I genuinely miss when AI was schizophrenic, now it's just boring slop and not even the good kind with leftover stew from the farmer's wife. Until we can use simple 3d environments to scene build and then use AI to insert objects we'll be limited by what random images get made and integrated into imagegen. Like imagine you have a bunch of general shapes, like a chair or table or even a house/room with windows and position the camera in 3d space and generate the image based on that result sorta like how controlnets work. That doom AI was kinda close but I don't think what I'm referring to exists yet