What AI generated imagery can't be
Ever since digital imagery created by artificial intelligence systems such as DALL·E 2 or Midjourney started becoming popular, I have been curiously following the discussions around their purpose and meaning in the art community. And although the technology has been around for quite some time now, its controversy could not be more present as of right now. ArtStation or in my case even Instagram and LinkedIn feeds are covered with many variations of "AI" written in big white letters with a red cancel sign on top, and "No to AI generated images" standing below, with many artists expressing their dislike and concerns towards AI generated "art".
In this blog post I would thus like to share my personal view on the matter, as I find the topic itself and the philosophy behind the meaning and very definition of art truly fascinating.
How do these AI systems work?
I think before starting to talk about the issues, one needs to understand how these systems work and create these images. In a nutshell, neural network algorithms such as DALL·E 2 use prompts - text descriptions such as e.g. "a dog on a skateboard wearing sunglasses" - to generate digital imagery. You can create variations, modify the created images by adding more prompts and get different "artistic" styles.
To do so these systems were trained with billions of images and their associated captions. And here lies already one of the main issue that many artist currently have.
The problem with AI "art"
As I said, these systems are also capable of generating images in established, popular artistic styles, such as from Van Gogh or Salvador Dali.
They are only capable to do so, because they were trained to "understand" what a Van Gogh or Dali looks like. And this goes beyond these famous artists and popular artistic styles. Artworks of many artists were used by these algorithms for their training, and some say it is not far off saying that this can be considered as being art theft.
And I dare to say that most of these AI generated images would not even look close to what they look now, without the already existing artworks of thousands if not millions of artists.
Of course this is also a matter of copyright and a reason why ArtStation recently, due to the following large online protests, stated in their newest Terms of Service update, that they will introduce a "NoAI tag". This will tell AI systems, that the artist "explicitly disallows the us of the project and its contained content by AI systems".
What AI imagery can be
Before starting to express my concerns and general dislikes towards these systems, I would like to share the possibilities I see and positive impact they could possibly have in my opinion, since it is not all just plain bad.
I think there are two areas at which these AI systems can be helpful, the first being inspiration. I like to compare these systems with Pinterest on steroids. Instead of looking through what already exists you get fundamentally new creations of whatever thoughts you feed into that system, which definitely can help to inspire or get new ideas. Especially if it is something so specific, that it is hard to find anything close on these platforms. This may also apply to finding references.
And then I also see its potential in helping those who may have amazing ideas, but are not able to visualise them in the same way an artists would. Having the possibility to translate the imaginative with a few words into digital images, could be incredibly helpful and create new opportunities. Since I am studying industrial design I like to take product design as an example. Someone may have great ideas for a product, but simply not the skills to visualise it in any way.
This is, of course, based on the very concept of AI imagery, with the use of image-information used for training being approved by the owner of the used images and perhaps even compensated.
What AI imagery can't be
Art is subjective. And although there is a definition of what >art< is, subjectivism does apply to it too, to an extent, I believe.
The definition you will get first of "art" as soon as you ask Google, reads as follows:
"1. the expression or application of human creative skill and imagination, typically in a visual form such as painting or sculpture, producing works to be appreciated primarily for their beauty or emotional power."
[Oxford Languages]
AI "art" may be based on existing expressions of human creative skill, yet I dare to claim that it is fair to say, that AI "art" is not art, and never will be.
Of course it does take human intervention and creativity to create these images, and definitely some skill in form of the complexity of the prompts one uses to create imagery that can be considered as interesting or good by most viewers. Yet, there is one thing that it will be always missing - "humaneness".
If you look at a drawing or painting, every single stroke tells a story. Every single detail you see, no matter how big or small, is the manifestation of this human's very consciousness. And this is something that no AI will ever be able to reproduce.
This is something I appreciate throughout all art disciplines and genres. The fact that everything I see in a video game, or hear in a song, is based on other lives being lived and other stories being told. The fact that everything I see is based on thoughts and emotions communicated in their purest form. And that is what makes them so beautiful and fascinating. That is the reason why I love to go to art museums and look at artworks that are centuries old, or just browse through ArtStation and look at the creations of others. And I do not believe that there are as many prompts as you would need to achieve just that.
What is the point of taking away the human, in a world already so artificial?
Conclusion
Apart from everything I just said, I think it is truly fascinating by the capabilities of modern technology. Yet I do stand by what I think about these systems, and believe they should be treated with more awareness and care.
The concept of AI systems creating digital imagery can be amazing for generating inspirations and helping those who may struggle with visualising their thoughts and ideas. But that should not be based on the work of others, taken and used seemingly self-evidently. And they should never ever replace an artist and the final product, which is, in a world where everything and anything is weighted in gold, a justified concern.
As of right now I believe the most important factor is transparency. Imagery created by AI should be tagged as such, and in case of ArtStation, an online platform which is solely meant to be a place for artists where they can showcase their crafts and skill, either put into its own category that is not shown by default on the feed, or removed completely.
As I said in the beginning, this is just my personal opinion. You are of course free to have a different take on this and I would be pleased and curious to hear what you think, no matter if you agree or disagree.