Nvidia demo by far. The both still look fake tho so don't know where the creepiness comes from. Also, after seeing the model for the PS4 character it made me not impressed of the incoming PS4 graphics.

What exactly does the face animation have anything to do with the devs not "connecting" the faces to the bodies?  NOTHING.
You even say "despite the faces" so you actually agree with me and just disagree for the sake of disagreeing?

BTW 
Naughty dog? lol  When they start to come up with at least 1 good character design (Nathan Drake is like the most boring "Hero" of any franchise that exists even the Tetrimos (Tetris blocks) are more interesting) then they can start to criticize others :)

I mean just describe Nathan Drake without mentioning what he does or Sony/Playstation or Adventure etc... 
He is a human with.... well....  - The only way to describe him is saying  He looks totally average and there is nothing special about this guy (this has nothing to do with making the players feel they are the character etc. its just bad design)

GOW's Kratos is an awesome character design. Sonic is. Mario is.

I mean try to guess who I am talking about.
1) Gorilla, red tie, barrels
2) Hotpants, turquoise top, boobs.

You can simply fix the "connection" problem then the illusion is great but being bad designers is another story :)

(my "BTW part" of the comment also has nothing to do with the faces but I am also allowed to help my argument by mentioning irrelevant stuff right? xD)

The NVIDIA one looks amazing.

Oh certainly. I was doubtful that the PS4 would be able to deliver impressive graphics and I can happily say I was proven wrong. From what you've seen, what kind of GPU performance would you put the PS4 at so far? I know the rumoured chip is the 7850 but with optimisation,  volume GDDR RAM etc, what GPU would you say it is the equivalent of? 

If by creepy feel you mean looks realistically creepy I would choose the NVIDIA one. 

Which one will give us the better games? My guess is PS4.

Of course, the head is only a part of the story. The hardest thing in my opinion is in animating the whole figure realistically. You'll need literally thousands of frames of animation to emulate a human standing up and walking to the bar for a pint. Imagine how many frames you'll need for when your character interacts with his environment. ND try to add these effects into their work, such as Joel clenching and unclenching his fingers, but it needs to go a lot further than this. It needs to appear 'infinite' even if it isn't.

Maybe your character puts his cigarettes away in a different pocket and you see him searching his pockets the next time you want a cigarette. Maybe when he lights a match it blows out because of a draft and he has to cup the second in his other hand. Maybe he burns his finger on the match because he's held it too long. Each and everyone one of these scenarios would require nuances of animations depending on where they are when they perform them and who they are standing with. Reality isn't cheap.

LA Noir has impressive facial animation technology. That's it. They don't look real to me. In fact they look very cartoony. The animation technology gives them very expressive faces, but the face as a whole is no more realistic than Enslaved or Crysis 2 or any of the other high end gaming engines. The old man above is a lot more realistic than the LA Noir faces, but since he's less cartoony he falls further into the uncanny valley while LA Noir teaters on the edge.

The intention of nvidia-demo was to show animating faces in realtime, so that it seems authentic. Especially subsurface scattering and tesselation makes it a lot better than previous face-demos. Technically I think nvidia's demo is way better but that doesn't mean it looks better.

Going by what we've seen so far from Cage, I'd have to agree with this ... but I don't really think the uncanny valley problem can be answered by tech, I think 'technique' is going to be the answer. You also have to remember that Cage's demo was specifically designed to show the eyes ( to give the impression of what the actor is thinking or feeling). I'm positive that once Cage's tech is used with full animation, people will begin to see the difference more starkly.

This comparison is ironic. Cage was trying to point out that telegraphed emotions is now a thing of the past whilst the NV demo did exactly that. It offered us almost cartoon like and exaggerated expressions that were designed specifically to telegraph an emotion, whilst Cage's demo gave us a moment of 'pensiveness'. It didn't just smile (happiness) or scowl (sadness) it had a much richer and deeper effect.

Watch this demo from Beyond on the PS3 and imagine this translated with Cage's advance tech and you'll get the idea. What is the policeman thinking? (Fatherly concern?)  What is the girl going through while she's being questioned? (torment and a sense of entrapment?) She longs to tell him, but she just can't. And all this is accomplished by very subtle nuances of animation and posture. Take note of her face at 1.4. It's actually pretty incredible. Then at 1.40 when her mouth wants to speak but the inner termoil won't allow her to:

http://www.youtube.com/watch?v=M48SZn83sPY