Personally I dislike DLSS so much:
If I enable it in Cyberpunk 2077, there are some visible noise patterns that pop in and out of view especially near shadows on flat textures all the time, it looks horrible and distracting. And if I disable DLSS, if the game updates sometimes it enables it back so I see the horrible things again!
Why do you need AI to upscale anyway if it looks this bad, why not just use Lanczos/bicubic/... upscaling? They are not only simpler and faster, but look better too, because they don't create the above mentioned horrible visible noise patterns.
Then they have the inter-frame thing that causes input lag, another bad degradation.
Then they 'solve' input lag by touting a vendor lock-in-sounding feature (Nvidia Reflex), while the real solution to input lag is not add vendor lock-in feature but ensure you compute 1 frame at the time and do so fast (which is of course not simple, but is a matter of optimization and not of brand-named features, and perhaps not generating and inserting additional frames would help).
Then there is the fact that DLSS is something that needs to be supported by games themselves: why? This again makes it sounds vendor-lock-in-ish because they can have it "exclusively" in some games. But nothing about upscaling, nor frame generation, sounds like something that requires game engine support to me. The graphics card should have enough info to do that by having all the pixels. I mean, it was always possible to play games at different resolutions (DLSS is basically also like playing the game at lower resolution), and there always has been some upscaling required for that (whether it's the monitor or the graphics card doing it). This never required special engine support.