Writing Imagery: Multimodal AIs in History and Digital Art
Daniel Hutchinson and Erin Jensen
Belmont Abbey College
We introduced undergraduate history and digital art students to multimodal AI models through the practice of prompt design. Students learned techniques for developing textual descriptions for generating images using models like DALL-E 3, Midjourney, and Stable Diffusion. Students then applied these techniques to explore art history through prompting of specific artistic styles on subjects of their choosing. After experimenting with these technologies, students then reflected on the broader social implications of AI-generated imagery.
Learning Goals
- Introduce students to writing textual prompts for multimodal AI models (DALL-E 3, Midjourney, Stable Diffusion)
- To explore the history of art through student-created imagery
- To engage issues of digital literacy concerning the ethical and social implications of AI models
Original Assignment Context: Sections of a fine arts “Introduction to Digital Art” course and a core curriculum history course “Western Civilization II.”
Materials Needed: Access to an AI image model like DALL-E (OpenAI), Midjourney, and Stable Diffusion.
Time Frame: One week.
Overview
Multimodal AI offers instructors an engaging approach to introducing students to an emerging digital art form where users can visually represent the vast datasets on which they are trained. Models such as DALL-E, Midjourney, and Stable Diffusion have enabled the generation of images of increasing sophistication and with significant commercial potential. These AI models can adeptly mimic various artistic styles, genres, and mediums. Crucial to using these models are a distinct form of writing: textual prompts. Our courses used two approaches in teaching this practice: 1) Instructing digital art students on creating prompts, and 2) Applying these approaches to explore the history of art and the ramifications of AI-generated imagery. Our objective is to cultivate among students an informed understanding of these technologies and their potential impacts on society.
Teaching Digital Art with AI Models
Students were asked to read about using AI art models and then create their own textual prompts to create their chosen images. Students were encouraged to keep trying different words in the text prompt to see how the AI images would change according to the words that were being used. Students were encouraged to use various free AI programs including DALL-E 3, Midjourney, Deep Dream, Night Cafe AI, and Stable Diffusion. Students then reflected on their experiences in creating the images.
A student in the class was writing a children’s book that stars a golden retriever puppy going on adventures. She decided to use Night Cafe AI to create all her images for the book. As she wrote in her reflection, “There is nothing more exciting than when your dog runs towards you in excitement so that was the look I went for. After picking my idea, I just had to experiment with the wording to get the images that I wanted and could use in my book. I like that I used my words to create this image and the rest of the images for the book.”
Another student in the class had spent her summer on a study abroad in Ireland. She was creating an Instagram collage and wanted to include AI art with her actual photos. She reflected on what she created by writing, “The description I provided the website with was: ‘Wild waves and wind off the coast of Ireland with cliffs in the background.’ The AI image that was created looks similar to the pictures I took of the ocean off Ireland and the project looks really cool to have both AI and real photos being used.”
Another student wanted to try and create as many “cute monster” AI images as possible. She described her experiences with AI in her reflection as she wrote: “I must start by saying don’t knock AI Art until you try AI Art. This assignment was so much fun, and I decided on AI due to how creative I could get with my ideas; most of the things I have heard about AI art are how it can be harmful or how people fear it. I began using Deep Dream Generator and developed some pretty graphics that mainly originated from entered texts (Monet, flower mandalas, during a golden hour with butterflies, smiling monster). This created some beautiful images with a Claude Monet feel, and the more specific I was with my language, the more detailed the image became and vice versa. The outcome was magical and extremely cute.
Students then reflected on their thoughts about AI art and the implications of being able to generate art through the use of keywords. Most students either embraced AI art and thought this experience of creating AI art to be “brilliant,” or thought “AI art is unethical and I never want to think about this again,” or took a more middle group approach with “while I can see the ethical issues, I am excited to go try to create AI art for various projects of mine and will continue to evaluate the ethical issues surrounding such art.”
Teaching Art History with AI Models
In an introductory Western Civilization course, students employed multimodal AI models to navigate the intersections of technology and creativity through prompt writing. Focusing first on the Industrial Revolution, we examined how emerging technologies like photography influenced artistic practices. This set the stage for discussing contemporary parallels with 'AI Art.' The instructor guided students in the practice of composing effective prompts. We explored how the AI represented specific mediums (e.g., oil painting, pencil drawing, drone photography), subjects (a celebrity, familiar landscape), and artistic styles (Impressionism, Salvador Dali, pointillism).
To demonstrate, Hutchinson generated an in-class example using the prompt: “An oil painting of the Charlotte North Carolina skyline at dawn in the style of Vincent Van Gogh.”
Generated using DALL-E 3
Anyone wishing to generate their own images were then invited to share their prompts. Below are a selection of those images:
Oil painting of a sunflower drawn in the style of Salvador Dali
Oil painting of Steph Curry by Van Gogh
Painting of a girl in a church pew in the style of Monet
The skyline of New York City from a bird’s eye view
The class then discussed the broader implications of these technologies. We noted how the generated images reflected the style of particular artists and genres, and how such models can be used for exploring and remixing different periods in art history. We also discussed the potential impact of these models on contemporary society. We debated how to weigh potential benefits (greater productivity, empowering personal creativity) against potential drawbacks (impact on commercial artists, privacy and copyright implications, dataset bias, misinformation via “deepfakes”). At the end of class, students were asked to respond to the following question: “Based on our discussion and experimentation with AI image generation today, what is your opinion on the potential benefits and drawbacks of these technologies for society?” While students noted the potential benefits of these technologies, they also consistently expressed ambivalence:
- “I think the potential benefits that AI allow us are unimaginable and it seems as if there is no ceiling right now. If you can express anything in words or think of it, AI can generate an image….Artificial intelligence can help us in many areas of life and will continue to be routinely used within our lives the more it improves. “
- “A benefit is that it will take less time and be easier to create an image and express yourself in words. A drawback is that there is no emotion in the image and it doesn’t provide you with feeling.”
- “It is very creative. For people who can not draw, this allows them to tap in their artistic side. It also makes beautiful pictures in a short amount of time. However, craftsmanship of a man made product is something it can not compare to. “
- “I think this is bad and will take jobs and skills from people who have an amazing talent.”
Teaching with multimodal AI offers students an opportunity to meaningfully engage in these debates while providing them with effective instruction in using this emerging technology. Effective writing will prove a key skill for this new skill set, particularly as these models improve. DALL-E 3 now uses chat interfaces to allow iterative refinement of both prompts and images. Such approaches empower students to harness their creativity and apply their historical knowledge to transform abstract concepts into visual narratives. Prompt design thus not only preserves but elevates the art of writing for both creative expression and academic inquiry.
Acknowledgments for any assignment, collaborator, curriculum designer, researcher or publication that influenced the assignment:
None
Assignment 1: Creating AI Image for a Specific Purpose
Think of a project you are creating in your life where you may want to generate an AI image to include. This may be a collage of pictures for an Instagram post, creating a stock photo for your internship, use on a website for your final project, illustrating a creative writing project, or any other project that you are working on for school, work, or your personal life. Please make sure the images that are being created are appropriate to the standards of our college.
Using any of the AI art programs discussed in the readings, lecture, or that you know about. I would recommend the use of DALL-E 3, Midjourney, Deep Dream, Night Cafe AI, and Stable Diffusion. Try various keywords to create a text prompt and see how the changes in words impact the image being created.
Please turn in the image you created and a paragraph reflection on what program you used, what your experience using the program was, what text prompts you tried, and how you plan to use this image in your project. Any other thoughts or reactions you have to create AI images would also be welcome to be included.
Assignment 2: Exploring Art History Through AI
Gather with your classmates to try out an AI Art model. Provide a text prompt and we'll generate an image.
Here's an example prompt style: Medium + subject + artistic style
Here's an example prompt: “An oil painting of the skyline of Charlotte North Carolina at dawn in the style of Vincent van Gogh.”
Consider the following artists and concepts mentioned in the lecture. Mix up the mediums to explore different combinations.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Follow-Up Question: Based on our discussion and experimentation with AI image generation today, what is your opinion on the potential benefits and drawbacks of these technologies for society?