You're not wrong. The early 2000s splatting from point-based rendering is what George Drettakis and his students realized could be applied in this new NeRF dom

You're not wrong. The early 2000s splatting from point-based rendering is what George Drettakis and his students realized could be applied in this new NeRF domain. Basically, all the reasons that point splats didn't work for regular surface rendering nearly 25 years ago (holes, inefficient, can't edit them like meshes) are less of an issue in a light field capture setup.