in Columns

Composers! Artificial Intelligence is Your Friend

by Uljas Pulkkis

Your friends on social media are not your friends; they are just people who are notified of updates to your account. Similarly, artificial intelligence is not intelligence; the term is a label for computational models of inductive reasoning.

What artificial intelligence (AI) is lacking is a feature of our brains that is crucial for intelligence: intuition, as visionary Erik J. Larson writes in his book The myth of AI. AI as we know it today will never gain consciousness and will certainly never threaten jobs in the creative industries. This is because even the very best of neural networks are unable to imitate how the human brain actually works. The reason for this is spelled out in Incognito by neuroscientist David Eagleman: no one knows how the human brain works, but it certainly does not work like artificial intelligence.

On a personal note, I constantly hear comments about my work: a computer can already write symphonies, are you not worried? An apposite response, paraphrasing Fermi’s paradox, is: if computers can write masterpieces, then where are they? They have produced artworks, yes, but what about producing excellent artworks? It takes a composer to guide the creative process of the machine in order to create artworks of value. After all, mathematicians did not find themselves out of work just because a computer can now perform the job of billions of mathematicians in one second; they were simply able to move on to more interesting challenges.

Artificial intelligence may not be intelligent, but it is a composer’s friend.

In composing music, you are faced with a multitude of choices all the time: pitches, orchestration, harmony, form, and so on. In a single orchestral work, there are so many parameters to consider and control that it is impossible to focus on all of them with equal intensity. There just is not enough time. This is why, in programme notes, we find things like “I focused on form in this piece” or “my principal interest here was orchestration”.

Artificial intelligence is shaping up to be an excellent assistant in managing this panoply of parameters. AI can sift out the best solutions on the composer’s behalf, even for challenges that are not in focus for the composer at that particular moment. The composer then has the task of selecting the best options for further development.

To take an example: when exploring ideas for orchestration, you find yourself leafing through an enormous number of scores from whose orchestration you wish to take inspiration. You may end up using a solution similar to something that someone else has already done but adding to it and adapting it to your particular purpose – in other words, doing what composers have been doing for centuries. The thing is, it can take days to solve just one orchestration issue in this way. So why not automate the reviewing of scores to ensure the best possible end result? Or to put it another way, why not harness AI to be an extension of your composer persona? This means having AI perform the inductive part of the process (deriving rules from data) but retaining control of the intuitive part (selecting the best options for further processing).

Of course, there are many parameters for which AI offers no help. For instance, choices regarding the large-scale structure of a work are difficult to extract by mining big data, because such choices are intuition-driven and depend on what the work as a whole seems to require. This requirement, imposed by the music itself on the composer, is something that is beyond the reach of AI. It has to do with how the human brain operates, the mystery of our consciousness and our understanding of what we believe sounds good and effective. These insights also include the fleeting moments that I enjoy in my favourite pieces of music, such as the sudden replacement of piccolos with double bass rumbling in Atmosphères by Ligeti or the whimsical repetition of the letters M–A–X in Knussen’s opera Where the Wild Things Are. AI is unable to model why these moments are memorable, because AI does not have a heart that leaps up at such things. There is, very evidently, something indescribable about musical compositions, particularly masterpieces.

So what is it? Philosopher Jim Holt has spent years of his life to find an answer to the question “Why is anything?” In his book Why does the world exist? he presents answers that to my mind apply well to the existence of music and compositions. Holt reaches a conclusion with two options. The first one is that everything exists by chance. It may well be that musical creativity is simply the product of the quantum foam of our minds, random flashes in the universe. This does not, however, explain the magic of a masterpiece.

Holt’s second option is that everything begins with goodness. This does not mean goodness of the divine sort, but of the Platonic sort: the world tends to organise itself according to what the best possible option is at any given time. Similarly, composers strive for goodness in their compositions, i.e. to create the best possible work out of all available options. Whether a composition is good is a collective experience that emerges in our brain during a performance of it and manifests itself particularly in those moments that we especially love, but also in the demands that the music itself makes of its composer. We can say with certainty that AI is not capable of goodness but a composer’s intuition definitely is.

The only thing that frightens me about AI is its reckless marketing. Mathematician Cathy O’Neill warns in her book Weapons of Math Destruction that we should not rely overmuch on the potential of AI. The greatest threat is not that AI will end up making decisions better than humans but that we will come to believe that the decisions made by AI are better than they actually are and surrender power to AI voluntarily.

Until someone makes the mistake of advocating for human rights for AI, I intend to use AI as my friend and a partner in the composition process. I never have to worry about its stamina or mood, and I can devote my time to compositional issues that are more interesting and more challenging. AI can thus help me and my colleagues create new masterpieces of art music that someone like Boulez or Stravinsky could never have dreamed of. What a time to be alive!

Featured photo: Uljas Pulkkis

Translation: Jaakko Mäntyjärvi