“AI art” made headlines recently with the $432,500 sale of a “painting” created by a generative adversarial network. Just for fun, I’ve collected a few additional “artistic” applications of AI.
Writing
Botnik trained a neural network using text from the Harry Potter series, and then generated a chapter from “Harry Potter and the Portrait of What Looked Like A Large Pile of Ash.” J.K. Rowling would be proud.
Students at Duke recently won a competition as part of the 2018 Turing Tests in Creative Arts for a poetry-generation algorithm.
On this website, bot or not, you can guess whether the poem you’re reading was written by a computer or a person.
Video
Here’s a weird abstract video made using DeepDream, a computer vision program based on a convolutional neural network.
For a more “realistic” (?) short film, you can watch “Zone Out”, which was written, directed, and produced using AI techniques.
Art appreciation
Google “Quick, Draw!” will attempt to identify what you’re trying to draw:
Style Transfer
In style transfer, the style of one piece of art is transferred onto another piece of art.
- This page has some beautiful animations showing style transfer.
- Page 5 of this article shows the style of one painting transferred onto another painting.
- I like the jacket in Figure 8c of this style transfer paper.
Visual Arts
Duke University recently held an “AI for Art” competition. You can view some of the winning entries here, at the Duke Forge blog.
Do you love cats? You can turn any drawing (ANY drawing) into a cat using edges2cats:
A recent face generation model based on generative adversarial networks can produce faces so realistic that most people can’t tell the difference between the generated faces and real people.
This website allows you to colorize back and white photos automatically using deep learning. Here’s one famous picture, colorized:
(original black-and-white image from here)
Here’s a gallery of AI-generated art, primarily based on style transfer.
“cloudpainter” won the 2018 Robot Art Prize. You can also browse the Robot Art website.
Nvidia recently announced neural-network-based software that will transform a simple sketch into a photorealistic landscape.
Music
A recent Google Doodle will automatically create a Bach-like piece based on a melody line of your choice. It’s based on a machine learning model trained on 306 of Bach’s chorale harmonizations.
A variety of websites including Jukedeck, AIVA, and Computoser will generate music using AI. Here’s an example from AIVA: “On the Edge.”
Food for thought
Google Dictionary defines art as
the expression or application of human creative skill and imagination, typically in a visual form such as painting or sculpture, producing works to be appreciated primarily for their beauty or emotional power
In all the above examples, humans created and trained the algorithms. What’s the difference between a paintbrush and an algorithm? What’s the difference between a programmed laser cutter and an algorithm?
Will we ever reach a point where a machine “independently” creates art? What are the differences between humans teaching machines to create art, and humans teaching other humans to create art? After all, even famous human artists had to learn their craft from somebody. Will machines ever teach other machines to create art?
Glass Box 
Comments are closed.