Generative art is art, AI art is shopping.

So when I am asked what is the difference between generative art and AI generated art I say generative art is real art, while AI art is just shopping. Both forms use computers and code but there is a fundamental difference in how the code works and the role of the artist.

With generative art the artist is creating the code. The type of coding used may vary depending on what development environment the artist is working with but ultimately the artist is writing the code with the intention of producing a particular visual result. Furthermore the artist will in most cases tweak and re-work their code multiple times until they get a result that they are satisfied with. The artists eye is part of the process.


Now let's look at AI art. You enter some text based inputs and are returned a series of generated images that should match the input you entered. Imagine for a moment you are in gallery and ask the owner to show you all the paintings they have of dogs. They return with four watercolours of canines and you purchase the one you like most. That's not art, it's shopping.

The person entering the input is simply asking “I would like this” and the AI software is simply delivering them the product they asked for. If this is art then you could argue that Google search results are also art.


OK, let's go back the shopping at the gallery analogy. You could ask “hey, what about the artists who painted the dog watercolours?” Maybe the person entering the input isn't the artist but the AI software itself is the artist. However there are couple of key differences between the real world and the AI world in this analogy. In the real world the four pictures of dogs were created by artists to satisfy their own creativity. In the AI world the pictures created by the software are generated in an attempt to accurately match the descriptive input. There is no creativity in the process. It may look 'arty' if the source images look 'arty' but ultimately all it really can be is a very sophisticated collage.

Have a look at the cat and dog on the left. Both are AI generated images and both look pretty realistic. Cat's are easy for software to generate because there are billions of cat images on the internet to use as a source material and ultimately cats all look much the same. Dogs are a little harder because of the wide variety of dog breeds. Look closely at the dog at it doesn't look quite right. The whiskers seem a bit off and there is some sort of extra tail at the back. But given time AI generated images will get more and more accurate. As the number of source images increases and the user input increases the software will get better at creating an image that looks exactly like what has been described.

---

To the right is an AI generated image with the input “robot frog eating ice cream”. It looks odd. The eyes aren't very raised, it seems to have ball ears and looks like it has been hit in the head with the ice-cream rather than eating it. It isn't really a robot frog and it isn't art. It's a mess, but it will get better. In a few years the same input will probably deliver an image that actually looks like a robot frog eating ice-cream. But it won't be art. There won't be any creativity. It will always give you the most common looking green frog, the ice-cream will almost always be in a cone and the robot side of the things will most likely make it look metallic. An artist might choose to represent a spotted frog, eating ice-cream from a tub with realistic skin but robotic limbs. The AI isn't going to be creative because that is not what it is meant to do. You ask if for something you want and it delivers what you asked for.

It's shopping.