What is AI art and is it okay to like it?

When a submitted piece of art created in part through artificial intelligence [AI] won first prize at the Colorado State Fair in 2022, people took notice.

Théâtre D’opéra Spatial by Jason Allen via Discord

It was not like the use of computers or technology to create art was new, but AI brought something new. No longer were tools being used simply to help bring an artist’s idea to fruition, these new tools were now being used in the process of conceptualization.

But before we hit the panic button on whether AI art will replace traditional human-made art, we need to dispel a few misconceptions.

What is AI art?

AI art has become a catch-all for a wide swatch of new art. And yet, not all of digital art is AI art. In fact, a lot of what is included in AI art only uses AI for a very select part of the process.

Let’s step back further and answer an earlier question: What is AI?

As its most basic definition, AI is the use of computer science and large datasets to solve problems.

This is essentially how our own brains work. Each one of us has a wealth of knowledge, accumulated over the length of our lives. In any given scenario, we use the knowledge we already have to help us either solve the task at hand or to point us in the direction of where we can find new knowledge to help.

AI art does this as well, but it only operates within the field of art. This is purposely restricted because it falls with a specific branch of AI, called generative AI.

By itself, generative AI is inert; it does absolutely nothing. In order to generate anything, it requires an input. The human artist inputs keywords or offers sample pictures (depending on what the specific AI art generator requires), and it generates a piece of art based on those inputs as well as the dataset it has available to it.

Sound simple? You (the human) give a machine raw material. The machine spits out a finished product based on what it has been told to do.

But that wins stuff?

Yeah, AI artists have started to win art competitions. But this is where we start to separate the accomplished artist from the amateur.

Let’s go back to the Colorado State Fair. The piece that won in 2022, Théâtre D’opéra Spatial by Jason Allen, did much more than put keywords into an AI art generator.

First, he did this enough times to generate over 900 individual renderings for him to choose from. From this, he chose his favourite three. Allen then used Photoshop and Gigapixel, two computer applications used by many digital artists today, to get the picture he wanted.

This took Allen roughly 80 hours, quite comparable to the length of time many artists take in more traditional media.

This shows two things. First, AI was only used in the first step of the process. It helped Allen to conceptualize his idea. Second, Allen used many of the tools and processes digital artists use in order to achieve his vision.

Now, could he have ended his work after the AI generator spit out the first rendering? Sure. But he did not. He put significant time and effort into his finished piece.

But when he won his prize, much of the conversation was about whether what he had created was truly considered art.

Criticisms of AI art

There are many criticisms of AI art, but we will try to boil them down to the broader ones.

First, critics argue AI art is simply an automated process. Because it removes human skill and talent, it cannot be an art form. Second, if we can agree that AI can help in the conceptualization process, then it is merely a stepping stone in helping an artist realize their true idea. But it is not art in and of itself. Third, AI art generators are so simple that anyone can input keywords. But that should not mean that it was created was made by an artist.

Interestingly, these are very similar to the criticisms used historically against another art form: photography.

It took many years for photography to be considered fine art. Photography was considered a way to capture an image in order to take it back to the studio so it could be painted. The photograph was not the work of art; it merely served as the sketch.

And when cameras became cheap enough that every family could afford one, critics argued that it could not be considered art because anyone could produce an image.

And yet, prestigious exhibitions and competitions are held all the time now. To say that photography is not art would likely get you some odd facial expressions in return.

Other issues with AI art

So, let’s assume AI art surpasses its critics and becomes known as a legitimate art form, just like photography before it. Problem solved?

Not quite.

For this, we have to go back to the first question we posed: What is AI?

Remember the datasets I mentioned? Yeah, those may end up being a real problem. A real BIG problem.

Datasets are essentially libraries of information. When you enter keywords in an AI art generator, it uses its dataset to find similar (and existing) works of art that find what was inputted. It then created a new rendering based on the keywords and the works of art it pulled from its dataset.

But is that legal? Well, we are currently in the process of finding out.

Art is intellectual property. That means it is owned by someone. At the point of creation, it is owned by the artist.

Within the Canadian Copyright Act, copyright of intellectual property exists for a term of 70 years after the death of the creator.

This means I could take the Mona Lisa (not the actual painting, but you get the idea) and reimagine it however I want in a new piece of art. However, I could not do that for, let’s say, Andy Warhol’s Campbell Soup Cans. For that, I would need to wait a few more years.

Now, within an AI art dataset are millions (or greater) of works of art. How do we know what it is pulling from is free of copyright? How do we even know what pieces of art it is pulling from at all?

Of the several lawsuits surrounding AI art, the largest would be that brought forth by Getty Images, the stock photo company. Getty has sued the creators of Stable Diffusion, one of the primary AI art generators, claiming Stable Diffusion has copied over 12 million images from their database without permission.

The potential cost of this lawsuit? $1.8 trillion US. That’s some serious money.

This, and many other, lawsuits are just the beginning. Copyright laws around the world are ill-equipped to deal with how fast processes around intellectual properties are changing.

AI art is just part of a much larger framework of intellectual property.

So, what does this mean for me?

Well, it does not have to mean anything if you do not want it to. As an artist, it can be a new tool to realize your idea.

That being said, AI art is likely to follow the same path as photography, only faster. It will be recognized as a legitimate art form because there are people who enjoy it as such.

The legal questions will eventually be ironed out because they have to be. There is too much money at stake for them to be ignored.

And as an admirer of art, you can look at it as having a new art form to add to what you can enjoy and own.

By: Nathan Durec