Shifting roles of AI and Artists
How do we assess the role of AI in Art? Dissecting two works and writing through my thoughts
The words ‘Art’ and ‘AI’ have spawned opinionated discussions on both sides of the art and tech worlds respectively. Safe to say the jury’s still out on whether AI can wiggle its way into the art world without any backlash. But this last week I saw two works that got me, an AI-art-sceptic, curious on what works made with artificial intelligence could look like. So join me as I endeavour to asses to role of AI in art.
The first work titled The Moon Underwater takes inspiration from the centuries old Japanese tradition of viewing the moon through its reflection in water. Here a sleek, techy-looking tree drops bubbles of fog that bounce onto a bed of water, eventually dissolving into its surface. The effect is a tranquil viewing of what seems like a natural phenomenon, randomized and sensory. The mechanical tree however, indicates a deviation from nature, visually marking out that this is not an organic act.
The artist duo A.A. Murakami worked with Anthropic’s AI model Claude for this project, describing it as ‘a studio collaborator supporting ideation and experimentation rather than replacing human vision’ (Anarchy Daily). At the core of this duo’s artistic practice is the merging of technology with physical phenomena like bubbles, fog and plasma. A string of code keeps the operation of their installations running behind the scenes, while the viewer-facing elements of water and vapour create the impression of something born out of nature.
The collective is interested in the physical manifestation of Artificial Intelligence, and it is curious that they place a heavy emphasis on this manifestation in the form of elements that are closest to, and hence representative of nature. Unlike various other use cases of AI in art that look like touch screens and image generating models, A.A.Murakami’s work seems to seamlessly slip into the realm of the natural.
A stark contrast to this is the next work, Game of Possible Lives by Tomo Kihara. Contained within a digital screen, this interactive game invites players to experience the potential lives of AI generated, fictional people who could statistically exist in Japan. Players make pivotal life decisions for these people. My boyfriend and I played this game last week and the person we were assigned was 22 year old Akari Yamada, a construction site coordinator faced with a plethora of big life choices. From choosing between moving countries to follow her love life or staying and prioritizing her career, to starting a new business venture while her health deteriorated, we argued over what decisions were best for Akari— someone I felt oddly close to and understandably responsible for.
At the end a genealogical chart showing the chosen and potential paths of Akari’s life flashed on the screen, giving us the option to restart the game and make different decisions for this human that we had been put in charge of. Even though I know Akari Yamada is not real, I couldn’t help but be very proud of her decision making skills (except the one that my boyfriend made), and overall was happy with where she ended up in life. But I couldn’t help asking myself, would Akari be happy with where she was? Would she regret staying in Kyoto for her career instead of going to Zurich with her long time partner? Or would she thank me for the financial security she had when she was in her later 50’s and able to invest in a sustainability start up?
The fictional people in this game are not bound by a single fate, and players can return to past crossroads to alter decisions— making this game extremely hard for indecisive people. Thankfully, I am not one of them. But this game did make me think hard, harder than ever actually, about the way I make decisions. What goals do I prioritize? What things in life am I okay to compromise on? Do I spend too much time on work and too less on building inter-personal relationships? It all makes my head hurt. In a good way I suppose.
This work allows us to experience other people’s lives in a way that data alone cannot capture. And in doing so, the integration of AI here is quite interesting.
Now on to the complicated yet nail bitingly interesting part. How do we assess the use of AI in these works? First let us compare the actual ‘AI’ part of both these works, for which we can only go on face value of what has been publicly shared. Full disclosure, I am far from an expert in AI. If anyone reading this is more proficient in the language of meep meep please do share your thoughts in the comments.
The first work by A.A.Murakami utilizes the infamous AI model Claude. Here, AI steps in as an expert on the science of the artwork, helping the artists understand the nitty gritties on the matter of fog and plasma and why bubbles bounce on water. My honest reaction to this is, perhaps a human could have done that. But I am more interested in the lexical field of ‘nature’ that has been created to describe this work. This can be justified given that A.A.Murakami’s work focuses on the elements of nature, but a big part of these installations is the tech behind it. However the tech part is seldom seen as separated from nature, always entangled with it, which is funny given the dichotomy of those words that literally make up the two ends of the spectrum. This is highlighted by Alexander Groves, one half of the duo, describing the artwork as a garden and Claude as the gardener that is silently keeping the whole harmony working (YouTube interview, 1:38). Similar metaphors of nature show up in the 3 minute video introduction to the artwork below. In simple words, this seems hella sus to me. But let’s get to the next work for now.
The AI in Game of Possible Lives differs in one main way, that it is a Local Large Language model. Unlike Claude which is a proprietary, cloud-based model managed by Anthropic, the local LLM is hosted directly on a particular machine. As mentioned on the artists website, the model is fed data from Japan’s 2020 national census, and the following description gives us a clue into what role the AI fulfills for this particular work:
Large language models are trained on massive datasets of online text. As a result, the worlds they generate tend to mirror contemporary social realities, including their biases and inequalities. The work thus becomes an experiential archive of our current moment, inviting players to navigate and confront the systemic forces shaping life today
We don’t have clarity on exactly what data the AI is trained on, but it’s clear that it steps in as a simulation of the realities of contemporary society. Here, the AI is at the centre of the work, creating narratives that reflect our current realities, flaws and all.
An important factor to note is that the images generated in the game are from an AI model developed by pixel artists themselves and the model ‘utilizes consented artwork data from a community of creators who are compensated for their contributions’ (artist website). This, I would argue, shows more of a reasonable relationship with AI than the previous work.
And this is how I would endeavour to assess the use of AI in both these works. For any healthy relationship, interpersonal or intertech, we need to lay down clear boundaries.
In Game of Possible Lives, the artist places AI at the heart of the work but retains the boundaries of creativity. Illustrations are done by pixel artists, and the fact that the AI operates on a Local LLM suggests a greater degree of control over how the model functions. The only aspect of which we are unsure are the large pieces of text that the model is trained on to reflect societal biases, and whether these were obtained with consent or not.
In comparison, The Moon Underwater shows a blurring of roles between the artist and machine. The AI has no exit point from the work, being described as ‘the technician that [the artists] leave behind’ (YouTube interview, 1:32). As the audience we are unsure what role the AI fulfills in this work. Is it purely a researcher? Or also the technician? What degree of participation did it have in the concept? And where does this leave us as human artists? Because the word human artist seems to be of great distinction in these times.
The role of AI in The Moon Underwater is clearly undefined but I’m willing to hope that maybe this was intentionally left vague as some thought provoking experiment into the role of AI and Art. But I will admit the more I am provoked the less I am convinced. The AI here has as much authority in the work as the artists themselves and no amount of nature-rearing metaphors can disguise that fact.
These works reveal two differing attitudes toward AI in Art. They bring into question the role and subsequent responsibility of the artist within AI assisted works to define the role of the machine.
Moving forward, I think we will see more work in the art world that calls into question this role of the artist with regards to AI, in perhaps some twisted re-telling of the Death of the Author. It is important we look beyond the glossy finishes of tech to understand whether it is simply expanding artistic possibility, or restructuring artist authorship itself. I hope this helped in discussing the different ways artists view meep meep in the creative space. Drop a comment with your thoughts, I would love to chat more on this.
Until next time,
Prachi
Sources:
Artist Tomo Kihara’s website
After reading through this article what comes to my mind about art and AI is the rise of artworks generated from AI becoming sort of a movement. Like AI becoming a medium. How art deco was a movement, impressionism was a movement. The artist will use AI as a medium to interpret whats in there mind.