Assignment: Artificial Creativity


Starting from the output of a machine learning system, add a “human touch” and produce a unique physical artwork, or a series.

Some examples of human interventions:
Using an analog printing tecnique such as cyanotype, riso, screenprint etc…
The generated image is a sketch for a traditional process: oil painting, pencil drawing, etc…
The images are printed on fabric and a item of clothing or accessory is made with it
The images are used for an original artwork made of paper: pop up book, board game, card game, origami
The images are used as illustrations for a piece of sequential art (graphic novel, illustrated book)
The images are printed and disseminated in the real world (sticker, stencils, wheatpaste, etc…)
The images are edited into a video.
A generated text is used as a basis for a performance.
The process of creation of images is social and/or game-like. It follow specific rules you designed.
The process and findings are presented as a slide lecture.


You will be presenting the raw output from the AI along with your final project.
The delivery and critique formats depends on the medium.

If the artwork is analog and portable bring it to class but also create digital documentation.

Write a project statement: find a conceptual angle, find a context (where does the artwork exist in the world?), situate the project in relation to history and contemporary art making, differentiate it from the thousands of users who just generate cool looking imagery.

Grading Criteria

You used the generator to create something that doesn’t already exist in the world, something you can’t find via google images, something that can only be conceived by an artificial intelligence, something that is not a bad imitation of existing imagery.

You made an artwork that interrogates our relationship with artificial intelligence, foreshadows new creative processes, advances your own practice. You didn’t just make something that looks cool or weird.

You used a human-centered process that makes sense for the concept and issues you are exploring.


Art Breeder


Parameters manual
eg you can change aspect ratio by adding –ar 3:2

some easily recognized styles you can add as modifier

Prompt sharing site


Chat GPT


Text-to-pixel art

Remove objects from pictures

Story starting from a theme

Set of photoshop-like tools using ML like upscale, remove background etc

Blender + stablediffusion – make a 3D sketch, filter it through stable diffusion (advanced)

Runway – design and video-oriented tools (subscription based, free trial)

Stable Diffusion + text inversion – fine tuning models (advanced users)
*Even more problematic, ownership wise

Animation / puppeteering tools

Warm-up exercise

Pick two of these exercises below, make them more specific and personal. Come up with prompts that can generate such results. Iterate, and create variations. You should try at least 10 generations for each of the exercise.

Your goal is to create a set of compelling and visually consistent images without the typical glitches and artifacts of AI generated images. 

– Create 3+ stylized self portraits
– Create 3+ portraits of famous people aged, de-aged, gender swapped, stylized
– Create 3+ graphics for a collectible card game (eg. Magic the Gathering, Pokemon) or a tarot set- photoshop them into a template, including name and description
– Create 3+ postcards from the future – write the message on the back too
– Create 3+ motivational posters using generated graphics
– Create 3+ images for a children book, include story bits
– Create 3+ credible images of cryptids, ghosts, UFOs or other exceptional phenomenon
– Create 3+ stills from an imaginary cartoon
– Create 3+ album covers from imaginary bands (you can photoshop text)
– Create 3+ illustrations showing different subjects but belonging the same fantasy world
– Create 3+ “paintings” belonging to the same imaginary artist, give them titles
– Create 3+ book covers for existing novels (photoshop title and author)
– Create 3+ images representing an imaginary style, trend, design movement
– Create 3+ graphics for an imaginary product (label for draft beer, cereal box) – you can add text manually
– Create 3+ ads for an imaginary product or service – you can add text manually
– Create 3+ credible photographs of a past event or situation that never occurred
– Recreate/reimagine 3+ famous artworks
– Create 3+ new meme templates or iterations of an existing meme

Post the images on discord along with the prompts that created them and a one line explanation (ie what the exercise was).


Student Work

Jennifer Shin asked her mom to describe memories of her experience in the US in the ’80 she didn’t have pictures of. She generated credible polaroids and collected them in a hand crafter photo album.


Tristan Hineman asked chatGPT to write a children book about an AI taking over the world and then illustrated it with midjourney-generated imagery.

Audry Stewart created a lifestyle instagram account from a bizzarro world in which most artificial things are quilted

/Imagine A work of iconic art by Ryan Chey
I ask MidJourney to reproduce the various famous artworks throughout history by describing the visual qualities of the work in ridiculously obsessive detail while avoiding actually mentioning the title of the artwork itself, the name of the artist, the movement or genre it’s associated with, or any other references to “reproduction” or “recreation” as to allow for MidJourney to use its own resources instead of directly referencing the original work 1:1. Was I successful? We’ll find out.

AI Doppelganger: Who’s more ME?
For this project I trained a roleplay-based AI to act/ talk/ think like me to the best of its ability. This was done through feeding it a mixture of my text messages, tweets and other personal dialogue.
The result will be presented in the form of a presentation/live Q&A guessing game–in place of a proper Turing Test– where classmates can guess which answers are mine and which are the Ai’s.

Kelly McDowell hid AI generated printouts of flowers and hid them in the campus

AI Spy by Annie Chong
Seized by the nostalgia for my childhood books, I took a try at replicating the object searching game, I Spy. In collaboration (and sometimes contestation) with Midjourney, I created a series of images in the style of the I Spy Scholastic books (written by Jean Marzollo with photography by Walter Wick). I then supplemented by writing my own riddles to match the generated photos. I hope to highlight the interesting oxymoron of AI images being best appreciated and unnoticed when being glanced over, and I Spy games requiring very thorough investigation of its images.

Do AI Deepdream of Electric Sheep?
When AI generated images were very new, I was fascinated by Deepdream reimaginings of images because of how the computer would create abstract and strange artifacts based on patterns that a human would never pick up on. Nowadays, the field is far far more developed and it’s creativity mimics that of humans. However, I was still curious to see what kind of strange “creativity” New AI could come up with. For this project, I wanted to have the AI create unexpected images with almost no human prompting, so I used old images by Deepdream as very noisy inputs, and used as few keywords as possible to generate new, strange iterations of these images. Basically, these are AIs dreaming back and forth between each other.

Assignments from Golan Levin’s session, he’s making students create a book or publication:

Most of it is interesting only due to the novelty of the tool, but the Alien Playboy issue is great: