Various experiments involving neural image synthesis. These were done with ruDalle, Midjourney, DALLE-2 & Stable Diffusion.

I'm not finding it so interesting to generate "art" with these tools; I prefer to generate "documents" or "objects". Something that is more about capturing a made-up reality (like a photographer/curator/gardener) than imitating artists (although actual captures of reality in the datasets are also due in part to artists). I will find a way to word it better some day.

There is definitely something exciting in experimenting with these tools, but I also can't really ignore how ethically questionable they also are, nor the ridiculous amount of processing power they do require. So I have pretty mixed feelings about all this. I don't feel like investing too much time and energy or myself into all this, and I believe one should only do it in a careful, responsible way that avoids blind fascination for the tech. I haven't fully figured it out yet.