AI Generated Art
What do you think when you look at the painting above? Does anything feel off to you? What if I told you no living person painted the picture?
In recent months, art created by artificial intelligence (AI) entered my radar through social media – mostly in the form of short form videos created by Ai to mimic popular movie genres. While interesting, the videos were mostly funny in the way that the AI didn’t really understand our movies and parroted them back with a bit more absurdity. When I began seeing AI art in the form of photos, my intrigue rose further. These images were so much more human, sometimes eerily so. Investigative as I am, I began looking into how these images were being made – curious to try it for myself.
Artificial intelligence is the ability for computers to learn, think, and perform human-like tasks. This ability has been created and programmed over countless years by numerous dedicated individuals, leading to the development of artificial neurological systems that mimic our very own. This artificial neural network acts as the “brain” of the computer and allows the users to input data with a multitude of outcomes, and, in our case, the production of an original work of art.
The most convincing AI artwork has been created through a variation on the GANs system. Generative Adversarial Nets (GANs) are a generative model artificial intelligence program that operates by having two programs with one attempting to dupe the other, making the two programs adversaries. If the second program is successful in creating a dupe, that dupe is the creation generated by the GANs. How this functions successfully in terms of art is best described by Art AI Gallery:
“Our Generator learns separately about style and content, which allows it to interpolate between styles and mix style and content in novel ways. This is an algorithm that can generate new original images from scratch. Training this kind of AI requires the training of two separate neural networks – the “Critic” and the “Generator”.
The Critic is given a vast database of human art in different styles from throughout history as input to analyze. The Generator, which has never “seen” art before, gets a random seed as an input and starts generating an image from scratch. The output then goes through the Critic, which based on its knowledge, dictates whether or not the generated image looks like art made by humans.”
The program’s aim – to make a piece of art that could convincingly be made by humans – is the primary reason many successful pieces of AI artwork have an uncanny feeling to them. And while the program generation may trick the AI critic, our own human inner critics have their good friend human intuition telling them that something is amiss.
Luckily for those with limited math skills and lacking the ability to program or code, more and more people are sharing their AI art programs for our exploration. The Wombo Dream app contains one of these programs and is shared with the public. Of course, upon this realization, I immediately went to work punching in words and phrases to see what I could get. Here I am pleased to show off some of the art Wombo Dream app created with inputs I provided.
By Tierra Deacon