AI art has been a talking point for some time now, beginning to emerge in earnest over the last few years, capturing a lot of media interest in the process. 2022 was really the first year that the technology hit a point of being able to produce images to a standard where it could be mistaken for the work of a talented human, and it has opened up a significant new “what is art?” debate.
Early iterations of AI art created some interesting abstract images, but things fell apart when they were pushed to something more conventional; figures would have arms growing out of foreheads, eight fingers, or two heads, something was always off. Improvements happened incredibly rapidly though, and suddenly in 2022 AI art researchers were producing some incredibly refined digital images.
Rapidly following the technological development, criticisms began to appear, which clearly did not see this as an interesting new face of digital art. Freelancers producing commercial art in cartoony and anime styles were finding that their services were being replaced by AI programs that were cheaper, faster, and often produced better end results. By September news emerged of AI produced digital paintings winning the Colorado State Fair’s fine arts competition; the judges weren’t told the art was AI produced, and assumed it to be the work of a human.
The big problem with AI art is in the way it is produced
Artists have always felt very limited threat from the world of computers; many jobs can conceivably be done by a clever machine, but they can’t create, so artists always saw computers as nothing more than a digital art tool. AI art has now shown it can certainly generate original artworks that are impossible to tell apart from the work of humans, which is something of a tragedy for those who produce art, but the bigger issue really is that AI, whilst impressive, still isn’t exactly creating anything.
AI art programs work by amassing a huge reference library of images; these can be specifically curated, or they may just be a large general selection of images such as listings on Google images. All of these images are labelled with keywords describing what’s in the picture, so when you ask an AI to produce an image of Elvis in a spacesuit on the surface of Mars, it references all the images with those keywords, and mashes them together into a new image.
As a result the final image isn’t so much the work of the AI, or even the human that gave the AI a text prompt, it’s the work of perhaps tens or hundreds of actual artists and photographers mashed together into a new form. Without artists to have created those images in the first place, AI art programs couldn’t produce anything, yet now threatened to put artists out of business, using their own work.
The coming storm with copyright and the film industry
As long as AI art is only putting comparatively small freelancers out of business little is likely to change, but the current push in AI imaging is to produce moving images; so entire films could conceivably be produced by a prompted AI. Of course, this is being talked up in terms of what an opportunity this could be for independent filmmakers, and the mainstream film industry would certainly be attracted to the cost savings. But, where will the reference library of film images come from? It can only come from the densely copyright protected film libraries owned by the same film industry.
So we could certainly see movie studios producing a new franchise instalment of a superhero film built by AI from their own library, putting film artists out of business, but any independent efforts will find themselves swallowed in copyright lawsuits if their AI references a film owned by someone else.
The lesson here really is that we are all replaceable by machines; AI art is still in relative infancy but has already largely mastered thousands of years of human artistic development. There’s also an argument as to how our own creativity works; if I asked you to draw Elvis in a spacesuit on Mars, you would probably think of other images you had seen of those things, the differences between our intelligence and that of computers is paper thin.
