The roots of AI Art goes back to 2015 when ‘automated image captioning’ allowed AI to give text to images and partially that’s where the idea of reversing this process came around i.e. making images out of text. However, it’s not as easy as it looks, especially when one concerns the ethics involved. The most glaring issue with AI art is that it requires usage of dataset to mimic them in order to create a new image. And this dataset is composed of many images, taken without the creator’s permission as per their copyrights. More than that, there is mere violation of consent of these artists as they are not asked before their artwork is put into the database.

Another pertinent issue is the connection between artificial intelligence and authorship of art. For instance, in 2018, a portrait generated by a machine learning (ML) algorithm called a generative adversarial network (or GAN) sold at Christie’s art auction for $432,500. The algorithm in this case was trained through the paintings of Renaissance masters, sourced from WikiArt. This points very clearly that AI Art is not something created entirely by algorithms as it rests on the works of so many artists.

Although people involved with AI art call it ‘prompt engineering,’ there are a lot of gray areas around the artwork which are used in the dataset without the permission of their creators. Even with many popular softwares like midjourney, the issue is with the input that requires a dataset which leads to output. All this database is full of art which has been taken without the permission of artists, thereby violating intellectual property rights.

When I wrote a piece about AI Art that was centered around queerness, I was amazed at how ‘queer’ it seemed in an optical manner, unaware that this inspiration comes from decades of queer art. More than that, we have seen how so many AI chatbots have picked up unfortunate abusive behavior from their interaction with humans. Similarly, there’s a high chance there will be a spew of misrepresentation or fetishization for the worst,, which has been already floating around the internet. For instance, a post about Bengali wedding recreation by AI art simply stereotyped them as fish eaters.

There is also a question of access when it comes to AI art as it provides access to everyone for making art, which is nothing wrong in its essence. But it’s the demolition of art as we know it. If we can make something through a few words, why would someone buy a painting from an artist? If accessibility is one aspect, it leads to an unfair world, which is already hard for artists and AI Art is only going to limit their creative potential. What value would an artist bring if everyone can do the same thing? But the difference is they are doing it through words while artists put long years of labor, hardwork and techniques in their work. The promise of AI that someone with zero experience can create the same image as someone with years of training is in the direction of replacement of artists.

Steven Zapata Art in their video The End of Art: The Argument Against Images AIs provides the most informed and nuanced understanding of what’s wrong with AI Art. The video essay points towards the flawed, unethical and deceptive AI system which is almost the definition of ‘unartistic’ as it violates consent and basic ethics. The basic argument is that these text to image based AIs cannot improve upon themselves as they rely on exploitation of creative works all the while marketing themselves as authentic or ‘game-changing.’ The video also extensively decodes the companies working behind the scene and how they are exploiting in the name of creating something between a for-profit and a non-profit (called as ‘capped profit’) which unfortunately doesn’t come under any pre-existing legal structure.

No one is contesting that AI art is art or not? The question is around future and ethical implication when some of these AIs have not disclosed how they make their dataset work. AIs are not tools for artists, they are replacement unfortunately. One of the comments on the YouTube video which I highly resonate with reads, “If they take away the ability and the incentive to create, we will only have the desire to consume. And deep down, it’s just that, consumption and more consumption. This is a strong step towards a less “human” humanity. Not to mention that there will be fewer and fewer jobs in which one can learn and enjoy what they do. This is horrendous, almost straight out of a sci-fi horror movie.”

Art since long has been an instrument of protest and revolution for many marginalized identities. It has been a key element in contemporary social activism because more often than not, it takes visuals and imagery to get an important message across. When AI art is self-generated and highly targeted, the authentic art would take a backset. It will also rob the artists of a space where it matters to them that they told their story. There’s also a Black Mirror-ish possibility where AI begins to interfere with all the intimate information it already has and the control it already exerts over us.

One instagram handle which uses midjourney software arrogantly asks in their bio “Who’s the better artist, man or machine?” without talking about how whatever the machine creates is actually stolen from the artists. In a video essay titled “Is Art Meaningless?”, transgender actor and philosopher Abigail Thorn asks the question about art – “If I don’t get it? How do I get it?” The pertinent question  that comes up here is trying to understand the artist’s intention. But coming back to the controversial relationship between authorship and AI art, whose intention are we looking at in terms of art made by AI? Whose subjectivity are we referring to? Are we equating AI as having human-like abilities or are we actually going to talk about the artists that fill up the dataset.