The Art We Let Machines Make

As tech progresses, the future of art feels like it’s traveling to an unknown place. As someone who likes to have one foot in technology and one in art—loving both coding and poetry, I now feel a constant push and pull as I work in these fields. Sometimes, I’m even feeling as if they are fundamentally now in conflict with one another. To try to understand how these areas intersect better, I recently helped run a survey between college students at USC and Stanford on how they thought about technology and creativity. The goal of the survey has been to understand how those of us engaging with new tech feel about how it might impact our lives and future artistic careers and pursuits. After all, we’re the ones that AI is really going to affect the most, especially compared to those who already have established jobs and careers. We talked to over 120 students, a few professors, and people already in the industry. What we found in some ways was perhaps unsurprising but also profound—people have very different standards about different types of art they create and consume.

Students were asked which creative tasks they would willingly surrender to automation versus which ones they would refuse to do so, and there was a clear pattern. Yes, our data indicates in some ways what we’ve all known all along, that there is a clear hierarchy of creative value. It seems intuitive that we largely agree that some forms of artistic expression are more “essentially human” than others. Some art we should hold sacred as a human product, other ones not so much.

The consistent theme amongst respondents was the description of the acts of writing and storytelling as fundamental. One student insisted that they “would never let the [machine] write the script itself,” while potentially being okay with computer-generated prompts as a form of inspiration. Others mentioned how they wished Hollywood would create more original scripts, and that this was only done through humans—with minimal tech intervention into the creative process. One data point in our survey even showed that 69.1% of people would be less likely to watch a film if they knew the writers used AI to help write the script.

Yet, as the conversation shifted away from writing into the visual arts, the boundaries started to blur. AI generated marketing was described as ‘saving a lot of time’, while still mentioning that for bigger scale projects, students wanted to see a human touch. Over 70% of our survey respondents said that if the editing process was sped up and optimized by AI it would not influence their likelihood of seeing the show or movie. Even photography, portraiture is viewed as less important, with students rather expressing excitement about stable diffusion tools and image editors. What many of us feel in our intuition was clear in the data, visual design work is viewed as more replaceable than writing.

Source: Stanford E-Lab x Art.Ificial USC Survey

The Biases Behind Our Hierarchy

While this realization seems to make us squirm, we need to dig deeper into “why.” Why have we constructed this particular hierarchy of art? Why do we seem to by instinct treat writing as more sacred than the visual arts or film editing, even if many of us engage in video games, movies, and music far more than we read books?

One explanation might be that film editors, sound engineers, and game designers themselves are interested in automating their work. Take a scroll on video game Twitter and you can see that people seem to really enjoy playing around with vibe coding their own games together. Others argue that the reason for our comfort with automating these roles is that they are more often implementing the vision of someone else anyway. A sound engineer usually must follow the directive of the artist, not the other way around. But is this really true? Is the work of a sound engineer so much more rote than a writer’s? Further, there is a real argument to be made that sound engineering is incredibly similar to coding—which far less people seem to be against automating. You’re always using the newest tools, and your central role is to implement the vision of the people you are working with. But this is a hard position to argue well from, as one could argue the same for the writers at the top of the hierarchy. Do studio writers not simply carry out the existing vision of the studio, editor, or publisher that they work for? Going further, sometimes people argue that it’s because writing is far more abstract and unique, but are there not formulas and structures to the stories our stories always seem to follow?

I love writing, I get that from my dad, and so I don’t mean to say this lightly. However, I think that we’re wrong to believe that writing represents some pure form of human expression, while other forms of art are simply good technical execution. To go back to my dad he always said that (quoting Samuel Beckett) crafting a great story is like crafting a shoe, “you work on one part, step back and look at it, then you work on another part, then another… you’ll never really know what each choice you make does, but in the end you can stand back and be proud”. I always liked this adage because it taught me from a young age that the two important goals of crafting a good piece of writing are to be beautiful and useful, just like a shoe. Yet, we’ve had no qualms about automating away our cobblers for generations now.

So why do we care about writing more than painting or more than shoemaking?

Perhaps our hierarchy stems from our cultural biases towards intellectualism. Language has for a long time been our privileged form of communication since those who could do it well were able to express themselves. It’s cerebral, abstract, feels close to the mind itself. We appreciate visual art, but we think in words.

But it’s more than that, because while there is a compelling argument to be made that writing has been a tool of the ruling class, that alone is not what makes writing what we wish to save. There are plenty of more privileged art forms in the visual fine arts, such as portraiture, that we as a collective seem to be quite okay with automating (in whatever style you want as much as you’d like). Another explanation might just be familiarity. Even if it’s just texting, most of us write in some capacity every day. Not many of us can say the same about visual art, and even less about film editing or music production. We might just protect the creative skills we personally possess and understand, being okay with automating the fields where we do not have knowledge or experience. Maybe writing truly is closer to the soul. We don’t necessarily understand how our brains work, and it could very well be that language is even closer to our internal states than our vision is.

It might be even simpler than that. These language models are trained on text, and thus our power and capability to use them is dependent on our ability to convey our ideas in text. We found that just about 90% of our respondents have used an LLM and 73% use one daily. However, generative image, video, and audio models are used regularly only 26%, 8%, and 7% respectively. If the language models ran on music maybe we’d all be clutching our recorders.

I don’t know which of these reasons is why we see our collective hierarchies about art. I don’t know if this should worry me, excite me, or simply take it as a given. There is much to learn and discuss about what we wish the future of art to be with AI, and the reality of our internal artistic hierarchies will be an essential part of the conversation.