Artificial intelligence (AI) is a type of computer program designed to simulate intelligence. Different AI programs can do varying things. Some write essays, some analyze images, and some play chess. Most importantly, AI can learn. An AI program completes a task, and the humans supervising it give it advice. The AI takes the advice and reruns the task. It retries the task until it can be entirely accurate.

Recently, AI has become more accessible. AI models to write essays were released and promoted worldwide earlier last year. AI to generate images has been popular since 2021, but just recently has been under speculation. Multiple online programs to generate “AI art” were opened to the public last year. People began to generate images and claim they themselves were now artists. Members of the art community were very concerned about this sudden wave of AI images. Some began to request the banning of these programs, and others suggested boycotting them. AI designers and users tried to explain the nuances of the issue but were mostly unsuccessful. It can be very difficult to explain, as you need an advanced understanding of coding and artificial intelligence to understand how these programs work.

How It Works

Simon Gardling, 12, has done some experimenting with AI. He has worked with multiple kinds of AI before and has a strong understanding of the coding it requires. He explained the basics of these programs.

“These AIs use billions of parameters, which are the properties of neurons used in these models for AI art generation. Anyone could technically create such a model. However, AIs are absolutely massive and are no small feat to train… [The pictures they use are] from publicly accessible images online that are labeled with human created descriptions. The model then learns how to create an output based off of the input description,” he said.

Basically, these AIs are trained to look at images and replicate the features they see. They don’t just replicate the image itself. They replicate each individual pixel. They look at the common colors, figures, and designs of thousands or millions of images. They find the similarities between images and create something that sort of combines aspects of each picture. There is some debate on whether these AI programs can actually create anything new. The code is not as inherently creative as humans are. Artists and AI developers debate whether its work is a collage of human works or something entirely new.

Artists and AI aficionados can agree on one thing: “AI art” is not art.

“It is not art. Art is created with the idea of communicating ideas, feelings, and emotions. While AI generated art is called art, it is missing the aspect that makes art truly art. As long as AIs aren’t sentient, AI art will be only an imitation of [art],” he said.

Art is based on communication, human experience, and imagination. With the current definition of art, only sentient beings can create art. Vitalia Badarancea, 12, is an artist in Ms. Flint’s AP Art class. She also works with a publishing company to make book covers. She has a lot of field experience and knowledge with art, and she has strong opinions about it too.

“AI art is not art. AI can not make art… It does not have artistic value. It just steals art from other people. It is normal for artists to use references, but AI art does it without consent and without creativity. It is an unethical practice,” she said.

Is It Legal?

Artists worry that sampling art to train AI is unethical. They feel that their copyrighted works should not be used without their permission. As of right now, there are no direct laws blocking AI developers from using copyrighted works to train AI. Developers argue it is no different from artists looking at works to get references or learn about composition. Gardling feels that artists aren’t understanding the issue.

“Well licenses exist for a reason. I am sure licenses restricting use [for AI art sampling] will be used by many artists in the future. But in my opinion, it is as much as stealing as someone learning how to create drawings based off of a limited understanding of art and limited examples. There is also a legal argument. [These works could] be considered fair use due to the fact these models are a derivative work,” he said.

Derivative work is a creation that contains work from another creation. However, a derivative work needs to be different enough from the original work to be able to be copyrighted. The idea of derivative work is that it allows for people to use preexisting ideas but create something entirely new with them. Musical arrangements and collages are types of derivative works that can be copyrighted, despite the fact that they contain other artist’s work. Some developers argue that any image generated by AI is a derivative work. The actual definition of a derivative work in copyright standards is vague, so whether it can apply to AI work or not is unclear.

The Future of Art

While there is no legal precedent right now, most artists and developers see legislation about “AI art” in the future. It is unclear which way it will go. As AI gets more advanced and smarter, the line between human creation and AI generation will be indistinguishable. Badarancea works for a publishing company that is very cautious using AI generated images.

“I’m working for a publishing agency that does book covers. They are asking us not to use AI because they are afraid it will eventually fall under copyright… It’s one thing to use as a fun little thing on the side. It’s a whole different thing to have it for a commercial use. There should be some regulations,” she said.

“AI can be used as a tool, which is different than it being used to make art. The problem is with AI, it requires no skill. It’s different than artists using it as references. It’s not even comparable,” she stated.

Some artists use AI generation as a way to generate ideas or interesting compositions. They get inspiration from the AI image and create something new with it. Some have even trained AI with their own personal body of work in order to produce new pieces in their original style. The question is, is the AI image as valuable as an actual human work? Would someone be willing to pay equal prices for a human work that took hours in comparison to an AI image generated in seconds? This is the real reason digital artists are so worried about AI generation. They feel a threat to their trade. Why pay someone $100 for a commission when you can generate multiple images every day with a $5 monthly subscription? Traditional artists working with paint, paper, canvas, and ink are safe for now. AI cannot paint images in the real world, but it is not impossible for it to have the ability in the future.

The AI is getting smarter. Right now, the AI struggles with delicate features like fingers, eyes, and hair. Soon, these programs will overcome these issues. Eventually, human digital art and AI generated images will be indistinguishable. AI may even be able to create new styles or techniques. Right now, anyone reading this story can search up “AI art generator” and access dozens of free or pay-to-use programs to generate images. These programs are not going away any time soon. They are only becoming more prominent and more useful.