Alexa, get rhythm: can computers compose good music?

Saturday 2 November, 16:0017:15, Frobisher 4-6Technological futures

In February this year, Chinese tech company Huawei unveiled an attempt to complete Schubert’s unfinished Eighth Symphony – on a smartphone. The new work used artificial intelligence to analyse the completed portions of Schubert’s original, creating melodies that were selected and orchestrated by Emmy-winning composer Luke Cantor. Despite the fanfare, the finished work didn’t seem to impress those who know their Schubert.

Nonetheless, the project has triggered considerable debate about the extent to which computers might be able to generate novel artistic works. It’s not like brilliant music isn’t sometimes formulaic – Johann Sebastian Bach’s rules of harmony have been taught to music students for centuries. Perhaps it would be possible for computers to learn that there are no ‘consecutive fifths’ allowed in a successful Bach harmony, and with a little random rule-breaking produce something new? After all, one thing computers are brilliant at doing is pattern matching. Throw in some AI and who knows what could be produced?

Identifying and applying rules is one thing, but critics argue that music making takes soul – something AI is (as of yet) lacking. In a blog addressed to fans, musician Nick Cave argued that when we listen to music, ‘what we are actually listening to is human limitation and the audacity to transcend it’. How could AI understand the concept of transcendence, he asks? Then there is the social, political or historical context that influences the creation of musical genius. Wagner, Schoenberg or even Little Richard were in part influenced by the world going on around them. Would AI be able to reflect a cultural or political context in a piece of music? If we programme AI to create a particular piece of music – say, something sad – is that merely an instrumental approach to creativity?

The more practical question is do we really need AI to make music, given that so much of it is already endlessly reproducible? Are critics of AI simply expressing the same kind of hostility to creative change that tried to stop new genres of music in the past? Could computers one day produce a great symphony or opera? Or does the possibility of computer-generated art go too far towards undercutting one of the ways in which we can still consider ourselves unique? All of this begs the question: what makes music, music?