By James Hannigan

“The enemy of art is the absence of limitations”

ORSON WELLES

The confluence of music and technology is nothing new; even standard orchestral instruments are historical artefacts of technological innovation. As a composer who has weathered a few minor paradigm shifts, I find myself anxious about the big one lying ahead. This time around, I sense more may be at stake than only the livelihoods of artists: some of our fundamental beliefs about the purpose of music, its personal nature, and the role it plays in our lives might undergo a transformation.

Early in my career, I branched into relatively uncharted musical territory: devising interactive music systems for video games after the introduction of digital audio and streaming.  This was a time devoid of ground rules and textbooks, with little in the way of formalised terminology or any theoretical foundation, and it felt as though I was making things up as I went along. 

Today, tools of this kind have become so familiar that their origins as inventions are long forgotten. Off-the-shelf solutions are plentiful and educational courses abound, but I can remember when simply grasping the principles of interactive music posed challenges, particularly for those rooted in the classical tradition or linear media.

One project from over twenty years ago, the BAFTA-nominated Republic: The Revolution, sticks in my mind, having afforded me considerable free rein to explore my ideas.  The game’s executive designer would later go on to co-found Google-owned AI lab, DeepMind – recently sharing in, no less, the 2024 Nobel Prize for Chemistry.

The apprehension I felt all those years ago, I suspect, isn’t unlike that experienced by many after the introduction of desktop music technology in the 1980s and 1990s. Initially daunting, sequencers, synthesisers, and hard disk recorders quickly became mainstream in music, and have since been lauded for the role they’ve played in democratising production, contributing to massive industry expansion.

Such change signified a victory for the tech-savvy musician but brought devastation for some: ‘Pen and paper’ composers and orchestrators – once dominant in film and television – were among those hit by the advent of MIDI and computer-based production, and their numbers have dwindled dramatically since the 1980s. 

Unsurprisingly, this shift was met with little public outcry at the time, not unlike the subdued reaction to the decline of professional photography after the adoption of digital cameras and smartphones.  Indeed, such advances appear to have been largely welcomed by many, viewed as having created opportunities for millions to pursue creative endeavours previously thought of as wildly inaccessible.  

Creative destruction – the process by which a technological shift opens more doors than it closes – inevitably produces winners and losers, and this is a cycle as old as innovation itself.   But will AI Music come to represent a paradigm shift too far for composers?  Are we looking at creative destruction this time around, or something more akin to total annihilation?

To get an inkling of what may lie ahead, it might be useful to look at how composers are perceived in the public imagination, both historically and in the modern day.

Edvard Grieg at the piano.

Despite widespread and persistent technologisation in music, the public’s perception of the composer – romantically depicted in Western popular culture and by classical music outlets – remains that of a stern, wild-haired maestro seated at a grand piano or feverishly sprawled across a desk, manuscript paper and quill in hand. Such figures are often seen as suffering for their art, agonising in their quest to express an intensely personal musical vision.

Conductors like Leonard Bernstein and Simon Rattle, known for their dramatic conducting styles over many decades, have reinforced this image – embodying the passion of the concert hall and becoming the public face of classical music for millions.

So ingrained is the archetype of the Western composer – who is, at the very least, a sentient being, and at most, a tortured and deified genius – that we continue to recount their stories, often over centuries, endlessly examining the inspiration behind their works. 

For instance: audiences for classical music remain fascinated by Mozart’s lack of recognition during his short, tumultuous life, and continue to reflect on Shostakovich’s artistic constraints within the Soviet Union.  And how many of us think of Beethoven’s music without contemplating the profound impact of his deafness? 

Artists’ stories and their emotional and intellectual responses to their circumstances appear integral to our appreciation of them, to the extent that it can become undesirable to separate artists from their art. We tend to romanticise such figures, requiring knowledge of their inner lives in order to process their creations, often looking towards experts and reviewers to validate and curate works for us.  How else, I wonder, can we navigate the immense sea of human creativity?

This principle isn’t confined to classical music. In pop music, too, a song isn’t created in a vacuum; it is experienced through the lens of an artist’s life, within the context of social change or a broader cultural movement.

The structure and duration of songs have stayed fairly consistent for sixty years or more – in part determined by the length of a 45 RPM single – and countless songs share similar motifs, chord progressions and production techniques, yet other mysterious factors appear to contribute to their ongoing appeal and cultural resonance. It would appear that there is no reliable algorithm for cool.

This perspective, I believe, is essential to our understanding and appreciation of music; part of the very way we perceive and ascribe value to it. It highlights the significance of a musician’s persona and the intrinsic motivation behind their works. 

While composers may be less visible to the public than performers, they are still widely regarded as the primary architects of music – the minds behind the musical blueprints, so to speak.

Towards the start of my career, the title of “composer” was laden with gravitas, and over a decade passed – and numerous projects – before I felt able to even present myself as one. I’d grown up believing that this particular job title, much like that of “writer,” had to be bestowed upon me by some musical institution or another before I could rightfully own it.

In stark contrast, in today’s marketplace, becoming a composer requires little more than self-identifying as one, acquiring some rudimentary music production tools and declaring an intention to write music. While this is a victory for the democratisation of music-making, I fear that AI may push this trend to potentially absurd extremes.

A similar phenomenon occurs among writers: how often do we encounter people who claim to have a novel inside them, even though such a novel rarely ever materialises? With tools like ChatGPT, bringing that elusive novel to life may now be a reality for many, but isn’t doing things that way missing the point of the exercise?  Much like climbing a mountain, isn’t the ascent itself the very reason for doing it? 

Using generative AI to create a supposed work of art feels, to me, much like teleporting a mountaineer to the summit of Everest: where is the satisfaction in doing so? Why would anyone care about a goal so easily attained?

I’ve always favoured a can-do culture and the notion of a meritocracy open to everyone, yet in the face of AI, even I now find myself wondering where, in future, we might look to draw the line in determining who – or what – qualifies as a real artist. Is such a distinction even possible anymore?

With the advent of AI music models, are we to begin considering individuals as composers based on their ability to craft effective text prompts for generative AI music engines? 

This notion is already being embraced by some who unhesitatingly equate their roles as prompt engineers to those of existing composers and artists. AI music tools such as Suno and Udio ostensibly claim to “democratise” music-making but, as far as I can see, appear to carry out all of the creative work on behalf of users.

Similarly, there are prompt engineers in the realm of the visual arts marketing themselves as the peers of seasoned professionals, with little or no regard for the enormous dedication, skill and craft such artists have cultivated over many years.  Is this really what society as a whole now envisions for its artists? 

Serious concerns exist, too, about how generative AI models may have been trained using copyrighted works, with a lack of transparency surrounding these processes – particularly when resulting outputs end up being commercially exploited. The issue remains unresolved and has, sadly, mostly been overlooked by the media. Artists of all types are growing increasingly desperate, as indicated by a recent statement from composer and Fairly Trained CEO Ed Newton-Rex, signed by tens of thousands of artists rejecting the use of AI in this way.

It would seem that the real and perceived benefits of AI – in fields such as medicine and healthcare, along with AI’s huge potential economic impact – have blinded governments and much of the public from considering the legitimate concerns of artists – even though, in 2022, the creative industries alone contributed £126bn in value to the UK economy.

Let us not forget: without human innovation and the impulse to create, there would be no concept of music, nor would there be anything for AI music to model or emulate. Shouldn’t we, therefore, pause and reconsider before handing over control of this precious human invention to legions of prompt engineers and machines? 

The quality of generative AI’s output largely depends on human-made input, yet very few major tech companies appear willing to even acknowledge AI’s debt to humanity for creating the foundations of personal creativity that they seek to mimic and automate. Not even so much as a “thank you” in many cases.

To many artists today – and by “artist,” I mean those who were motivated enough to actually create works before the advent of AI tools that apparently automate the process – it feels as though cherished, intrinsically human art forms are being prised from them and handed over to tech bros – often for profit and under the guise of democratising creativity for all.

I don’t buy the democratisation narrative – not yet at least – but can see how ingenious a marketing ploy it is. It allows subscribers to indulge their fantasies of becoming the artists they’ve always wanted to be, while simultaneously lining the pockets of tech companies at the expense of human creators. The only people who lose out in this equation are the real, working artists.

If this attitude towards creativity is inevitable for our society in the long run – which it very well may be – then surely, at the very least, the works used as training data should be licensed.

SOME DIFFERENCES BETWEEN AI AND HUMAN-MADE MUSIC

Differentiating the various potential niches for AI deployment in music could, I believe, hinge on understanding their primary musical objectives. After all, not all music seeks to provide deep insights into the human condition; sometimes, simply being fun, relaxing, exciting, danceable, or merely interesting will suffice for many listeners.

Clearly, not all music serves the same purpose.

This variance leads me to be fairly optimistic about the coexistence of AI-generated and human-created music in the early days of AI Music, albeit with some overlap and the blurring of boundaries here and there.

In the long term, however, if intensely personal, human-centric music – as many of us conceive of it today – is to retain its value, surely it will be necessary for it to continue being written and performed by actual human beings – in ways transcending any mere computational or statistical exercise. 

Some – admittedly not all – technologists appear to consider the role of prompt engineer sufficient for generating meaningful music. But here – just as with the less-than-convincing democratisation narrative – the cynic in me questions the depth of this belief: Is it genuinely held or merely a superficial claim?  A token gesture to rationalise and morally sanction the invention of AI music tools by overstating the true extent to which humans will actually be involved in the creative process? 

In the face of already dwindling human involvement in music making over the last few decades – from automated music production tools to the increasingly infrequent use of living musicians in favour of samples – I find myself wondering: just how much human involvement in music, and what type, is enough to authentically capture the personal experience of human beings? 

A first step in answering this may be to look at the commonalities of human and AI-made music, because some appear to exist, and then to consider a few differences between them.

Much of the music composed in any era can be somewhat derivative, fitting established stylistic frameworks – not every piece of music has to break new ground to have value.  There are benefits to adopting existing musical languages, enabling audiences to extract and perceive meaning quickly and easily, just as is the case with spoken language.

It’s feasible, too, to suggest that the creative process within the human mind shares some fundamental similarities with an AI’s, involving the assimilating and recombining of musical elements. 

And, I dare say, someday it may become impossible to distinguish – at face value – between isolated artefacts of creativity as being uniquely human or AI generated.

A key difference remains, however, regarding the origin and purpose of human music: it is created by and for sentient beings.  This perception gives music enormous intrinsic value.

Given this value, then, what is the bare minimum level of emotional investment and musicianship that audiences will tolerate before losing interest in music altogether, or consider it to be soulless or valueless? 

Are abstract text prompts really sufficient to guide the creation of a tangible musical artifact – often in the form of a mere audio file – when the output is left up to a machine? If they are, then I would compare this process to conceptual art, where the idea behind a work is vital, transcending its physical expression. It’s an idea I find myself unable to embrace in most spheres of music-making.

In the future, I wonder, will the mere knowledge that a human being has participated in the creative process – even minimally as a prompt engineer – satisfy listeners or will AI be able to get away with implying human involvement instead? 

Given that AI has mastered human-like conversation and tools such as ChatGPT are already anthropomorphised by some, should we allow ourselves to be deceived in such a way?

Today, I’m reminded of the uncanny valley phenomenon in computer animation, where increasing attempts to mimic human likeness can lead to a peculiar eeriness – detectable by some, but not all human beings.  My own feeling is that AI technology might be better used for tasks highlighting its uniqueness, rather than trying to mimic human artists. 

RISKING PUBLIC INDIFFERENCE

In the long term, I don’t believe that the adoption of AI music tools can be prevented, nor do I think it should be.  And my concerns don’t only lie only with AI’s soon-to-be extremely impressive capabilities, either.

What worries me more is the question of whether audiences will, over the long term, become indifferent or even oblivious to the very idea of differentiating between human and AI-generated works. Will the fundamental principles of craftsmanship and visible human involvement in the arts, existing for millennia, continue to be valued in the age of AI – particularly as we face an inevitable flood of computer-generated works?

Someday, I can imagine a situation where promoting music created by humans and AI is done side by side, with less and less distinction required between the two. Without a clear and direct connection to artists themselves – or any insight into the creative process – seeing beyond this form of manipulation may become challenging. Many of us, I expect, will simply stop caring, as making the distinction – if it isn’t advertised – will require too much effort.

Thankfully, some AI labs, such as Google DeepMind, are trying to tackle this problem with tools such as SynthID – created to watermark content as AI-generated at the point of creation in order to keep the public informed.

LOSING CONTROL

Ways of curating and distributing music for public consumption continue to evolve all around us. Long gone are the days when only a few major entities, such as the church, academic institutions, or record companies, held the reins over such activities. Historically, this type of control has invariably been combined with the power to determine, validate, and even distribute content to the masses.

This effect is amplified by those with the most resources and means of promoting music – further relegating its intrinsic value to a secondary role.  In an overcrowded marketplace for music, more and more needs to be spent in order to gain attention.

Much of the music we encounter around us is driven by aggressive marketing or linked to highly marketable entities such as films, TV talent shows, games, or social media influencers. To stand a chance of being heard, artists often have to piggyback on these mediums and platforms – yet another shift away from the idealistic notion that artistic merit alone is the path to success. Perhaps this is the sad consequence of what has essentially already become an oversaturated arts landscape in the era of democratisation. 

The widespread accessibility of marketing tools themselves – ironically, in part thanks to AI and algorithms, offer to level the playing field in what will, I’m sure, be seen by some as yet another victory for the individual. But in a world already overflowing with music and messages, it remains to be seen whether such changes will really be of much use in a global, AI–fuelled race for attention and engagement.

WHAT COSTS NOTHING IS WORTH NOTHING

The scarcity of recorded music once significantly determined its market value. Before desktop tools and extensive sample libraries emerged, access to expensive recording studios and music ensembles was extremely limited, making it challenging for musicians and composers to realise their musical ideas. The complexity of externalising and refining musical ideas might have deterred many from even trying, possibly ensuring that only the most privileged, talented, hardworking, or committed rose to prominence.

At the time of writing, over 100,000 tracks are uploaded to streaming platforms every single day.  And this is before the widespread adoption of AI tools promising to make the creative process easier than ever.  With more recorded music in existence than anyone can realistically listen to in a thousand lifetimes, will the adage, “what costs nothing is worth nothing” begin to apply as the cost of creative thought theoretically tends to zero?

This question gets to the heart of my motivation for writing this article. While increasing technologisation appears to have made the creative process easier than ever, it has also blurred the lines between talent and technical proficiency. The adoption of AI music tools will, I believe, only accelerate this change – for better or for worse.

Perhaps in the face of future generative AI, the very notion of attempting to ‘qualify’ a composer will become futile – seen as an outdated, almost aristocratic folly in an era where were all of us are artists. Yet, in the kind of a meritocracy many of us claim to aspire to and support, the idea of rewarding the laziest among is surely pretty unpalatable.

SEEKING OUT THE HUMAN TOUCH

Traditionally, art has often been idealised as a reflection of life and the human condition.  But if this is to remain true, how can AI take the place of artists who are motivated by this aim?  

AI currently lacks an inner world or any sense of purpose.  It has nothing to kick against; no motivation beyond a text prompt. Even when it acts to create, it is presently unaware that it is doing so. 

Fostered by our attachment to screens and our growing detachment from one another in person, the deep human connection that has defined music for centuries – if not millennia -is, I believe, under threat of becoming increasingly intangible. As a result, music is now often reduced by corporations to being mere “content” or “assets” – stripped of its essence as any sincere form of expression.

Compounding the problem, composers often find themselves asked to adopt the style of other successful productions or artists, guided by temp tracks or stylistic decisions made by executives – further alienating them from their own intuitions and personal motivations. 

Removing the personal from music will, I believe, become increasingly dangerous for composers in the era of AI, as it will render them even more replaceable by algorithms capable of executing similar tasks. 

The devaluing effect on music through the potential use of AI may be further aided by the aforementioned archetype of the composer: a remote, shadowy, and enigmatic figure – known to exist, but rarely observed or directly interacted with. This distance has only been increased by the ongoing technologisation and commodification of music. Perhaps it has even contributed to growing public apathy towards artistic authenticity in general – a dehumanising trend emerging long before AI hit the scene. 

Audiences appear open to substantial technological intervention in the creation of artistic works, so long as they believe – or it is implied in some way – that a key human figure is involved somewhere in the creative process, even if only symbolically or conceptually. 

The potential “handover” of all music creation to AI in future – if such a thing ever occurs, and I sincerely hope it doesn’t – could simply result from changing public attitudes toward artists, and a growing sense of detachment or indifference toward human creators in general. 

This, to me, seems much more likely in areas such as media music, where artists are already largely invisible to the public. To counter this, composers may need to step out of their darkened studios and into the public eye if they wish to endure – demonstrating what it is that makes their work and ideas unique and personally motivated.

Rather than leaving music to tech bros, composers and musicians should position themselves at the forefront of AI technology, becoming its driving force. If history teaches us anything, it’s that very little can prevent the march of technology in any domain of human endeavour, and it’s incumbent on each generation to explore new ways of staying relevant – innovating and keeping art intrinsically human – even through tumultuous technological change.

WHAT NOW FOR COMPOSERS?

My perspective on the long-term impact of AI on human artists tends to lean towards pessimism. I do believe, however, that the direction eventually taken will be determined by society at large – by composers, musicians, publishers, and audiences.  

If audiences cease to care about the origins of music, then so will the markets who serve them.  If, however, audiences take the lead – finding themselves concerned enough about human involvement in music to actively preserve it – then perhaps markets will be incentivised to follow them.

AI Music doesn’t, of course, spell the end of human-made music, nor does it mean that we, as individuals within our local communities, cannot continue to celebrate what we cherish most.  But we may need to act swiftly to solidify the valuable relationship existing between human artist and audience, or risk losing the essence of such a connection to future generations.

More broadly, if we value human-made music, we may need to show more interest in the human beings around us who live and breathe it.  If we fail to do this, then the gradual removal of humans from the creative process will become ever more likely in many settings.

Instead of choosing only to celebrate the globally famous, who are relentlessly marketed at us online from every direction, perhaps we should learn again to show some interest in the composers and musicians around us, seeking meaning in realms unreachable by AI. This might take the form of live recording, performance art and concerts – any and all areas where the human aspect is most evident and impactful.

Looking far ahead to the potential of AI in the arts – particularly if it gains some sort of inner life or sentience – this affords its own intriguing possibilities.  Maybe someday we will be led away from thinking of AI as a mere imitator of human beings towards something complementary to them; a distinct form of intelligence with its own stories to tell – as far-fetched as this sounds today.

I take some comfort in noting that the development of robots surpassing human capabilities doesn’t appear to have lessened the value of human competition in sports – rather, such technology has highlighted the unique qualities of each. We continue to value human endurance, strategy and emotional engagement, while marvelling at the dexterity, speed and efficiency of robots. It’s my hope that AI in music and art follows a similar pattern by not replacing people, serving as some kind of augmentation to a diversifying spectrum of artistic expression – maybe even giving rise to new art forms and an entirely new category of creativity.

The challenge for artists will become one of harmonising technology with the spirit and essence we cherish in human activity and interaction.  

A certain degree of inevitability seems to exist over AI’s adoption, but I believe society ultimately retains control over how much, and in what ways.   Finding a balance that ensures all forms of art maintain their profound connection with our collective human experience will, I think, be necessary if our lives are to be enriched, rather than eclipsed, by artificial intelligence. 

Copyright James Hannigan 2024


AFTERTHOUGHT: THE DIVIDED BRAIN

A question arising in relation to the adoption of AI tools is whether we are collectively losing our ability to discern what is genuinely human in our online interactions – musical or otherwise. In his book The Master and His Emissary, Psychiatrist and Philosopher Ian McGilchrist explores the concept of the divided brain, suggesting that the world – particularly in the West – is increasingly leaning towards left brain dominance. He posits that the left hemisphere is largely reductive in its outlook, perceiving the world in terms of discrete parts and certainties, whereas the right hemisphere is more questioning, viewing the world holistically and recognising beauty and interconnectedness.

In this sense, perhaps the reductive left brain qualifies a piece of music as such simply because it superficially matches the description of music, whereas the right brain questions its authenticity and purpose in a broader context. 

The right brain may be particularly attuned to the nuances and deeper meanings that go beyond surface appearances, perceiving emotional depth inherent in artistic creations. For the right brain at least, human beings and their motivations clearly do matter. Is it any surprise, then, that some the left brained leanings of some technologists miss the point of why we have music in the first place?

Perhaps this hemispheric tension and the tendency towards left brain thinking in today’s detached, online world, is making us more prone to being seduced by the illusion of personhood generative AI apparently seeks to fabricate. 

That we can so easily be fooled may say less about the capabilities of the AI itself than it does about our own diminishing ability to understand one another as human beings. Many any of us, I believe, would be horrified by the notion that we can be lulled into perceiving music as heartfelt, when the AI model behind it not only has no heart, but the lights aren’t even on in its figurative head.

<< BACK TO WORDS