Does it matter to you if a robot writes your favourite song?
When you listen to recorded music, how important is the musician? These days we rightly assume that what we listen to has been created by humans, but what if it weren’t? Could you imagine yourself enthusiastically awaiting an album produced by a robot?
For years we’ve wondered if and when artificial intelligence (AI) is going to take our jobs. The economic and social implications of automation have given journalists a lot to write about, but we rarely discuss the ramifications of AI as a creative endeavour. AI is good at replicating human decision making, and that can also include the judgments we make while we’re trying to be creative.
Recently, on The Economist’s Babbage podcast, data editor Kenneth Cukier interviewed Jon Bruner, a journalist covering AI for O’Reilly Media. You can listen to their discussion here:
Brener explained how algorithms, known as generative adversarial networks, are allowing computers to imitate examples of art by generating data based on other examples of an artistic genre. Here’s more of the AI-generated music we played during the interview:
As you can tell, it’s pretty basic and hardly very original. But we before we rush to judgment it’s important to remember that almost every human musician who has ever lived started off by producing something that was derivative of the artists they admired. The generative adversarial networks can dismantle a track and then rebuild something similar with the same musical ingredients, but it will be a long time before it can create something that’s original, never mind seminal.
The Babbage segment reminded me of “We are the Robots”, by Kraftwerk. This German band was described by Mark Fisher, a cultural critic, as the perfect example of original “futuristic” music. Fisher pointed out the dearth of progress we’ve made since Kraftwerk:
The ‘futuristic’ in music has long since ceased to refer to any future that we expect to be different; it has become an established style, much like a particular typographical font. Invited to think of the futuristic, we will still come up with something like the music of Kraftwerk, even though this is now as antique as Glenn Miller’s big band jazz was when the German group began experimenting with synthesizers in the early 1970s. Where is the 21st-century equivalent of Kraftwerk? If Kraftwerk’s music came out of a casual intolerance of the already-established, then the present moment is marked by its extraordinary accommodation towards the past. More than that, the very distinction between past and present is breaking down. In 1981, the 1960s seemed much further away than they do today. Since then, cultural time has folded back on itself, and the impression of linear development has given way to a strange simultaneity.
So yes, for now, it takes a human being to make something that creates or redefines a genre of music but is that what humans really want? If Fisher was right about musical progress, then the cold algorithmic musicians of the future could be perfect for listeners who don’t really want to be challenged by new ideas. What if human artists are already behaving like the algorithm and simply rearranging elements of the music they already like? What if what we as listeners really want is to be fed more music that sounds like the music we already know we like? If so, then computers could soon create music for us in the same way algorithms used by iTunes, YouTube and Spotify recommend music to us today (if you liked “The Smiths” then you also might like the new album by “Car Seat Headrest”). Perhaps one day we will all be listening to our own bespoke songs, generated just for us on the basis of what we enjoyed yesterday.
This blog post was written by a real human being who would love to hear your thoughts before a machine renders him redundant. You probably have an emotional attachment to certain songs and artists. But if, in years or decades to come, AI could create what you perceived to be perfect sounds, could you still have an emotional attachment? And are you impressed by the sounds of the generative adversarial networks?
Please post your comments below so we can share your thoughts and questions in an upcoming episode of Babbage. Or, if you don’t have a Medium account, you can e-mail them to firstname.lastname@example.org.
Nicholas Barrett is a social media writer at The Economist.