As an artist, I appreciate original works of art that capture a moment in history, a concept, a theme, or a philosophical idea. Jacques Louis-David’s “Oath of the Horatii” or Claude Monet’s “Water Lilies” series are superb works of art created by human beings with a creative eye and a brush. But what happens when computer programs generate art based on text or voice input? Is that art per se? Or is it just random computer-generated dots grouped together to create an image, without a creative mind behind it? And if the program takes elements from other works of art, is there a copyright issue? These are questions that are now being asked as programs like DALL-E 2, MidJourney, and Stable Diffusion, which are capable of creating an incredible range of images based on text or voice prompts, make waves across the art world.
An article in Wired makes the intriguing claim that AI-generated art reinforces stereotypes against women and people of color. According to the article, tests have demonstrated that DALL-E 2 “leans toward generating images of white men by default, overly sexualizes images of women, and reinforces racial stereotypes.” That’s… not good. From where are these biases emanating? Is there a line of code that tells the program to produce such disturbing results?
Ironing out these details is essential as AI programs such as these begin to proliferate and become popular. In 50 years, classrooms will be using AI for everything from math to art so the kinks need to be worked out before our children are exposed to the quirks and dangers of computer technology. There are plenty of detractors who dislike the idea of AI-generated art, decrying it for its immediacy or availability to the masses. Others, see such tools as a gateway for everyone’s creativity and imagination.
According to Scott Baker at opuresearch,
“AI cannot replace expertise in composition or color theory that’s right for your art or brand strategy, and it will not replace the premium we place on the imagination of experienced professionals, but it may drastically shorten the ideation and conceptual design phase for any project.”
Whatever the nuances of creating digital art, it didn’t matter to judges at the Colorado State Fair in September when they awarded first prize to an artist who submitted his AI-generated image “Théâtre D’opéra Spatial” to a fine art competition. Artists are fuming.
According to the website creativebloq, “It's an epic scene of sci-fi fantasy that seems to show women in elaborate dress looking out through a huge circular portal. However, it only emerged after the prize was awarded that Allen had created his artwork using the Discord-based AI art generator Midjourney before upscaling it using AI Gigapixel to put on canvas.”
It is possible the judges were unaware of the origin of the work, but they have not retracted their decision, so it seems they are satisfied. Regardless, people have called for an AI category in next year’s competition so as to avoid such controversies. One fascinating aspect of AI-generated art is being able to create an image of, say, a girl wearing headphones while riding on an electric bike in medieval London surrounded by aliens and dinosaurs - just by asking the computer to generate the image by itself. A few second later, you’ve got your artwork.
Yet another fascinating tool becoming popular now, is the ability to “outpaint.” This cool concept allows users to generate an image that imagines what existed outside the frame that artists had painted. Outpainting is a creative technique that will probably generate lovely images for years to come and there is no doubt that more tools and techniques will materialize as AI becomes more advanced and developed.
In the meantime, there is definitely art in Artificial Intelligence. The question is, is there art in you?