Exploring creative problem spaces with Gen AI

The advent of Generative AI (Gen AI) has introduced a plethora of new use cases, significantly enhancing our daily office and work tools. These AI features are now integrated into text editors with predictive text and code completions, image editing software with generative capabilities, and presentation tools offering AI-generated templates. We are witnessing the initial wave of these innovations, with substantial developments anticipated in the near and distant future.

In my personal time, I have been experimenting with Gen AI tools for music, sound, voice, image, and 3D model generation. I am currently developing a 3D action-adventure video game featuring a walrus, utilizing Gen AI to generate most of the assets. Anyone familiar with game development understands that creating game assets—such as graphics, sprites, textures, models, level designs, and sound effects—consumes a significant portion of development time.

Through my experimentation with Gen AI tools for creative projects, I discovered a novel use case: exploring creative problem spaces. These tools excel at helping refine final ideas and designs by presenting multiple versions of an initial concept. This capability enables creators to iterate through the design space more efficiently, ultimately producing superior results.

For instance, let me describe my process for creating a non-player character (NPC) for my walrus game using OpenAI’s latest version of DALL-E.

1. Conceptualization: The process begins with an idea. I needed an NPC for my game who assigns a quest in a lighthouse involving the replacement of a lighthouse lamp. I named this character William.
2. Prompt Creation: I described in my prompt that I wanted William to be a lighthouse keeper, an anthropomorphic goat with a long tailcoat and a scraggly beard. I submitted this prompt.
3. Initial Outputs: DALL-E generated two images of a lighthouse keeper—one with an eye patch and another with a hat. One had an olive coat, the other a black one. Analyzing these images fueled my creative process: I decided I wanted the character to have the olive-colored tailcoat, the eye patch from the other image, and a similar hat. I refined my prompt and resubmitted it.
4. Refinement: DALL-E produced two more versions of William. The first image depicted William as I envisioned: a big scraggly beard, eye patch, hat, and long olive-colored tailcoat with boots, but he was not facing the camera. The second image was similar but had incorrect expressions and details. I adjusted my prompt to have William face forward.
5. Finalization: DALL-E finally returned an image of William that met my creative requirements.

This example illustrates how DALL-E facilitated my exploration of the problem space for “the best design for a lighthouse keeper NPC.” Steps 3 and 4 could be repeated indefinitely until the creator is satisfied with the results, allowing for deeper consideration of stylistic and aesthetic goals.

I hope you find this method of using Gen AI tools for creative purposes insightful. It is just one of the many ways these tools can be leveraged in the creative process. I believe Gen AI can be an incredibly powerful tool for game developers and anyone interested in the creative arts and design. Its ability to quickly visualize and iterate on design ideas makes it ideal for iterative design.