Auras and ouroboroses: AI as a creative tool

What I find interesting is not only how AI is this mirror of who we are and what we make, but rather how having unlimited access to it has started to change how we think. 

AI has learned how to bend to and understand our language and, in turn, we have learned how to bend to the language of AI. How we speak to a ChatBot and write prompts, and rely on it to accomplish tasks is reshaping our minds. Just like the internet and social media have rewired my brain for better or worse (mainly worse), AI has taught my brain to see the world differently.

AI can be seen as the collective consciousness of human output, the bad, good and ugly, amalgamated into one nebulous blob-like soup. 

Since first experimenting with AI visual effects, I now observe and absorb my surroundings in a completely new way. My mind wanders off into new directions knowing how I can capture and manipulate the world to an infinite degree. In this sense, AI has become a collaborative partner, allowing me to explore paths I would have otherwise ignored and experiment in ambitious ways.

AI can be seen as the collective consciousness of human output, the bad, good, and ugly, amalgamated into one nebulous blob-like soup. My latest film Absolve, made in collaboration with French musician Jacques, is about our relationship with this collective consciousness and the context of AI in the beacon of institutional art history. 

Shot at the Musee Du Louvre, the film follows a morphing teardrop that transports Jacques into an emotional realm. The teardrop grows beyond a droplet of water and becomes a catalyst – or virus – that alters its surroundings. The floating blob of water metastasises, infects, and erodes the classic works of art to reveal the pain hiding underneath the surface – it is a representation of our collective consciousness that has absorbed human creation and spits it back in a new a warped view of what we’re accustomed to seeing.

AI gets our neural pathways to diverge, sparks new ideas, and creates a more fluid creative discovery process. 

This act of transformation is a visual metaphor for the erosion of the 'aura', where the unique presence and authenticity of these artworks are challenged by their generated counterparts. There is certainly a tongue-in-cheek aspect to the film as it shows AI destroying art history. 

The film has over 80 visual effects shots that combine traditional VFX with AI-assisted tools like Stable Diffusion, Runway and Luma Labs. The point wasn’t to use AI because it’s easier but rather a way for me to explore a novel narrative with new visual techniques and aesthetics. To blur the line between what is AI and what is traditional VFX because at the end of the day, it’s all artifice. It’s the back-and-forth conversation with AI that allowed this idea to happen.

Paul Trillo's film Absolve sees the unique 'auras' of genuine artworks being eroded by AI generated counterparts.

The creative ping-pong match where you serve the AI something and it instantly serves something right back is what makes it so exhilarating. It gets our neural pathways to diverge, sparks new ideas, and creates a more fluid creative discovery process. 

These happy accidents lead to new ideas and have inherent beauty on their own. It’s up to the composer, the user, to choose which of these glitches speak to them and to curate their results. It’s a revolutionary way of creative decision-making. 

However, I make sure that it always starts with my idea and ends with my final finesse, sandwiching the AI ping-pong match in between. Curating options in front of you is what directors have always done, we’ve just never experienced it at this scale and speed.

The choice architecture that AI presents gives more opportunities to explore options quickly so that the creative ideation phase can flourish.

The speed aspect is easy to take for granted but in production, we deal in the currency of time. A lot of time is wasted pulling the right references or having to translate what’s in your head to another person. Then you’re backed into a corner to decide because of a deadline. The choice architecture that AI presents gives more opportunities to explore options quickly, so that the creative ideation phase can flourish. However, with this abundance of options comes the paradox of choice. Having too many options can paralyse the artist. Knowing when to stop and commit to a decision is the true art.

However, being able to generate to our heart's content can obscure the value of our inherent taste. AI is only as good as what it’s trained on, and the user is unaware of what is missing from a dataset. What has been deemed creatively pleasing is limited to a pruned dataset scraped off the internet. 

The phenomenon of 'model collapse' in which AI is trained on AI-generated data is concerning.

AI’s choice architecture can be deceiving in that way because when we only make decisions based on what the AI presents us, we are turning down all the options the AI wouldn’t even be able to come up with. Generative AI is only as good as the options it has to generate from. It doesn’t know what it doesn’t know. Surely, that’s not going to push culture forward.

The risk of an AI-driven creative ouroboros a self-consuming cycle of synthetic data—is real. We already exist in a homogenous, reboot, remix, rehash culture and I fear AI will only exacerbate that problem. The phenomenon of 'model collapse' in which AI is trained on AI-generated data is concerning, especially when we consider the long-term implications for culture and creativity.