‘Superficial Intelligence’ – The Role of AI in Creativity

“What is art?”

-everyone, at some point (probably)

---

It’s a question as old as art itself. Ever since the first cave-dweller decided to spend some time doing anything other than look for food, eat, sleep, or die; and found himself etching some crude shapes on the walls of his abode. Inevitably, those nearby to witness this first attempt at elevating consciousness to another dimension must have felt less than impressed: wishing this useless layabout would instead go back to doing one of the aforementioned four useful options. Who was he to presume to waste time, and indeed ruin a perfectly good cave wall, by making this demonstration of human will? Nevertheless, it was bound to catch on.

We’ve come some time since those early examples of people escaping the bounds of purely literal expression by abstracting the sights and sounds around us. We’ve now experimented with countless types of media: paintings, sketches, music, theatre, novels, and learned something about ourselves from its best examples. And, at each step of the way, we have always asked ourselves that same question: in an effort to not only explore the human condition in all its inglorious detail, but equally to push the boundaries of the art forms themselves. This deep, complex, ever-evolving struggle with the question of who we are is, I believe, at the very core of what makes us human.

Today, the question of what makes us human may be changing more rapidly, and more fundamentally, than ever before. Over the last couple of years we have seen a radical shift in the way we use technology – the problems it can help to solve, the tasks it can simplify (or even do for us), and the complexity and depth of the solutions it can provide.

I am of course, speaking of Artifical Intelligence, and specificially, the very recent advent of generative AI. The technology is not fundamentally complicated. The most ‘visible’ AI model are Large Language Models (LLMs), which are essentially extremely advanced predictive text generators: using complex mathematics to ‘guess’ (with a very high certainty) what the next word (technically, ‘token’, or part of a word) is most likely to be in their answer to a human prompt. They draw from billions of data points, across their grounding (or training – where they are exposed to huge volumes of media from articles to manuals, images and more) and web sources which they access where required. Other types of AI are available (particularly diffusion models, which can create visual or audio media by generating and removing random noise), and work using similar technologies.

At the moment, we (the layperson) are just starting to get to grips with how to utilise these tools. For most of us, it doesn’t go far beyond the odd harmless (!) comedic prompt (‘generate an image of a politician as an exotic creature’, for example), or an embarrasing question that we otherwise would have asked Google (or, heaven forbid, an actual informed person). However, for many, they are increasingly relied upon as a core part of daily life. Each day, AI becomes more embedded in our personal lives as it becomes indispensible.

I myself regularly use ChatGPT in myriad ways on a daily basis – both in my personal life and for work. I’ve found it very useful in many cases, though overhyped in others. For example, I use it to transcribe and organise hand-written notes from client calls, which saves me probably 5-10 minutes each time; and even to help me EQ some of my instruments (this works on a broad level, though it isn’t close to beating someone with a good ear, obviously). However, while I have tried asking it to help me write blog posts or client proposals as an experiment, beyond simple structure/formatting advice (though it’s not even that good at this), it fails miserably at generating anything remotely interesting or pleasurable to read. This is also true for diffusion models, which consistently produce images and video with a consistently cold, corporate, same-y look.  

This brings me to what I regard as the fundamental flaw in generative AI within a creative context: it cannot produce anything new. It is incapable of speaking in any novel artistic voice, and is utterly devoid of any sense of style. It cannot move you, because ultimately, it does not experience the world as you do. Real art is produced by people with demons, regrets, and shame. Real art is flawed – can be judged by others who feel they could do a better job. Real art has a purpose – representing something fundamental to the human experience by making an impression on something: be it a page, a song, or a canvas; and exposing it to the world where others can take something from it. Real art is a struggle by people and for people, and as such, no set of computer code can begin to replace it.

So why, then, is much of the discourse around AI today concerned with how AI can replicate creativity? Music YouTuber Adam Neely produced a terrific video recently (https://www.youtube.com/watch?v=U8dcFhF0Dlk) highlighting the issue: discussing the rise of platforms like Suno. These platforms aim to replicate the creative process, radically speeding up music production and empowering users to create entire albums in a fraction of the time a band might usually expect to record in a studio. At first glance, it appears that a service like Suno stands to radically democratise music production, empowering individual creators to produce high-quality music, and taking some power away from large brokers such as record companies, large studio conglomerates, and streaming giants. This could only be a good thing, right?

Well, I suppose this problem exemplifies what I’ve been thinking about. There are plenty of moral and financial arguments against getting involved in platforms like Suno (many of which are well-examined in Neely’s video), but at the end of the day, I think it comes down to one practical truth:

Making art is supposed to be hard.

Once you start to cut corners, you may well create the illusion of producing good music, but it will ultimately fall flat emotionally.  If there has been no struggle, then art is meaningless.

We must then resist this temptation to outsource creativity to a cold and uncaring ghost in the machine; because if we don’t, we will inevitably lose our ability to create great art. This won’t be easy: tech billionaries and their money-men have a vested interest in enticing us to maximise screen time, and surrender as much agency as possible to our devices. It won’t be easy, but we know the cost of losing our human autonomy – our human art – is too high to fail.

I believe we have a reason to be optimistic, though. Historically, when a new technology has ‘disrupted’ the market (effectively, threatened the status quo and forced everyone to adapt to new ways of doing things), it has generally been a positive thing. New technologies have given us cheaper and easier ways of completing tasks, and at their best, disruptive technologies have radically democratised work: eliminating needless repetition and enabling us to focus on more stimulating things. The printing press, dishwasher, and internet are great examples – they give everyone more power within society: to educate themselves, communicate more effectively, and have more time to do things they enjoy. When that happens, we find more time to spend doing creative things.

The truth in this can be found all around us: today, everyone is (or at least, can be) a creator. The cost to individuals looking to pursue creative outlet is effectively zero today, because we all have smartphone and we all have access to the internet/social media. Specifically in a music context, producing music today is orders of magnitude cheaper than even 20 or 30 years ago. A free DAW (Digital Audio Workstation – software used to record and produce music) like Reaper, paired with inexpensive audio equipment will get you pretty close to what was achievable with thousands of pounds of kit 50 years ago, at least on a purely technical level. Along with the advent of social media and streaming services, entire albums can now be recorded, mastered, and distributed for a fraction of the cost of what a single could be produced for in the 1970s.

All this, I think, means that a significant outcome of increasingly useful AI models will result in a revolution – an erasure not of jobs or human worth; but of tasks. Eliminating needless drudgery in everyday life and in workplaces can enable us to use our brains as the engines of creativity they have the potential to be. This possibility, paired with the rapidly dissolving barriers to entry for new artists, can result in more expression that speaks to more people. What could be better than that?

So what are we to do with all this, then? Well, if I were to advocate for a single behavioural change in all of us, it would be this:

We don’t need to use AI less, but rather consume and participate in human output more. 

Finlay