We’d like to think creativity is the last uniquely human skill, but if developments in the arts are anything to go by, we may need to reconsider.
In 1906, the composer John Philip Sousa decried the invention of technology for recording music, such as the gramophone, in an essay entitled "The Menace of Mechanical Music". Perhaps somewhat selfishly, because composers were poorly compensated for recorded music at the time, he posited that it offered no substitute for a live performance.
Even more relevantly, he cast his mind back to centuries prior when attempts to turn music into a science had brought about the mathematisation of composition. Human algorithms, as it were, for creating music.
Could you tell the difference between a song written by a human and a machine? You might think music written by algorithms would be unmistakable for anything else – until you’ve heard it.
In an episode of the "Switched on Pop" podcast, the show’s host, musicology professor Nate Sloan, was challenged to differentiate between AI and human compositions in a musical Turing test. Spoiler alert: he failed. And (for any music snobs) it wasn’t just pop – Sloan also struggled to tell the difference between Beethoven and computer-generated classical music.
Yet, despite their similarities, most music fans will say they want AI to have no part in the music they consume. Something just feels wrong about machines in creative domains such as the arts.
In 2016 I saw the musical Beyond the Fence, written from plot to score by AI. It was average, at best, even as a curiosity. But, pardon the pun, it set the stage for the future of AI in the arts.
In the same way that AI can be fed scores by Beethoven and generate something that is Beethoven-esque enough to fool a music professor, the AI involved in the musical was fed a corpus of data, including other musical scores and libretti. It then "learned" from them how to generate something novel.
This level of AI is readily available within existing music software. So there’s a chance you’re already listening to music in some ways algorithmically generated. Whether made by man or machine, much of today’s music is far from the traditional singersongwriter model.
If you look at a modern piece of music’s credits, you’re likely to see upwards of half a dozen names. Software gets no such credit, but its fingerprints are already all over today’s hits.
So are we being oldfashioned in our distaste for AI-generated music? Maybe we’re just continuing to be unaccepting of new technologies, just as 20th-century composers hated the arrival of "mechanical" music.
Speaking to musicians and fans at The Great Escape festival in Brighton – which featured a day dedicated to AI in music as part of its industry conference this year – the same words cropped up when talking both about AI-assisted music and "pop factory" chart hits: inauthentic, lacking soul, contrived. These aren’t criticisms of their musical qualities so much as the creation process.
You could say the worst of human-created pop is made to pay bills and be passed off as the performer’s work. Hardly how we like to imagine the creative process. Whether made by AI or a team of hitmakers, this type of music lacks authenticity.
Like the notion of the "uncanny valley" (when we’re creeped out by artificial human faces that look almost, but not quite, real), there is perhaps a feeling of "uncanny melody" when we know AI is involved in composing a song.
It hits us hard that one of the things that makes us human, creativity, is being taken away from us. But unaware of this, we feel the same connection between us and the artist that has made music the engagement art form it always has been.
Meanwhile, in the world of commercial creativity, perhaps we should just accept we will be collaborating with AI to make advertising and figure out on how we will do so. Because, just like the hit parade, most consumers don’t care – they just like what they like.
Thomas Scovell is the executive strategy director at AnalogFolk London.