Four Ways in Which the Robot is Surpassing the Human Composer


Music caters to people. Yet, does it also have to be created by people?

1. Humans are inclined to listen to human or even machine performers, less to connect with whoever organised the music

Listeners who are willing to hear current music should be motivated to listen because of the qualities that make it interesting and functioning music, not for the fact that it was made by a living composer, a person you know, or even by a human. Recent technological advances in artificial intelligence might find their way into several previously human enterprises, and perhaps music will not be spared from this. We all enjoy performing or listening to musicians perform, although we have in a sense delegated performance to machines already – I for one prefer hearing recordings to live performances. It happens with both that I sometimes listen inattentively, or miss details due to extraneous noise, yet with high quality recordings I can immerse in the moments I want.

What I see for the future is the next step of this automatisation tendency. If an audience is to appreciate music at all, it will not feel entitled to seeing a composer-demigod as part of the concert etiquette, but only to hearing good music. This is where composition machinery steps on stage, or rather, on the backstage. Different than a robotic vehicle lacking a network of crucial and complex human reflexes, when this machinery gets enough time to do its job and learns from feedback to its mistakes, the resulting compositions will become convincing. Since nobody can credibly claim ownership of machine-created compositions, the exchange of musical material will not be hindered by posthumous copyright or other bizarre considerations of possession or consumption. Likewise, the artificial distinction between folk and art music shows us that during tens of thousands of years of music making, humans have dealt comfortably with a blurred authorship.

2. Automation has kicked off long ago and might flood us with new compositions faster than ever

In many ongoing branches of automation and robotisation, development of artificial intelligence relies on the assumption that each creative algorithm in the human brain can be located, broken down and emulated inside circuits. Such early automation in notation software is already complementing human errors and deficiencies, even if a human has to be steering the computer’s actions and thus workforce is not completely freed.

Something to look for in the next generation of notation and composition software is a feature that learns and intelligently continues the music already typed in by the composer. This is based on the notion that new computer intelligence is capable of learning from mistakes and monitoring highly individualised rules and preferences of notation in supervised environments, much like an apprentice of a busy composer in the olden days.

If and when computer-generated music can no more be discerned from the best of human composers’ works, and when this quality and activity starts to surge, the costs of composing will drop, as will the amount of available grants. New high-quality works can be churned out at musicians at an alarming rate. That might counterintuitively soon account for an unwarranted inflation of and disengagement with any new music.
With immersion in psychoacoustics and brain scanning, perhaps even personally tailored listening experiences could save music from becoming omnipresent and thus impersonal. If new music fails to implement personalisation at such a stage, composed music (by humans or machines alike) had better disappear completely.

3. Within the revolution of employment in the arts, music will come out more diverse

During the next two decades as many as 35 percent of jobs, estimates vary, might vanish in favour of robotisation. [] Jobs that more directly address issues peculiariar to humanity will prevail; also humans in jobs in which communication skills and a mindset of consultation, sharing, as well as originality of thinking are needed. Only some of these apply to conventional human composing. Knowing the age-old myth of composers equaling male hermits, all of these traits also interestingly happen to lie decisively outside what has been called hegemonic masculinity. To comprehend this, one need only think how relatively few human composers from any minority group have been and will be given the early opportunities and representation to develop to be good composers. During a stage of human–machine transition, such monolithic concepts of composer geniuses will crumble while composers other than light-skinned middle-aged cis-males are gradually being presented and taken more seriously. Machines are a threat to composers’ identity politics, since machines don’t buy into notions of fixed compositional style, nationality or gender.

In short, depersonalisation of composition is a cause for everyone who wishes the most surprising and positive results for music and not necessarily for the continuity of composership as it is known today. We living composers have to prepare for this distant but possible scenario.

4 a. It is cost-efficient to have robots compose for us and refine themselves while at it

The conventional route of bringing a composer through intense education or trial-and-error to artistic maturity and social significance is a costly enterprise. The limitations of a composer’s brain can be overcome in more traditional ways, not just with computer-aided composition. Just like the early works in the Western(ised) classical tradition were created and refined in simple collectives of chorister-composers, more and more composers seem to have access to such fruitful collaboration – even though they seldom have previous hands-on expertise on even all common instruments, of which there now are simply too many to master. Is any number of hours of musical experience worth paying for if there were a miraculous way to teach all of it to a composer robot in a minute?

….on the other hand… 4 b. There’s still hope for human composers at least as maintenance staff for composer robots

Future people deserve the music that helps their lives in the best possible way. To create an enjoyable, functional or effective piece of music, one should no more rely on just one brain, however genius it may seem. The interpretation and delivery of that work is still best left to human musicians, though. Attempts at entire computer-generated compositions have thus far been either uninteresting or disturbing to hear for that reason, especially in the absence of human performers. We will perhaps never know even how to bridge the perceptual gap between electronic music and the acoustic one.

Despite point 1, humans are looking for at least a tiny fingerprint of humanity speaking back to them, even if the mediating score was created by a machine. If the machine cannot feign that, audiences will revert to human composers.

I clearly don’t encourage robotisation of composition at the cost of more relevant fields such as surgery, genetics, environment, rescue, interconnectivity, transport, or food production. Evolution has made human brains quite the optimal apparatus for creating art (and teaching it to robots), after all. By any account, entire or partial robotisation in the field of the arts will happen in the distant future, if at all.

Meanwhile, all fellow composers should keep up the good work surpassing computer-created music – we might be living the last of viable times.

Juhani VESIKKALA (Helsinki, FI)

Composer etc., secretary of Ears Open.