Given we have engines that can render complex 3d worlds, can maintain consistency far longer than a minute and simulate physics accurately, why put all that bur

Given we have engines that can render complex 3d worlds, can maintain consistency far longer than a minute and simulate physics accurately, why put all that burden on a GenAI world generator like this? It seems like it’d be more useful to have the model generate the raw artifacts, world map, etc. and let the engine do the actual rendering.