It’s Time to Regain Power Over Digital Representation: A Look at AI ‘Art’ Generators and Predatory Image-Making

By Laurie Luna

From hypersexualised AI portraits by Lensa and TikTok to the stealing of work by deceased artists, the overconsumption of AI-generated images can be draining, boring and uninspired. The prompts can be creative – we’ve seen things that haven’t happened (yet): Scooby-Doo the Musical, and an ‘80s live-action Powerpuff Girls. While these particular ideas are harmless, image generators have been used to push dangerous stereotypes.

Midjourney is a popular AI image generator. While exploring their public prompt threads, it doesn’t take long before sexualised, whitewashed and infantilised images emerge in endless streams. Created by users with names like Tom, Josh, Robert (…you get the idea), this is a breeding ground for men’s fetishes.

When ‘woman’ is used as a prompt, the results are young, blonde, skinny white women with large pushed-up chests, minimal clothes and a child-like face. Expressions devoid of emotion, eyes soulless, posing submissively to the male-gaze. This raises similar questions as DeepFakes, as images of real people can be uploaded: what is being done to protect women? Because Midjourney’s users don’t need permission to make these images, they can save and post them wherever they wish. Predatory behaviour thrives, while autonomy is non-existent.

Writing for Science Focus, Dr Kate Darling’s piece, ‘Stop Blaming Avatar-generating AI for Needlessly Sexualised Images – Fault the Creators Instead’ focuses on the Lensa filter app, which is responsible for sexualising and fetishising Asian women. She says that the app is not a reflection on our society, but ‘a reflection of historical bias and injustice that a company is choosing to entrench and amplify’.

Women’s bodies have been historically exploited and unethically used in art, and in AI generators we see the same thing happening. The more we use AI and train the machine to see women as people and not objects of sexual desire, the better able we may be to combat the biases fed into the machine, regaining power over digital representation, and call out artists exploiting the tool.

When artists use AI tools like text-to-speech or AI images from easily accessible websites like Artbreeder, questions of ownership can arise. Who made the art? Who owns the art? Do the programs own the art? How much of YOU is in this piece? And the classic – is this art?

Depending on how conceptual you want to get, mediums, tools and processes alone are not art, and AI is indeed a tool. Just like throwing paint is an action, writing an AI prompt of a ‘robot eating ice cream seductively’ and posting the result to Instagram crediting yourself as it’s ‘creator’ is a series of actions.

You can choose to believe or not believe something is art. Its meaning is ever-expanding. In Ben Davis’ book, Art in the After Culture, he writes ‘“art” stands in symbolically for the parts of cognition that do not seem machine-like’ and he speaks on the ‘unnerving excitement’ of AI generating new compositions in the style of famous or deceased artists.

But rather than seeing if a robot can mimic the soul of a human, AI can aid our souls as a tool, as part of the artistic process, a collaging experience that aids decisions making. Not be the end goal or govern the artists' choices but allow them to explore alternatives and to push our ideas outside of our comfort zone.

In a conversation held on Twitter Spaces with Laurie Simmons and Claire Silver, hosted by @Fellowshiptrust, Claire Silver, an artist working collaboratively with AI, said: ‘Photography is a camera for what is ...and AI is a camera for what isn't’. AI has incredible capabilities to remove the soul from an idea, and reflect it back to you, informed by whatever fuel it runs on, whether that is your own images or an unknown anonymous bank of stolen artworks from the internet.

They went on to discuss if an artist should share their word prompts for the AI, as this is a recipe to create the image, and the artist should have the right not to disclose that. Do artists owe their viewers an explanation of the process or who/what created the work? Or will this be a never-ending cycle of credits and disclaimers?

But when I look at an AI image, I ask: who/what is in the image? Does the artist relate to the image? If so, how? Does the artist contain harmful biases? What are they doing to combat this? Have they pushed the boundaries of the medium? Will they share what prompts they used to achieve this result? Did they input other artists’ names?

AI generators output the information of the past, which includes all the misogynistic information we’ve been fed since the archive existed. So we cannot place this kind of AI on a pedestal, or judge it as a ‘new’ concept. AI-generated images can be obvious and predictable. AI-prompt-generated images allow artists to act quickly on ideas without commitment. There is so much potential to explore what art is and how it can be created to enhance the world withstanding unethical art creators. Woman artists can use AI to experiment unapologetically and safely.

What would an intersectional feminist AI generator create? Could it create a female gaze, or even utopias? AI images can be a tool for women artists seeking alternative methods of art creation, challenging male-dominated techniques. I believe this is an exciting time for feminist artists, cyber feminists and image-makers.

I’d like to give a special mention to @loopartcritique, an experimental art feedback program on the MUD Foundation’s Onland.io metaverse, Loop Art Critique founder and artist @arielbaronrobbinsart and the artists-in-residents I’ve been working alongside that are creating AI-assisted works that challenge AI to create works that put the artist's voice to the forefront: @slimebubble @marcocaridad @nstgrm.cha @zhoupengstudio

This text was written by a human, not chat GPT.