (life-affirming) music will always be created by those who have lived, loved, (ah, fuck it) laughed and lost. The good old fashioned ‘squishy, flesh and bone, cut me, and I will bleed’ human being. The likes of which can be found below.
