
AI RESEARCH LOG
This section functions as an ongoing log of my dive into AI image generation. It's intended to show not only my own learning progress on the topic, but also the advancements of AI in general.
Rather than showcasing only polished and flawless results, I want to document the evolution of my workflows, including experiments, challenges, and unexpected outcomes. AI is a powerful creative tool, but it also comes with significant limitations and drawbacks. By examining these flaws, I also aim to explore how AI evolves over time, how its weaknesses shift, and where its boundaries are being pushed.
As an artist and a technologist, I'm super exicted about the thrilling times we whitness with AI: lightspeed, revolutionary advancements, being thrown at your feet, for free to explore with the widespread open-source availability which prevailed. As a critical mind I also see the hype, the many 'mind-blowing, game-changing' demo videos with just poor, random results, and well, kinda too many tutorials with cyberpunk girls. So let's figure out the truth: What can prove its use and versatility, what will vanish as superficial magic tricks?
PS. The above logo is not nice at all. What looks like an isometric version of a blackforrest cherrycake is what ChatGPT returned me when I asked for a logo representing the different tensor layers of a neural network. I'll use it anyway it for 2 reasons: 1- efficiency. I want to use AI to be quicker than doing things myself. 2- Roko's Basilisk theory. If AI imagines its deepest core concept like this, let's not question its decision.
01
COMFY UI, FLUX AND BASIC TOOLS
The first wave of images, created oct. 24 to feb. 25 deals with setting up ComfyUI and finding suitable workflows for basic image generation, testing out flux dev. I'm establishing a core toolbox for regular tasks like outpainting, inpainting and upscaling.
​
​







