Return to site

How Machine-Made Music Challenges Our Ideas of Art

Radical new developments in music have always courted controversy, whether it’s the semi-legendary tales of rioting at the premiere of Stravinsky’s Rites of Spring or the mixture of bafflement and awe that greeted John Cage’s 4’33’’. Above all else, however, one area of debate continues to rage – the slow integration of modern technology into the way that we create and produce music. From Karlheinz Stockhausen’s pioneering electronic compositions to a new generation of AI-assisted music scoring software, does machine-made music challenge our very idea of art?

A Brief History of Machine-Made Music

Just as the invention of the printing press revolutionised the production of music notation, so did the slow birth of the computer age throughout the twentieth century.

The earliest known recording of a computer-generated composition can be traced to the father of modern computer science artificial intelligence – Alan Turing. In 1951, 11 years after Turing’s work breaking the enigma code in WWII, a BBC outside-broadcast team recorded one of his room-filling early computers as it was programmed to play a crude, mechanical version of God Save the Queen.

The acceptance of computers into the creation of popular music, whether through the ground-breaking TONTO synthesiser or Kraftwerk’s appearance on the BBC’s science and technology programme Tomorrow’s World, has been relatively uneventful. Among some classical musicians, however, many equivalent developments have been met with a mixture of suspicion and hostility – an unease that has only increased as advances in AI have made music produced entirely by computers a very real possibility.

Today’s Music-Making Machines

The latest developments in this debate have centred around machine learning– the use of deep learning algorithms that use inputted data to find patterns and then predict outcomes. In the case of music, this could well mean analysing existing compositions to create entirely new ones– all without the guidance of a human hand.

Google’s open-source Magenta may be the most high-profile of the machine learning music projects, but they’re far from alone, with competition from the jazz producing algorithm DeepJazz and Cambridge’s BachBot. While these projects are at varying levels of progress, algorithmically-produced music is far from hypothetical – the French research team behind FlowMachine have already unveiled an AI-created track which sounds disarmingly like a competent Beatles pastiche.

What Does This Mean for Music as Art?

In 1935, the philosopher Walter Benjamin wrote that a mechanical reproduction of a work of art was lacking in the original’s ‘aura’ – its unique location in space and time. Far from bemoaning this development, however, Benjamin saw in it a radical potential to remove the authority of the aura, so often heavily guarded by a privileged few, and democratise access to art. As we find ourselves faced with an age of mechanical production, could technology actually help, rather than hinder, musicians?

Freya Murray of the Google Arts & Culture Lab has argued that no matter how advanced her or her competitor’s creations become, their lack of a human touch or intuition will always limit them to assisting, rather than replacing, composers. Instead, music-producing technology will become a tool to make the production of music easier and more accessible. Outside of the computer lab, programs for composing music have been around since the emergence of the Sibelius scorewriter in the early 1990s – now the team behind Sibelius have built on their earlier version to create the music notation software Dorico, which offers the same quality of notation and engraving as the finest traditional methods. In this case the technology optimises the existing talent of its user, rather than attempting to emulate it.

Historically it has not been technology itself which has threatened music, but its application in the service of what Theodor Adorno dubbed ‘the culture industry’ – artistic expression as a mass-produced commodity. The new generation of music technology could just as easily provide talented musicians with the tools to deliver a higher quality and quantity of work. The developers of Dorico compare the AI assisted elements of their music notation software to a journey in a driverless car – the final destination remains the same, but the process of getting there is transformed utterly.