wondermash

View Original

AI art is human. And shares the same flaws.

Much of today’s AI art debate revolves around answering the question: is AI art, art, or soulless regurgitation of actual art? I’d say it is a very human craft that can lead to art - but honestly, this discussion bores me to death. It is smothering the debate about a potentially transformative tool, and leads away from AI’s real dark issues.

AI art tools are algorithmic programs that harnes the power of artificial intelligence, digital art and self-learning computation. Since the launch of many AI tools in the summer of ‘22 the discourse has shifted from frantic experimentation by everyone and their neighbour to harsh criticism on AI for killing the art scene and on the ethics of the tools. These points, although valid, bypass the most promising aspects of AI tools. Let’s not throw away AI art tools for the wrong reasons.

The most heard criticism on AI generated imagery is that it will put talented artists out of a job. Conjuring up (a synonym for AI art generation I learned from Ambika Joshi from @computational_mama) an image out of thin air has never been easier or cheaper for the masses, so why hire an artist to do it? Many from the artist community have shared their concerns about machine-generated art replacing human-generated art.

The animosity against AI art tools only grew fiercer when artists found out the algorithms were being trained on their art without compensation or permission.

Imagining birth in the future using AI tools (c) @computational_mama

I am not convinced AI tools are a threat to the art community. Sure, I see artists working hard on their art but hardly getting by and suddenly there is someone with no art skills monetizing on the AI art they create with a fraction of the effort of a fully human created artwork. It is cashing in on a novelty. Soon people will not only tire of the specific look and feel of AI art (seven fingers anyone?), but the non-artists creating ‘art’ with AI will soon realize that skill is needed to create a custom picture, and either become actual artists, or move on to the next gig.

The real danger in AI is not that it produces mechanical art that puts artists out of a job (which in the long run I honestly doubt), but that it is flawed and all too human. It’s founded on the flawed capitalistic and hegemonic society we have today and will exacerbate the effects. Like the catchy instagram meme I caught a while back said: You don’t hate AI art, you hate capitalism by @seedingsovereignty.

(c) @seedingsovereignty

AI art: human or machine generated?

A third point of criticism raised, is that AI art is not art, it is a soulless, mechanical regurgitation of human-created, actual, art. That touches the heart of the age-old debate ‘what is art’? I am tired of this debate before it began, but here are my two cents: art is not what it literally is, it is the idea that is behind it and what it provokes. Art is a concept, as most famously demonstrated by the artist Marcel Duchamp’s urinal. An editorial attributed to the artist Beatrice Wood reads: “Whether Mr Mutt with his own hands made the fountain or not has no importance. He CHOSE it. He took an ordinary article of life, placed it so that its useful significance disappeared under the new title and point of view—created a new thought for that object.”

So AI images in itself aren’t art by default, but they can be, if they are so used by an artist. Why can’t we think of AI algorithms as mere tools, no more and no less? Why do we feel so threatened by it? One could argue that AI image tools use state of the art technology in a similar fashion that the first photo cameras used technology to assist the creation of an image. Of course, many photos aren’t art, and most people taking photos are not artists. But there are artists that use the technology of the photo camera to create true works of art. I can’t create a piece of art using my camera for the life of me, but my friend and photographer Wiosna van Bon can and does, and the fruits of her labour are exhibited.

A final point I’d like to make is that in these debates there tends to be a sharp distinction between what is ‘human’ and what is ‘tool’ or ‘machine’, but we all know that this is a leaky distinction at best. In this day and age of pacemakers, mobile phones as extended memory, cars, implanted chips, we have all become cyborgs and our machines are human extensions of us. In fact, humans have used technology since prehistory, and it has become a part of us.

Ownership and AI

I do not assume to know a lot about the algorithms behind AI. In fact nobody does, it’s not publicly shared information. It has been confirmed however that many of the newly launched AI art tools use public, copyrighted images, to train the algorithm. AI tools like DALL-E and Midjourney offer no way for artists to find out if their images have been used, for they're not open-source. Even the AI LAION (claiming to be a truly open AI) offers no insights into how images are scraped to train it.

In a recent lawsuit filed by the American Copyright Organization an artist who created a graphic novel using AI generated images stands to loose his copyrights. The grounds of this lawsuit: copyrighted works must be created by humans to gain official copyright protection. Conveniently disregarding the efforts of the artist in question imagining and writing the story, plus the work that went into manipulating the AI to create an image that is just right. It is difficult to disentangle the machine from the human effort in the end result. But anyway.

The fact that AI tools use stolen content is valid and raises serious ethical questions. I’ll get to that later.

The true potential of AI art

As a designer myself, my play with AI tools sparked curiosity, not dread. Art or not art is up for debate by people who are interested in drawing lines between wat is actual art and what isn’t. I think the potential for AI imagery lies beyond creating images on the cheap (although yes, I’ll admit I’ve succumbed to the lure of re-creating myself in AI selfies and was strangely fascinated by these images, another hours of my life I will not get back).

Eerie AI generated selfies art using Lensa (c) Marjolein Pijnappels

I see AI’s true potential in imagining things that are just beyond our reach. Architect Andrew Kudless (@matsysdesign) uses AI tools to imagine new building materials, conjuring up images of skyscrapers made of textile. The fact that it is partly out of human control is part of its appeal and potential. Kudless, who often shares his process on instagram, reveals that it’s the misunderstandings, bugs and mistakes that give him the best inspiration: “It shows me something I didn’t even know I wanted.”

Architectural experimentation using AI (c) Andrew Kudless @matsysdesign

Kudless likens AI art tools to the painter and sculptor Cy Twombly’s practice of painting in the dark, the automated drawing of the Surrealists, and the use of random chance from Ellsworth Kelly, in the sense that it transfers control to another system, which “allows us to escape our own limited imaginations”. Relinquishing control partly to a machine might come up with solutions for sustainable building materials, future visions of what we want society to look like, and challenging the status quo.

It takes effort

Unfortunately, AI isn’t beyond human limits. It takes effort to get AI to imagine the unfathomable. Like Ambdika Joshi from @computational_mama I am using AI art to re-imagine our society by creating subversive images. Ambdika has a series on motherhood that sparks discussion of how we depict mothers in our current culture and we are both imagining what birth might look like in the future using AI.

When creating images of birth, I couldn’t for the love of me create something that even resembled actual birth. AI did come up with interesting conceptual images that I found very interesting, but birth somehow wasn't’ on the radar. AI inevitably spewed out highly technical, medical imagery of birth, continuing an existing discourse, not challenging it.

When creating humans with AI tools, the women were invariably thin, full-lipped, attaining to Eurocentric beauty standers and as a default: white. Not so when I started imagining mother goddesses, these were imagined as black women and women of colour. Stereotypes and biases are perpetuated and strengthened.

image created using Midjourney (c) marjolein pijnappels. I had to specifically prompt ‘black woman’ or all my birthing people turned up white.

But with careful prompting, it is possible to overcome the biases and trick the system into creating something truly different. In their project ‘Robots in Indian Healthcare’ Indofuturists @antariksha.studio @thiruda.2040 and @murthovic are charting alternative histories and speculative futures of AI - robots, cyborgs and androids in Indian society and culture - exploring nuances of how they integrate - from ancient times to 2079. The accessibility of AI art makes it possible to create a subversive, countercultural exploration of the future, providing alternatives to the predominantly white, Eurocentric and tech-angst-like futures imagined in mainstream media and art.

AI generated Indofuturist imagination by @antariksha.studio @thiruda.2040 and @murthovic

The deep dark flaw of AI art

Let’s circle back to an uncomfortable truth about AI tools: that it is being build on other people’s art? Is using art in any way or shape unethical? How about human artists taking inspiration from other artists? How much ‘alike’ can an AI generated image be until copyrights are infringed upon? How can we even keep up? Many people are busying themselves with this topic, and I am curious to follow this debate. But it is not my main concern with AI art tools.

My main concern is that it is perpetuating Eurocentric, misogynistic, techno-optimistic ideas. If even medical AI and algorithmic decision-making systems show racial biases, what hope is their of ever creating a fair and unbiased AI art tool?

The short answer is: none.

At least not when it is created by a small, uniform group that has all the power and access. In a lecture about’ Algorithmic Injustice’ by Abeba Birhane from Dublin University, organized by the TPI AI Lab of TU Delft I came to understand that an AI ‘for everyone’ is impossible. It is unrealistic and unattainable to create a generic AI tool that is fair and ethical - as the current AI tools painfully show. A decolonial AI is only possible if the AI is owned and managed by the groups it works for. A fair unbiased tool for the masses, owned by a small minority, is an illusion.

But that doesn’t stop AI artists from tricking the system and using it in their advantage Donna Harraway style. In her A Cyborg Manifesto, cyborgs symbolise a rejection of rigid boundaries, notably those separating ‘human’ from ‘animal’ and ‘human’ from ‘machine.’. A ‘cyborg author’ is someone who, when “retelling origin stories, (..) subvert the central myths of origin of Western culture.”

These tools are here, and they are here to stay. But instead of dreaming of Paradise, let’s accept we are all to an extend cyborgs and hack these algorithms to challenge the system that created them, and transform them for the better.

Got any additional thoughts? Would love to hear them and start a dialogue around this topic. Add your insights below.


wondermash offers a space for curious and boundless exploration using insights from trans-disciplinarian research, future speculation and design. Reach out for collaborations and enquiries.