I find this version of DLSS too taxing for the hardware. They should have focused on fixing the artifacts and improving the material rendering only. Now they decided to change the lighting to the point that it makes the graphics look like being AI generated. They shouldn't launch it with those changes, as the image looks incorrect and fake.

It's a beauty filter...

The eyes are bigger, the use eyeliner/shadows, the earring looks changed, the eyebrows look different, the mouth looks different... ect.

Does it look closer to human? more realistic? yes.
Did it CHANGE the artistic integrity of developers that did the art for the character models? yes.
It does do the whole A.I thingy, its not quite just lightning. 
Its apparent it does more than that.  

In the videos with motion, you can see shadow pop-in on the face too, as they smile or move their mouth/face. 
There is a A.I filter feel to it.

as I've seen more of the dlss5 videos, I'm leaning more towards a "no I don't want this, in games".
I think, I dislike it changing the day/night times in pictures, and beauty filter like effect is too much.

However they do it for next gen titles (in the end, it's built for that), I'm hoping they make it close to path traced ground truth of original artistic vision, so in essence "trick" to make PT less computation heavy - cause a lot of what they've showcased in this presentation looks very different and off compared to originals.

Oh, cool. Yet another thing that will deeply divide the gaming community. We needed another one of those.

It seems like the reponses are overwhelmingly focusing on the Grace facial features in street scene.
When DLSS 5 is equally being used in the other ¨portrait¨ images released, as well as background elements of street scene.
I´m not sure where these are normally linked from, but you can just edit the html to see screenshots 1-3:
https://www.nvidia.com/en-us/geforce/news/dlss5-breakthrough-in-visual-fidelity-for-games/nvidia-dlss-5-resident-evil-requiem-geforce-rtx-comparison-screenshot-003/

The portrait images show that DLSS 5 itself is not tied to the facial changes seen in the street scene, which don´t show up in the portraits.
You just see much more detailed lighting, which brings out details already present in the underlying model and source content.
I think the background elements of street scene also are worth addressing, e.g. of girders against light grey sky, or even qualities of leather jacket.
The comparison is harder because the DLSS 5 version has brighter over all illumination, obviously a choice of Capcom not fundamental to DLSS5.
But my eye gaze can dance around the scene and every area seems appropriate, whereas without DLSS 5 many areas feel greyed out.
(although the real effect when the world is in motion is not readily apparent in still screen shots, like many other graphics technicalities)

I would be interested how the technique could be applied to render scenes not going for such realism in the first place...
Because Capcom´s source content clearly already had super high level of detail, i.e. they were aiming for realism style.
(and of course their choice in human models, who in fact have the prominent facial bones people focus on when lighting improves)
(but with a different human model and different facial bone structure obviously the visual outcome would be very different)
What DLSS 5 is achieving doesn´t seem out of bounds for ´CGI cinematics´ even if in-game treatments had worse lighting etc...
But even ¨cartoonish¨ styles could plausibly benefit from richer suppleness of lighting variation. 

Overall I think the negative reactions do point to ´photorealism´ not being seen as universal positive anymore.
And this isn´t specific to DLSS 5, but the modern 3d gaming industry has just blindly pursued that goal for 25-30 years.
But as we are more closely approaching that goal, the outcome is people don´t really care, it can be a negative just as much as a positive.
The universal desirability of that goal was really just assumed by so many companies, and frankly many consumers bought into it as well.
I think even putting aside the specific anti-AI stances, people aren´t as prone to accepting this goal as universal good, but ask ¨what else?¨.

That's the reason we all are charged triple on RAM and SSDs, btw.

Grace's idle animation does open her mouth from time to time for a few seconds.

I just took this screen myself:

I like how DLSS 5 brings out the outdoors quality of light and space here, including for the shadowed areas of the patio cieling...
(although again, scene differences impede 1:1 comparison with the DLSS 5 version clearly having cooler light than non-DLSS 5 version)

The Hogwarts street scene is also great... just the stacked cauldrons is great, and over all there is more contrast, it feels more HDR.
Although it seems that there is less overall dependence on ¨strong shadows¨ to imply space and depth, with subtleties coming out more.