Mr Nikrang, how musicality does AI have by now?
(Laughs) There’s no simple answer to that. Even we humans don’t really know what qualities certain sequences of note have to have to be perceived as music. All cultures have music; it is intrinsically human, but we still don’t understand how and why it works. So if a machine is capable of composing music, my motivation is to investigate this machine to see what it learns from it.
What are you working on at the Munich university of Music and Theatre?
I’m basically involved in two different areas: On the one hand, I am developing the AI-based Ricercar system, with which students can compose. Ricercar is trained with a broad classical repertoire of public domain works from the Renaissance to the first half of the 20th century. And then there is the data the AI composes itself. With a few tricks, you can get it to learn from your own pieces and to get better all the time. The other, equally important, question is: Why use artificial intelligence from an artistic perspective?
And what is your answer?
We know that AI systems have great creative potential. The history of art shows us that new technologies have often yielded novel formats and ideas. Modern AI is a bridge technology that gives us completely new possibilities: it enables interaction with the audience, can visualise music and is better able to convey content. We can mix and match all of these components as it suits us, and even people without programming experience can use the technology. But it does require close cooperation between art and technology. In joint projects, we must find out what is possible and what makes sense from an artistic point of view.
How can we imagine this in practical terms?
In collaboration with the Munich Philharmonic and the Munich-based company Brainlab AG, for example, composition students at the HMTM have explored our AI and written a concerto for a self-playing piano and orchestra. The piano played what the AI had composed, while the orchestra played the part of the composition written by humans. Our aim was to address a range of questions: What does it mean to work with this technology? What is the role of humans? People think that machines do whatever we want them to do. But AI is a creative system with a certain degree of autonomy. You can’t tell it exactly what the result should be. This often leads to interesting situations where we fight with the AI, negotiate with it, consciously manipulate it to create something that corresponds to our own artistic ideas.
Does this mean that AI cannot do without humans?
That depends on what you use it for. For functional music, for a game or for advertising, it is perhaps not essential for music to have a personal touch. That’s where AI can save us time because it is fast and productive. But this carries the risk of driving down prices and replacing people with technology. Our focus is on how to use AI to create an individual work of art. One that is made by humans. Because AI is not capable of producing anything like that. Moreover, it would never feel the need to create something this personal. It is, after all, a lifeless machine; but we humans can utilise the full potential of AI. The more personal my work with AI, the more indispensable I become.
What have you learnt from AI?
It changed my perspective on music a great deal. When you pay attention to what AI learns quickly and where it has problems, that changes your own perception of music. I see AI as an exploration system. It is trained with masses of data and learns to solve a specific task. It represents a multidimensional space in which all the data points are represented, much like in our brain. When I compose, I move within this space. In a nutshell, composing in this way is like a walk in the woods: AI would never feel the need to go for a walk, but I, as a human being, want to get to a certain place in the woods. So I have to get the AI to take me there. Based on mere chance, it makes various suggestions and I pick my path along these.
What role will AI play for musicians in the future?
I believe we are dealing with a generation of artists for whom AI will be a matter of course. Whether they actually use AI to create is another matter, but it will play an important role. If you want to come up with individual creations, you shouldn’t be afraid to customise the systems. I want to convey AI in a way that emphasises this individuality. Active involvement of the art community in its development is very important here. AI is not something predetermined and static. It should be regarded as an open system, whose parts can be changed at any time. Humans remain at the core, navigating and making decisions.
Text: Maria Zimmerer
