1
0
Photo credit: Abby Mertzlufft
At the beginning of the Fall semester in 2022, I noticed a plethora of flyers around campus that said, “Art made by computers?” with QR codes. I was intrigued, so I followed the QR code, which brought me to an Instagram page @artificial_brush_stroke. This anonymously run account posts artwork in the form of artificial intelligence (AI) generated images, a new medium popularized by DALL-E mini (a version of DALL-E, a deep learning model run by the research company OpenAI that is open to the public). I thought this would be worth pursuing, so I contacted the account, who said they would be available for an interview if their identities were left anonymous.
The person I interviewed declined to give his name or whether he was related to the Roberts campus (or even a student). He wore a mask, sunglasses, and a baseball cap to conceal his identity. I got the impression that he and the Artificial Brush Stroke were serious about their anonymity. What I learned was fascinating – the team behind the account has a unique process to create art, stay creative and work with the advanced technology they use.
The Artificial Brush Stroke started as a group of people that came together because they saw AI art rising in the art world (usually, teams like this are only one-person shows). They wanted to hop on the trend and started the account and have been doing it since. To get a good following when they began, they placed around 50 flyers around campus. The goal of Artificial Brush Stroke is to explore how close to an actual photograph their produced images can look and to ask, “how far can we push the bar to make people question whether a computer or a human made this?”
Rather than pay other companies, which can get pricey, they use their computers to run every aspect of the account, so the art’s creation is in their hands from beginning to end. They run a deep learning model called Stable Diffusion on their personally-owned computers, generating thousands of images for every prompt. Only about three to four of these thousand images get used, and the rest get deleted. The only downside to using their technology is that they have to export the produced images in lower resolutions than they’d like, typically 512×512 pixels, because they don’t have industry-scale technology. Because they want the final images to look as good as possible, it can be time-consuming. However, other times, images don’t come out how they would like them to, no matter what they do, so they will have to throw out prompts.
Many people believe AI art to be unethical. I got feedback from a few people that are generally against AI art. A few things they said were that AI art is too easy, it steals images from other people, it is brainless, AI art could replace fine art just as computers replaced graphic design, real art has to have intention, and AI art can further reduce art’s quality by democratizing it. I asked the representative what he thought of the ethicality of AI art, and he said that “when it comes to AI, we don’t see any issues with it because it takes millions of photos and creates something new out of it. Sure, it’s inspired by those things, but it’s not the same.” In other words, because AI art takes heavy influence from images but does not directly copy the photos, there is no way that someone could even prove that AI used their image in making the thousands of images used by the AI. The team isn’t making money on the art; they’re just making it. So it is impossible even to trace back ownership of images produced.
Because AI art has been growing, the team would love to collaborate with other teams. They see AI Art as a field that has just yet begun to be explored; however, it is prolonged on the rise. I asked the representative how far he thinks AI art will go and whether it could take over the entertainment industry. He said he believes that AI won’t replace the entertainment industry in the long run but could be a tool for it. There was an experiment in which someone asked an AI bot to rewrite a Harry Potter novel. The produced story was awful. There is still a need for storytelling from humans – all the AI is doing is using people’s other ideas and combining and changing it. Coming up with new ideas is very difficult for a bot.
When asked how the creative process plays into the technology, the representative said it depends on the computer’s wants. Sometimes, a person will have an idea for an image, and it will only take a five-word prompt to get the desired effect. Other times, however, the computer is much more finicky. “Say you want an image with a man sitting on a guide rail over the city. If you don’t tell the computer that the guide rail is on the side of the road, it’ll put it over the city.” Sometimes the opposite effect is true: If you give the computer too much detail, it’ll mess up, too. One strategy that the team uses is to tell each other their ideas for images to try to generate prompts with each other before they try them. However, some things the computer just can’t get – fingers, for example, are very hard for computers that don’t understand their functionality. Sometimes they just have to throw out prompts. There is no way to do it perfectly, and it’s like art because of the sheer amount of trial and error.
You can find Artificial Brush Stroke on Instagram at @artificial_brush_stroke, where they have added a new tab to allow for idea submissions. Check them out!
Average Rating