MmmmMm .. . I wants me one o' them robot corpse thingies. . ..
June 12, 2004 6:52 PM Subscribe
"Toss in realtime surface diffusion of lighting and realtime radiosity"
Realize I should clarify here so people don't think I'm repeating myself.
The former refers to the fact that skin is tranclucent and photons diffuse within the medium of skin - this is a very subtle but noticeable effect and if absent you WILL say something along the lines of "WTF IS THAT?". Golem was rendered with this in mind. To be frank I'm too lazy at the moment to look it up but there was just a big article on this that made the rounds, as some prof. recently came up with a pretty good way of doing it for pre-rendered CG sequences.
Realtime radiosity is simply the fact that a brightly lit, colored object placed near another object transmits some it's chroma value to the neighboring object via photons. The algorithms behind this - while a PITA to do realtime - are well-known and understood. Expect the first trappings of this in another 2-3 generations of video cards (of course, people have been saying that for a while - and the problem is that unlike polygons and pixel shaders you can't slowly implement this feature over successive generations of hardware - either you do radiosity or you don't).
posted by Ryvar at 7:03 PM on June 12, 2004
Realize I should clarify here so people don't think I'm repeating myself.
The former refers to the fact that skin is tranclucent and photons diffuse within the medium of skin - this is a very subtle but noticeable effect and if absent you WILL say something along the lines of "WTF IS THAT?". Golem was rendered with this in mind. To be frank I'm too lazy at the moment to look it up but there was just a big article on this that made the rounds, as some prof. recently came up with a pretty good way of doing it for pre-rendered CG sequences.
Realtime radiosity is simply the fact that a brightly lit, colored object placed near another object transmits some it's chroma value to the neighboring object via photons. The algorithms behind this - while a PITA to do realtime - are well-known and understood. Expect the first trappings of this in another 2-3 generations of video cards (of course, people have been saying that for a while - and the problem is that unlike polygons and pixel shaders you can't slowly implement this feature over successive generations of hardware - either you do radiosity or you don't).
posted by Ryvar at 7:03 PM on June 12, 2004
Ryvar, that's interesting stuff but it speaks only to things like surface textures, diffusion and lighting, which are algorithmic and therefore will eventually fall to technological solutions. What the article talks about is our visual comprehension of another person, which has to with the intricacies of behaviors and the context in which those behaviors occur.
How much is nature vs experience is hard to know, but it's probably quite a bit of both, and either way we learn to comprehend others (accurately or otherwise) through a million interconnected subtleties that have a lot to do with how our brains and theirs are wired, and through cultural experience. The uncanny valley is when you get to the point of realism and then suddenly the subtle wrongnesses are magnified inordinately; the the point where they're actually upsetting to look at. Gollum works because his behaviors were mapped from a live actor -- Andy Sirkis; and even then, I suspect it onlly works because he's meant to be creepy and repellant.
Making realistic humans will long be the purview of art as much as technology, and an AGP card can't call upon art to solve problems of realism in real time.
posted by George_Spiggott at 7:18 PM on June 12, 2004
How much is nature vs experience is hard to know, but it's probably quite a bit of both, and either way we learn to comprehend others (accurately or otherwise) through a million interconnected subtleties that have a lot to do with how our brains and theirs are wired, and through cultural experience. The uncanny valley is when you get to the point of realism and then suddenly the subtle wrongnesses are magnified inordinately; the the point where they're actually upsetting to look at. Gollum works because his behaviors were mapped from a live actor -- Andy Sirkis; and even then, I suspect it onlly works because he's meant to be creepy and repellant.
Making realistic humans will long be the purview of art as much as technology, and an AGP card can't call upon art to solve problems of realism in real time.
posted by George_Spiggott at 7:18 PM on June 12, 2004
Actually, I've yet to see a videogame (or for that matter, anything that'll render in realtime on normal PC hardware) that gets anywhere close to the "uncanny valley".
posted by fvw at 8:30 PM on June 12, 2004
posted by fvw at 8:30 PM on June 12, 2004
I agree with fvw -- I think Clive Thompson is confusing shitty art with the genuine uncanny valley. And I'd say we've got a while before developers need to start worrying. The Final Fantasy movie was absolutely state of the art, probably decades ahead of the realtime world, and it didn't even touch the uncanny valley.
posted by Ptrin at 8:38 PM on June 12, 2004
posted by Ptrin at 8:38 PM on June 12, 2004
The "Uncanny Valley" syndrome is a pile of shit. Anyone whose brain works knows what they're looking at and is not automatically repulsed by it.
posted by interrobang at 8:47 PM on June 12, 2004
posted by interrobang at 8:47 PM on June 12, 2004
Are you sure about that, interrobang? I've certainly been startled more often by life-like prosthetics than their more functional (but less "natural") equivalents.
posted by Eamon at 8:53 PM on June 12, 2004
posted by Eamon at 8:53 PM on June 12, 2004
interrobang is a robot, and he's quite sensitive about it.
posted by eddydamascene at 9:00 PM on June 12, 2004
posted by eddydamascene at 9:00 PM on June 12, 2004
Slightly OT, but has anyone seen this Unreal 3 Engine demo? The lighting effects are insane.
posted by gwint at 9:34 PM on June 12, 2004
posted by gwint at 9:34 PM on June 12, 2004
It was interesting the first time somebody name-dropped the uncanny valley theory. After that it has gotten annoying, especially the quotation marks.
posted by Hildago at 9:36 PM on June 12, 2004
posted by Hildago at 9:36 PM on June 12, 2004
Yeah. Next they'll be saying "hagiography". Bastards.
posted by George_Spiggott at 9:43 PM on June 12, 2004
posted by George_Spiggott at 9:43 PM on June 12, 2004
gwint: That does look pretty cool, but looking cool at 200x200 (which is all that video gives alas, though the actual presentation was probably a lot higher) is kind of easy. Also, showing fictional-world stuff is a lot easier than actually trying to do that kind of "realism" with realistic stuff.
posted by fvw at 10:00 PM on June 12, 2004
posted by fvw at 10:00 PM on June 12, 2004
Just so we know what the article is talking about here's an Alias game cinema... which I guess is a tad creepy, some games, for example Onimusha 3 cinemas and I hear Riddick's in game graphics do it a lot better. I think there are some better samples of Unreal 3 over at Gametrailers too.
posted by bobo123 at 10:31 PM on June 12, 2004
posted by bobo123 at 10:31 PM on June 12, 2004
The best designers will be pretty careful about avoiding the valley, I think -- the Sims 2 would be a good place to look for it and it seems like they're keeping that a bit cartoonish. But it'll be a factor in a lot of games.
posted by Tlogmer at 11:23 PM on June 12, 2004
posted by Tlogmer at 11:23 PM on June 12, 2004
In respect to computer graphics, we are at the very least 5 years away from touching the surface of the uncanny valley idealism, assuming it even exists. Focus on how animatronics differ from CGI.
posted by Keyser Soze at 12:17 AM on June 13, 2004
posted by Keyser Soze at 12:17 AM on June 13, 2004
Spiggott you may have missed this:
"The human body will need to be parameterized so that you can throw in a few specific modifiers/random seeds and the appropriate mesh is spat back out."
That would include things like IK constraints, proper skin-over-bone/muscle animation nuances, things like the ways eyelids bulge beneath moving eyeballs etc. specific to the parameters entered with formation of the mesh. It will at some point almost certainly have to be done this way because content development is hugely hugely eating into profit margins across the entire industry - especially console gaming right now. The surest way to get anything done in any industry is to start eating into the profit margins and increased reliance on dev work done by the machines is just one difficult problem that will eventually be solved over the next decade.
We're already approaching it with automated algorithms that scale multi-million-poly character meshes down to 5,000 poly + bump/spec map meshes for realtime rendering - take that same principle and run with it a ways and you're already at the next logical point I'm suggesting here.
The reason we're admittedly getting minor uncanny valley problems right now is because every last idiot and their mother is implementing bump/spec with the surface normal magnitude for skin texture turned up way the hell too high not to mention shadowing systems that, while at least taking into account world geometry now in addition to the usual self-shadowing, only renders hard shadows.
Those two problems won't be factors for too much longer (the former is largely just bad judgement on the parts of artists leads), and then eventually the other things I've talked about will kick in as well.
posted by Ryvar at 12:26 AM on June 13, 2004
"The human body will need to be parameterized so that you can throw in a few specific modifiers/random seeds and the appropriate mesh is spat back out."
That would include things like IK constraints, proper skin-over-bone/muscle animation nuances, things like the ways eyelids bulge beneath moving eyeballs etc. specific to the parameters entered with formation of the mesh. It will at some point almost certainly have to be done this way because content development is hugely hugely eating into profit margins across the entire industry - especially console gaming right now. The surest way to get anything done in any industry is to start eating into the profit margins and increased reliance on dev work done by the machines is just one difficult problem that will eventually be solved over the next decade.
We're already approaching it with automated algorithms that scale multi-million-poly character meshes down to 5,000 poly + bump/spec map meshes for realtime rendering - take that same principle and run with it a ways and you're already at the next logical point I'm suggesting here.
The reason we're admittedly getting minor uncanny valley problems right now is because every last idiot and their mother is implementing bump/spec with the surface normal magnitude for skin texture turned up way the hell too high not to mention shadowing systems that, while at least taking into account world geometry now in addition to the usual self-shadowing, only renders hard shadows.
Those two problems won't be factors for too much longer (the former is largely just bad judgement on the parts of artists leads), and then eventually the other things I've talked about will kick in as well.
posted by Ryvar at 12:26 AM on June 13, 2004
That's weird. I was thinking similar thoughts just the other day about movies like Final Fantasy, Shrek, and the upcoming Polar Express. For my part, I don't think it is necessarily how lifelike these things look, but how unlifelike they look. The graphics distract from the story and the experience for me, because I am constantly critiquing them in my mind. It's not something I even intend to do, it's just something that happens whenever I see movies like that.
That the author uses a lot of the same words that came to my mind when I was explaining this to a friend yesterday ("zombie", "ghastly", "ghoulish") is a little eerie in itself. Not so sure about he Uncanny Valley effect. It might make a good research project or three though.
posted by moonbiter at 1:44 AM on June 13, 2004
That the author uses a lot of the same words that came to my mind when I was explaining this to a friend yesterday ("zombie", "ghastly", "ghoulish") is a little eerie in itself. Not so sure about he Uncanny Valley effect. It might make a good research project or three though.
posted by moonbiter at 1:44 AM on June 13, 2004
The "Uncanny Valley" syndrome is a pile of shit.
I have to agree for the most part. I think that many people have a reaction to seeing human-like but clearly artificial images, but I think that's a familiarity and philosophical issue rather than some kind of spooky innate mechanism. When movies first came out, many people had a negative reaction to them at first as well, but they (mostly) got over it.
After all, why would we have evolved to reject "almost human" imagery? History documents no episodes of having to distinguish and fight off alien simulcra.
posted by rushmc at 9:32 AM on June 13, 2004
I have to agree for the most part. I think that many people have a reaction to seeing human-like but clearly artificial images, but I think that's a familiarity and philosophical issue rather than some kind of spooky innate mechanism. When movies first came out, many people had a negative reaction to them at first as well, but they (mostly) got over it.
After all, why would we have evolved to reject "almost human" imagery? History documents no episodes of having to distinguish and fight off alien simulcra.
posted by rushmc at 9:32 AM on June 13, 2004
Well, right. It's an evolutionary glitch, a matter of being incredibly fucking sensitive to human imagery (2 dots and a line make a piece of anatomy?? :) and not particularly sensitive to nonhuman imagery; something human enough to get analyzed as such, with all the modules turned on, can be deeply disturbing if it's off.
posted by Tlogmer at 11:22 AM on June 13, 2004
posted by Tlogmer at 11:22 AM on June 13, 2004
Ptrin: The Final Fantasy movie was absolutely state of the art, probably decades ahead of the realtime world, and it didn't even touch the uncanny valley.
Which is funny because I had the exact opposite reaction. While the look of Final Fantasy was good, it was almost as if I was watching high-school drama students mouthing their lines. The little nuances of facial expression that makes acting believable just were not there.
rushmc: After all, why would we have evolved to reject "almost human" imagery? History documents no episodes of having to distinguish and fight off alien simulcra.
But there certainly are cases in which we might want to not be associated with someone whose face was subtly wrong. Smallpox, Syphillis, and Herpes comes to mind as communicable diseases that leave physical traces. Jaundice and pallor can be signs that something is basically wrong. And to be blunt and say the things that most people don't want to hear, we have an instict for xenophobia, for treating people who don't look like clan members with suspicion.
The other issue has to do with recognition of deception. Part of what separates a good actor from a bad actor is the ability to decieve the audience into thinking that the emotion expressed is authentic. The Final Fantasy digital actors, Princess Fiona in Schrek and even the motion-captured Tom Hanks in The Polar Express preview all feel much less expressive than, Chuck Jones' Grinch.
posted by KirkJobSluder at 12:11 PM on June 13, 2004
Which is funny because I had the exact opposite reaction. While the look of Final Fantasy was good, it was almost as if I was watching high-school drama students mouthing their lines. The little nuances of facial expression that makes acting believable just were not there.
rushmc: After all, why would we have evolved to reject "almost human" imagery? History documents no episodes of having to distinguish and fight off alien simulcra.
But there certainly are cases in which we might want to not be associated with someone whose face was subtly wrong. Smallpox, Syphillis, and Herpes comes to mind as communicable diseases that leave physical traces. Jaundice and pallor can be signs that something is basically wrong. And to be blunt and say the things that most people don't want to hear, we have an instict for xenophobia, for treating people who don't look like clan members with suspicion.
The other issue has to do with recognition of deception. Part of what separates a good actor from a bad actor is the ability to decieve the audience into thinking that the emotion expressed is authentic. The Final Fantasy digital actors, Princess Fiona in Schrek and even the motion-captured Tom Hanks in The Polar Express preview all feel much less expressive than, Chuck Jones' Grinch.
posted by KirkJobSluder at 12:11 PM on June 13, 2004
hi, my boss requires that I write articles. Heres one of the fillers.
posted by Satapher at 1:18 PM on June 13, 2004
posted by Satapher at 1:18 PM on June 13, 2004
It's a mistake to treat "the uncanny valley" as fact. It's a theory, and not a particularly well-supported one.
posted by NortonDC at 1:52 PM on June 13, 2004
posted by NortonDC at 1:52 PM on June 13, 2004
The uncanny valley syndrome is a pile of shit.
No, no! Valley concave. Pile convex. Valley not pile. Valley not pile of anything. Valley absence of pile. Valley what left after make pile.
posted by George_Spiggott at 6:10 PM on June 13, 2004
No, no! Valley concave. Pile convex. Valley not pile. Valley not pile of anything. Valley absence of pile. Valley what left after make pile.
posted by George_Spiggott at 6:10 PM on June 13, 2004
« Older from realism to near abstraction | Bet your bottom dollar Newer »
This thread has been archived and is closed to new comments
Probably another 10 years until we see all of that implemented well enough for general-purpose in gaming and such.
posted by Ryvar at 6:57 PM on June 12, 2004