Raw LLM Responses

@RandomYTubeuser Lets stick with drawing. Lets set a low bar. I would define an average artist as someone who has understood the fundamentals of drawing. The AI already outperforms this artist in rendering quality. But what this average artist already can do better, is understand what I want him/it to draw. With that average artist I can have a back and forth with a couple of sketches in between and in the end he will draw exactly what I want. Granted it is not looking great, but try this with an AI. It will change unrelated things. It will refuse to place stuff where I want it. It will ignore half of my requests or misunderstand what I meant, no matter how well I describe it. And the reason is always the same: It is limited by the images it has stored. The more you deviate from those with your wishes the less you get what you want. Edit: just tried it on ChatGPT "Give me an Image of a kangaroo please." -> really good image. "Now spray a green circle around its tail. Not as a contour but wrapped around the tail." -> the circle is not on its tail. instead its a hoop around its legs with a fur texture. Why?! Because it has a lot of images of people in a hoop. And these go around legs and hips. This is not so easy to train away, because for these things to work it needs to understand what it is being asked, instead of finding some interpolation that might fit your prompt. Coming back to my original point: The sad thing is, the same technology would probably be really really good for searching images. If an artist is looking for reference, how great would it be if you could describe what he are looking for, no matter how niche. It would find all the things. Instead we got this thing that can't be used for professional work. The suits will try anyway and we will have to suffer through a period of products with mediocre art, till even they realize that it was easier to work directly with artists instead of giving them AI images to fix and rework.