So...personal preference is always going to be on a non-interpolated image. That being said, what I hear from Nvidia and from AMD is the same thing.
Taken in broad stroked, and with some silly numbers, I'd like to try and express this from a gameplay perspective. Because you are taking 1 frame, and generating 3, on paper you get 4x the FPS. In reality, you get much less. The interpolated frames could potentially be twice as long to generate, meaning 4/2 = 2x the actual frame rate. Because you have variation in complexity of what is rendered, one frame quartet could take 1 second, another could take 2 seconds, and you thus get frame timing issues despite having many, many more frames. FSR and DLSS both take this, attempt to apply more algorithms to smooth everything out...and in my opinion really don't offer anything more than trying to deliver more frames that simply retard the motion which would be intended. IE, this would be absolutely the way to generate extra frames to be smooth...if everything moved fluidly.
Unfortunately, the pattern is that there's another derivative. The derivative of position is velocity. The derivative of velocity is acceleration. The derivative of acceleration is jerk. Frame interpolation, which is what DLSS and FSR attempt, basically flattens out any jerk and people who know natural systems have a very hard time whenever they detect spikes in jerk that their eyeballs cannot call out, but that they cannot un-see once pointed out. Call me old fashioned, but objects with instantaneous spikes in jerk are silly difficult to deal with, and this is why I'm against the idea of frame generation, interpolation, or whatever you want to call it. I'm also someone who has a splitting headache at a 70 degree FOV...so power to the people that like this stuff. I just can't deal with this crap, and would choose 60 FPS over 180 every day of the week if the later was framegen...because bigger numbers are not always a better experience.