Prof. Mark Gotham Involved in the Completion of Beethoven’s 10th Symphony

In his research, Prof. Mark Gotham develops computer-aided methods that can expand knowledge about musical structures. By appointing him, TU Dortmund University has brought on board its very first musician capable of cooperating with the Department of Computer Science in the field of artificial intelligence (AI). He demonstrated his capabilities by contributing to the completion of Beethoven’s 10th Symphony.

AI project with the unfinished 10th Symphony

When Ludwig van Beethoven died in 1827, he left behind nine complete symphonies, but also one that was incomplete. Only the initial fragments of his 10th Symphony have been found. How would Beethoven’s 10th Symphony have sounded if he had managed to finish it? An AI project at Deutsche Telekom has been looking into this question since 2019. A team of ten international experts from the fields of AI and musicology developed an artificial intelligence system that can “understand” Beethoven’s style.

“The project raises a lot of interesting questions, particularly about the nature of the possible interactions between humans and computers,” says Prof. Mark Gotham, whose role within the project is to combine AI with music. “How can works be completed? Beethoven’s 10th Symphony offers fertile ground for experimenting with such a task,” he adds. There is only a small number of surviving drafts, but these are enough to provide valuable starting points for the project, but they are so sparse that most scholars have come to the conclusion that the work cannot be completed through traditional means. This is where artificial intelligence comes into play.

Teaching AI with around 10,000 pieces of music

In order for the AI to “think” like Beethoven, it requires a lot of data. Since the start of the project, the developers have taught the AI using around 10,000 pieces of music. They used compositions and notes by Beethoven, but also works by musicians and composers who can be proven to have inspired and influenced Beethoven during his lifetime, such as Johann Sebastian Bach. This made it possible for the AI to continue writing – and ultimately complete – Beethoven’s 10th Symphony by using an algorithm and interacting with the experts.

With a computerized project such as this, there are many musical decisions to be made. What kind of music should be generated? What examples are relevant for AI to learn? It is not all about interpretation, flexibility and creativity, as important as these aspects are. It is also necessary to incorporate the computational aspect of human musical knowledge – because developing as a musician involves learning about recurring patterns by dealing with existing works. “For me, finding satisfactory answers to these questions is a very interesting and rewarding challenge,” says Prof. Mark Gotham.