Artificial intelligence is becoming increasingly prevalent in our everyday lives, including the field of art. With this post, I want to explore the question of whether there is a concern that AI could take over my job as a composer and producer. Additionally I would like to share my thoughts on this matter and address the question of whether these advancements make me fearful.
What is AI?
What is this AI that everyone is talking about? AI stands for artificial intelligence and its usage may often be misunderstood. Please forgive me for continuing to use "AI" throughout this post. Because when you think of movies and fiction, you might quickly associate AI with robots that destroy humanity. So many people may think of it as machines that can completely think for themselves.
However, through friends who are active in the field of Cognitive Science or Computer Science, I have come to understand the term "AI" in a slightly more concrete way. The term Machine Learning is a crucial factor in this regard. It is, so to speak, a field of artificial intelligence. In the article "What is Machine Learning?" by IBM, the different areas of AI are well categorized and explained. Ultimately, patterns are consistently learned.
What is my job?
When I compose or produce music, I consciously or (in my belief) subconsciously rely on patterns. I often see the creative aspect of my work solely in combining the right elements to achieve the desired outcome of a piece. Therefore, it is important to recognize what the goal should be, which features should be emphasized, and which musical tools could be used to realize these features. Additionally, there is always room for personal taste, but ultimately, I consider myself someone who recombines patterns in terms of creativity.
This may sound somewhat uninspiring and could paint a rather sobering picture of art for some. It is currently my belief that art always embraces familiar patterns and recombines them.
After the creative process, there is also the craftsmanship process for me, such as producing the compositional idea. In this realm, I find there is even less creativity involved. As a creator, there are many more conscious automatisms that I have to repeatedly execute.
What is AI already capable of?
In the field of music, there are already systems that can compose music. At Aiva.ai, for example, you can generate new melodies in combination with harmonies directly in the browser of your online account, and they will be played back by virtual instruments. There is also a plugin (a software extension) called OrbComposer available for music production software, which can generate new musical content. For sound synthesis, Google has developed the NSynth project, where a synthesizer can create entirely new sounds using AI.
Creating images using models such as Stable Diffusion has become more well-known. You prompt a text describing how the image should look, and after a rendering time, you receive a result. I have personally used the latter, for example, to generate one or another album cover.
Does it scare me?
Now, let's get to the crucial question of whether this scares me or if I see it as a threat to my job.
Certain aspects of my job (the more creative part, not the bureaucratic side - that part is even more boring!) involve recurring processes. Processes that could be automated for sure. In this area I already see great potential for AI to take over these processes for me. It would mean that I could invest even more time and energy into the truly important and interesting aspects.
My job could change
Let's assume that AI will eventually reach a point where it can take over these more interesting tasks from me. In turn, that would mean that my job would simply undergo even more transformation. Perhaps I would become more of a coordinator. Instead of specifically composing music, I might coordinate text prompts and make a suitable selection from the outputs of a machine. The question then would be, what qualifications would be required for selecting the outputs and choosing the prompts? Would it be beneficial for a project that now requires music generation if I intuitively select what to use? Or would I need to have a more reflective understanding of how I assess music quality and base my selection on that knowledge? In this case, I would still be a specialist who possesses a certain qualification as a prerequisite for this job. Otherwise the process could be handled by the customers themselves. Based on my experience, I am already a human interface between customers and music. Often times it is a process of communication between the customers and me to find the perfect music production that aligns with the project as closely as possible. The question of how and whether this communication process between my current customers and a future machine could take place remains open.
Perhaps I would also start programming certain algorithms myself or training models. If access to it were easy for me, I see it as an opportunity to learn new things and gain new experiences. If access, however, were difficult, I would need to undergo further training or make more extensive changes to my job. I find a conservative view that everything should stay the same to be rather counterproductive when it comes to change.
Ultimately I often see significant opportunities in increasingly advanced technology. For me it usually involves eliminating processes that humans probably don't enjoy anyway. So why keep these processes alive?
The fun factor
Another crucial aspect, in my opinion, is the ability to have fun while making or creating music. Even if a machine can eventually do the work that I have always done, it doesn't mean that I would stop doing that work. Music is a perfect example of this. Why do people make music, such as performing live together? There are already digital tools available to create music using computers. And sometimes there are even way too many of them. Yet, people make music because it brings them joy!
In this regard, there are two aspects that come to my mind. Firstly, I enjoy exploring new things. While testing and playing around with Stable Diffusion, I noticed how interesting it was for me to make slight modifications to the prompts and receive different outputs from the AI. It made me feel like an explorer, searching for fascinating things in a little new world. Personally I found it enjoyable and it satisfied my curiosity to some extent. Similarly, when producing music, I might experiment with a synthesizer to discover new sounds for myself. I also treat musical ideas as thought experiments, trying to bring them to life and hear them because sometimes I couldn't have imagined them otherwise. The second aspect is interpersonal connections, which are present when making music together, such as performing live with others. Perhaps during rehearsals, we make a mistake that creates a funny sound, and we all laugh together. Maybe we can identify areas for improvement in each other's performances and offer suggestions for better playing, and we may receive tips ourselves. It can also happen that the situation inspires us to engage in conversations about unrelated topics.
Therefore, I find it highly unlikely that people will stop making music themselves just because a machine can now do it too. At least for now, I can't imagine sharing laughter or having a conversation with a machine. And ultimately, I will probably still want to explore new things regardless.
Costs for customers
"But what if a machine can create music faster and cheaper for customers?" one might ask. In that case, I would recommend this machine if it truly delivers more suitable results for potential customers! It would be hypocritical of me to want to provide optimal results to the customer on one hand, but then withhold a better option just because I wouldn't get that job myself. Moreover, I believe that there will perhaps always be a certain individual tone (quite literally) when it comes to musicians. I trust that musicians are commissioned because the customer likes their style and wants their sound specifically. Reality (also based on my own experience) often differs because the offered price can unfortunately determine whether one gets a job or not. However, I find a diversity of sound to be more optimal, as it can offer customers and audiences so much more. In such a case, the customer can then decide whether the "special something" should be given more weight to the human musician or not - and I would find both scenarios legitimate.
Music in transition
If music could be produced faster and cheaper, it would undoubtedly accelerate the pace of change in music itself. New genres would emerge more quickly. For instance, at some point, I could simply input as text "riff of a metal band played by nine harps, a tuba, a synthesizer, and a jazz drum kit" and immediately invent the Jazztalchestra-Dance-Music genre, or have it created for me.
Generated music might eventually serve as a reflection of society. Once highly praised artists who were played on the radio over and over again would lose their charm, as anyone could generate similar music using simple text prompts. Or perhaps every radio station would become a single generative output, and one could choose a particular station based on personal taste and mood? Would that be a good thing? Personally, I would appreciate a shift towards reflecting on art rather than idolizing artists.
11.08.2023: Suitable to the topic I found the article "KI kills the Radio Star" on the german news site taz.