Continuing the discussion from SketchUp Diffusion Feature Requests:
Diffusion appears to be using raster image based AI to detect shapes and generate images. It would be amazingly cool if you could take Diffusion to the next level by enabling it to understand the 3D geometry. That’s quite a bit different from raster based AI image generation and, undoubtedly, would require a lot of work. But it’s what SketchUp needs.