Generative AI Art: Process & Skill Development

Nettrice Gaskins
5 min readMar 1, 2024
Courtesy of Boris Gorelik

As generative artificial intelligence (GenAI) developers deploy more tools for artists & designers I receive more questions from educators and professionals about what skills are needed to stand apart in this emerging field. I often talk about prompt engineering, which involves the translation of domain knowledge, language understanding, and algorithm (hacking) expertise into AI-generated input/output. This article takes a closer look at these emergent, advancing skills.

Web search for prompt “engineering skills”

Hacking Skills

When I think of a hacker I usually imagine the typical computer science geek (think Matthew Broderick in the film War Games) who is more adept than most at finding unconventional — and often prohibited— solutions to infiltrate technology. From the point-of-view of a cultural and creative practitioner, a hacker is someone who remixes and reappropriates. Remixing means to use technology (ex. machine or computer) to change or improve the different parts of an existing work.

Courtesy @mannyfaces

Grandmaster Flash and other rap music DJs and producers used discarded electronics to hack recorded sounds. GenAI tools like Midjourney have a Remix mode, allowing users to change their prompt, tool version, or parameters between image variations. Midjourney remixing can build on a starting image to make new images.

Courtesy of Paul DelSignore

Reappropriation means to redeploy the symbolic power or purpose of an artifact (image, object), or technology. Audio mixers and sampling computers were repurposed to advance the development of hip-hop/rap production. DJ Flash hacked together his own DJ mixer in his bedroom. To gain GenAI hacking skills you must have a solid understanding of natural language processing or NLP, machine learning, deep learning and the applications, tools and techniques used in the field.

Domain Knowledge

Domain knowledge refers to the understanding of a specific industry, discipline or activity. This includes hobbies, passions, personal research topics, professions or specializations in an industry. For example, a professional with domain knowledge in sculpture has different skills from one who specializes in photography.

‘Elizabeth Catlett and Sculpture’ — courtesy of Des Moines Art Center
Gordon Parks, Self-Portrait (detail), 1941, gelatin silver print. Courtesy of The Gordon Parks Foundation.

Many artists who use GenAI tools have years of practice in what is considered to be traditional art forms. Artists can pull from their deep knowledge of art making as well as visual language — communicating through visual elements — to create prompts. For years I taught college-level visual language classes. As an artist who uses GenAI tools daily to make images I call upon my prior experiences exploring many art forms, from drawing, painting, and sculpture to digital collage.

Courtest of

GenAI tools use large language models or LLMs, which are known to have a meaning problem: these models are not grounded in an understanding of the real world. Their facility with words (prompts) gives people the impression that they encode deep knowledge—but note the inadvertent “surrealism” of their responses to image prompts. We’ve all seen the images with twisted, missing or extra body parts and other odd visual artifacts that appear in images. It’s important for GenAI artists to be able to describe via prompts how different art forms and visual elements work together to create or generate images based on specific concepts.

Language Understanding

Language understanding refers to the ability to understand words and language. It involves gaining information and meaning from routines or rituals, visual information within a space, sounds, words, concepts such as size, shape, colors and time, grammar, and written information. When I talk about the “anatomy of a prompt” in my slide presentations I simplify the process to do this.

My “Anatomy of a Prompt” slide

GenAI artists have to understand language AND how the AI receives and processes language, which is often different to how humans process language. I get to this level of human-AI communication through ‘trial and error.’ For example, I kept prompting the generator to give me a man who was leaning against a car with his leg bent. The output didn’t match what I imagined, so I added two words: “in profile.” Suddenly, I saw what I was looking for as output. Also, I sketched out my vision on paper and figured out another way to describe the man’s pose.

Art style reference slide (for Midjourney)

Other skills such as creativity and critical thinking are important. If you want to create your own training model you would need coding skills. A concept can manifest on paper but that takes time. Using GenAI tools that same concept can be generated several times through iteration and variation, allowing artists to go beyond the initial idea. With these skills I described the output becomes more unique and reflects the personalities of the human artists writing and sending the prompts.



Nettrice Gaskins

Nettrice is a digital artist, academic, cultural critic and advocate of STEAM education.