AFAIK we already develop games with maximum quality assets and scale down to meet average gaming machine. Most textures used are at least 16K with billions of polygons in a single character. The upscaling tech is used on the gamer machine not on the developer machine.
The game engine already dynamically optimize mesh and textures by scaling down resolution and loading low mesh count assets. But that is still not enough scale down for most gamer machines.
Frame generation is not a problem in the gamer side. Its all perception.
I play cyberpunk 2077 at 130 fps on overdrive at 1440p and its awesome and smooth with a RTX 4070. I play on a hisense 55" tv at 11 feet on my sofa and I don’t notice any artifacts. Its just way better than classic SSAO and other global illumination “hacks” that raster uses.