It is hard to compare the gaming techniques with what is used at Enscape (at least without going into technical detail) because the goals and requirements are much different. We will most likely not use DLSS but improve our existing state of the art filtering and upscaling solution.
Has the annoucement of DLSS 2.0 changed anything? It sounds like it's gotten a lot more reliable than before. Theoretically you could use it on top of your in house filtering and upscaling, and improve performance even further. It sounds like they've trained the network to be pretty universally applicable, so Enscape should theoretically work with it out of the box.
If not though, perhaps there's still a way to take advantage of the tensor cores in RTX, which presumably are sitting dormat within Enscape currently, correct? Though that would require training your own networks. Neural networks are one of the hardware accelerated features alongside raytracing that are only going to become more common in future graphics cards (including from AMD).