The story so far: On October 3, acclaimed comic artist, Kim Jung Gi, died after experiencing chest pain while traveling to the New York Comic Con. He was 47. Within days of his passing, a Twitter user claimed they had trained a deep learning model to draw in the style of the South Korean illustrator, and had shared alleged images created with the model. The backlash was swift and stormy as artists and art lovers expressed their feelings regarding the advance of AI art - and what it could do to human artists.
(For insights on emerging themes at the intersection of technology, business and policy, subscribe to our tech newsletter Today’s Cache.)
How AI creates art?
Machine learning and deep learning models are now available for people to create art work. Stable Diffusion, released in August by StabilityAI, is one of the models that can be used as its source code has been made available.
Several AI-based art generators use Stable Diffusion to help people create their own images. Many of them ask users to enter a text-based prompt or even an image, which is then used to redraw, mimic, sketch, paint, synthesise, adapt, or alter images to create the desired result.
In other words, even a person with no art experience can enter a prompt such as “a parent and one child standing on the beach and watching the pink sunset together” to produce an image on an AI art generator, using the Stable Diffusion model.
With additional keywords, image prompts, or even pre-built filters, this image can be edited infinitely to deliver a certain style (such as photorealism or anime) and show specific details (such as a dolphin in the background, or even an asteroid crashing into Earth).
Why are some artists against AI art?
While art created with Stable Diffusion can get the job done quickly, the result can also range from slightly unnatural to downright ghoulish. In order to learn how to produce passable images complying with the prompts typed in by the users, deep learning models are fed millions of works of art for analysis. These artistic data sets also include the copyrighted products of artists who are still living today. In the process, the artists’ names can become prompts for the AI art generator to imitate their style or aesthetic when producing pictures using Stable Diffusion. Non-artists can now freely use these produced images, and even monetise their art work.
Greg Rutkowksi is a digital artist who was caught in this drag net. His name is one of the top prompts on AI art generators, despite his discomfort with the idea. Many published pieces on AI art platforms are eerily similar to his own extravagant projects featuring dragons and epic fantasy landscapes.
Critics have claimed this is a violation of copyright law and that artists should have the freedom to remove their work from the data sets used to train AI models.
The Twitter user who claimed they trained a model to imitate the late Kim Jung Gi’s lush, ink-based style of drawing characters was also criticised for treating the illustrator’s lifetime of work as a generic end product that could be replicated by a machine.
However, the Twitter user claimed they were paying homage to Kim Jung Gi.
Could “AI art” replace human artists?
Freely available AI art generators could make it possible for almost anyone to carry out this task in just a few minutes, at far lower costs. These models could also eliminate the need to pay for licensed art by making it possible to choose a subject, copy an artist’s style, and create a customisable result - all for free. This has the potential to completely restructure fields such as fashion design, architecture, cinema, book publishing, and more.
Artists are also concerned that any work they upload or share online could be used without their permission to train deep learning models. This could let others create derivative works of art and profit from them without having to give any credit to the original artists whose efforts made the AI art generator possible.
In September, this very scenario came true when American designer Jason Allen won the first prize in an art competition for his piece ‘Theatre D’Opera Spatial’ which was created with the help of an AI program.
As deep learning tools reach more users by the day, artists are waiting to see how different sectors will embrace or reject the rise of AI art.