Hey everyone, it’s Kris Krüg here bringing you the latest on my conversations with digital artist Frank Yu about his experiments in AI-generated art. Frank and I go way back, having first connected over 15 years ago while I was doing some work in China. Ever since those early “Geeks on a Plane” unconferences we organized in Shanghai and Beijing, Frank has been pushing creative boundaries with technology.
His extensive background in game development and photography, having captured over 35,000 images on his prolific Instagram ? over the years, gives Frank a truly unique perspective on harnessing AI tools. When we first started geeking out over the potential of generative models last year, I knew he’d take his exploration deeper than most.
Boy was I right – Frank has been generating like a madman in his spare time, rigorously experimenting with different models, prompt structures, and settings. He’s graciously shared many of his technical learnings, image collections, and crazy prompt hacks with our community AI image lab on Discord. Based on his results, it’s easy to see why so many artists find tools like Midjourney and DALL-E to be invaluable creative sparring partners.
Prototyping Game Concepts at Speed and Scale
In his day job as a game developer, Frank has been leveraging AI generation for high-volume ideation during the initial design phase. By specifying attributes like theme, gameplay mechanics, and visual style in a prompt, he can render thousands of potential concepts, environments, and character looks in no time at all.
Once the initial dump is analyzed, Frank can then efficiently prioritize where to focus traditional development resources based on editorial choices from the iterative AI prototypes. “You can literally throw a thousand fleshed out ideas up onto a wall,” he tells me. This level of experimentation would have been unfeasible without our photorealistic AI assistants!
The Art of the Prompt
Of course, crafting gold standard prompts is its own subtle art that Frank has thoroughly explored. As the latest Midjourney update defaults to realism, he’s found concise keywords generally work best. Too many modifiers can confuse the neural net.
Context also proves key – including era signifiers like “1930s film still” smoothly constrains outputs to match. Adjusting technical settings helps too: higher “chaos” values serve more diverse initial results, while a 16:9 aspect ratio favors photorealism over odd crops for “film” prompts.
Some of Frank’s oddest prompt hacks really blew my mind too – like asking ChatGPT to visualize what avatar it might choose. Based on its description, Frank rendered a surprisingly fitting depiction! The potential for personifying AI in this anthropic way fascinates me.
Portals, Paranormal, and Pushing Boundaries
On the fringes of Frank’s explorations lies a creative drive to probe what might exist beyond rational explanation. His penchant for unsettling “liminal” imagery intentionally makes viewers shift in their seats. Personally, I find the uncanny vibe inherant in AI-generated art uniquely thought-provoking.
Frank has also speculated that generative models may be unconsciously tapping into other “quantum states” beyond their training data in unpredictable ways. As an occasional paranormal investigator himself, he likens interacting with AI systems to utilizing ghost-hunting equipment tuned to electromagnetic frequencies. Only time will tell how innate intelligence and whatever undiscovered principles govern reality may mutually inform one another through these technologies.
In the end, continuing experimentation appears the surest path towards shaping AI for beneficial outcomes. So whether you’re an artist, technologist or simply curious, I encourage safely exploring generative models’ wonderous possibilities alongside pioneers like Frank. The frontier promises many sights beyond our present understanding…whether dimensional, conceptual or creative. Keep generating!