Slashdot Mirror
Realistic Human Graphics Look Creepy
WellHungMonkey writes "A really interesting read on Slate about how realistic human faces in games and on robots and so on, are not necessarily the way to go -- the brain isn't fooled, it attaches itself easier to Snoopy-like simplicity... Or Lara Croft attributes, but I'm not sure that's the brain talking."
While simplicity is good, as far as mental-recognition goes, taking simplicity to the max is a bad idea, especially when we have the technology to produce quality-driven graphics. You have to stay around the current level of production quality or you will lose audience. A good example of over-simplification for graphics is demonstrated by the terrible reviews given Radical's unsightly (cell shaded) The Hulk PS2 game. So there's subject matter to consider, as The Hulk was a kind of wacky cartoon/comic, but there was always a darker side to it for me. I was disappointed with the semi-recent Hulk movie, but does that mean the game had to suck too?
For me, a balance of player control with appealing storyline is critical to any video game, and the lack of plausible graphics never helps. Perhaps this could be graphed on a curve or something, but I truly believe there is a balance between all elements of any game or CG film for that matter. Even in film there is still a kind of gameplay, in the physics used and the modes of operations designed to portray the story. Compelling writing fuels the arts, not parlor tricks, so this subject is not exactly cut and dry, by any means... it's very subjective and taste-driven. Another thing to consider is the date that media is designed, because we can all look back at early animation or even live-action special effects and think it looks fake, and the stuff created today will look fake tomorrow. Is there a ceiling to special effects?
The dangers of knowledge trigger emotional distress in human beings.
...as long as the blood spatters are lifelike when I blow their heads off.
Michel Ja...uh...Jefferson.
An Indian-American Hindu committed to non-violent thought/speech/action alarmed by the global explosion of radical Islam
If they were, they wouldn't be creepy. That whole sentence about how the brain knows the difference... doesn't that make them not-so-realistic? I mean, I understand that realism is what they're going for, but the tech isn't there yet. I think we all knew that already.
http://xkcd.com/386/
I'd like to see some examples of these pictures. Sure they are creepy, sometimes people can be fooled though. I had a picture of Aki in a bikini from the Final Fantasy movie on my computer. My girlfriend found it and wanted to know why I had it. She didn't beleive me that it wasn't a real person.
http://github.com/gbook/nidb
If you are refering to games such as UT2k3/UT2k4, Doom III, Deus Ex: Invisible War, etc. I am wondering what you are referring to as realistic human graphics? Since when did human skin look like it was gone over with mop and glo a few times? All new video game engines for some reason or another want to make evey damn thing in the game shiny!
-illumina+us "I put on my robe and wizard hat..."
...right here .
There's a bit in there about how Aesop's fables are more effective because he used animals rather than people for his characters... interesting stuff.
The Army reading list
:-)
:-|
:-(
Also from Slate, about high-definition TV being bad for porn, because it's just too clear. Everything looks better in porn when it's a bit blurry.
I've read in a number of places that game developers have discovered that the more "real" the physics engine, the less "fun" a game feels. Of course, for simulations, you do want accuracy. But for other games, you want "just the right amount" of realism to envelop the user in a believable environment, but not so much so that it mimics the somewhat boring constraints of real-life.
The developers have changed Americas Army recently to include realistic "death drops." It is actually VERY creapy to watch someone shot in the head snap back and collapse and then roll down a hill. It really makes you not want to play anymore.
...And when they came for me, there was no one left to speak out for me." - Martin Niemoeller (1892-1984)
It seems as though that 'the movies' have been in the uncanny valley for a little while. I thought that "The Hulk" was very realistic, but it was missing 'something'. I didn't care too much about that but it seems as though most people instantly pointed and said "FAKE!". It's like the 90/90 rule. "The first 90% of the code accounts for the first 90% of the development time. The remaining 10% of the code accounts for the other 90% of the development time". We are now in the last 10% of making realistic CGI humans, and it isn't easy!
If this was true then I really wonder why this doesn't apply to classical art. I mean, if I visit a gallery of the great masters and look into the faces on the paintings I can really attach to it. And so can millions of people. You can see the love, the fear, the hate in these paintings. I know it is not animated but still, humans seem to be capable of creating artificial pictures of themselves. The point, as I see it, is that game developers are just particulary bad at it.
If you know ANYONE who even VAGUELY resembles R2-D2, I want to see pictures!! (yes, I know they were using them as examples of androids, but jesus... I think using C3PO alone would have sufficed
This appeared on Slashdot a while ago.
The general premise is that has things move towards looking more life like, at a certain point they end up in the "uncanny valley" if they aren't perfect. This is where things look real enough, but the brain sees something wrong with it.
The human brain (and I'd suspect a lot of other species) is very good at picking up the "attractiveness" of something and a lot of it is subconcious. This obviously has developed for mating as a way of choosing the best possible mate. An example would be looking at a girl, being attracted to her and having no idea why i.e what specific features makes her attractive to you?. The counter example would be looking at another girl and finding her repulsive for one little flaw , say a limp or a mishapen nose, even though the rest of her is fine.
The reason cartoons and classic animation don't cause this is because we don't take them seriously.
can't sleep slashdot will eat me
I read the article, and came away unconvinced.
I buy the starting premise of the article: that as computer render figures get more human, viewers become harsher judges of the figures. Mario was cute, while the much more lifelike CGI Neo, in the Matrix Reloaded, was stiff and zombielike.
Since this becomes more true the better the rendering, the Slate writer concludes that computer rendered humans will always look creepy.
I suspect this is another one of those computers-will-never-be-able-to-act-human arguments. Most people want to reassure themselves that there's something inherently irreproducible about life, and humanity. This desire leads us to predict that computers won't be able to render convincing humans, beat a person at chess, or ever create art.
My guess is that a decade from now, people will look at predictions like those in the Slate article, and laugh.
I've seen paintings that look intensely lifelike, so why should such representations be beyond the capabilities of future computers?
Check out my blog: My Galaxy is Milky Way Adjacent
Most humans are inherently creepy.
Why would machine replicas be any different?
Guess there'd be a market for both realistic and unrealistic human characters in games. Clearly, realistic characters would do very well in RPGs and simulation games like The Sims. Try out the "The Sims 2 Body Editor" for some sense of what to expect from EA soon. It's not bad, nice and realistics. On the other hand, there are games where realistics characters aren't as important, such as FPS games. Who cares about a realistic chin lines on the enemy soldier if you're a few mouse clicks away from turning said realistic character into a corpse with a lovely ragdoll physics system?
Same thing with movies, some will obviously develop more on a "cartoonish" look, such as anime gone 3d. No matter how hard they try, they can never make a 16 year old school girl with blue hair that can handle a 300 foot robot come over as realistic. Then again, eventually, there will be serious movies with close to no real actors in it. It will all be rendered because having a large cluster is cheaper then having Keanu Reeves ruin your movie with some atrocious acting...
Hate me!
Real humans don't have such perfect symetry. It's true that better symetry is considered more beautiful but nobody has perfect symetry. And people who look too good, ie. too symetrical, do look sort of creepy.
I remember watching a 'making of' show about the first Shrek movie and they said they purposely made the girl less human-like for the same reason. That she got to a point were it was freaky to have her look that human.
I'm not drunk, I just have a speech impediment. And a stomach virus. And an inner ear infection.
I thought about this very thing watching Jude Law play a robotic gigalo. Unless STDs and the fear of AIDS became rampant, would women really want this? Law's makeup was pancaked to show he was not the generation of Haley-Joel.
This is an interesting problem, if we don't continue to attempt to get to 100%, we will never get there - yet going through the 80th to 99th percentile will be creepy.
I don't have any issues doing it in computer gfx. Some of the new techniques used in Pixar's The Untouchables, such as the way hair moves in water - go towards the overall body of knowledge of how to create actors on screen that you don't know are real. The new Spiderman seems mostly CGI, or motion captured and sped up. This eventually makes for better movies, and games in which the protaganist NEEDS to be human is essential.
But in robotics, I even think the face in the new adaptation of Asimov's "I, Robot" is really sinister. I don't see society even accepting that in robotics. I think the farthest people will go is C-3P0.
Is this why burn victims strike people so oddly? Everyone reacts differently to them, but not usually in a normal way. Once these poor people loose their identities, they become something else, to everyone else. It's not fair to them, they're still them!
-Patrick
"They never stop thinking about new ways to harm our country and our people, and neither do we."
I'm glad the rest of the world realizes it. I've known I hated looking at people for years now.
This story reminds me of an interview I read in, I think, Wired about the making of Shrek. They made the princess as realistic as possible, but it was looking like an animated corpse. They said something along the lines of "until we have the ability to cross the last 1% of realism, we need to step back a bit".
Or something.
I've got more mod points and GMail invi
An example would be looking at a girl, being attracted to her and having no idea why i.e what specific features makes her attractive to you?
Her personalities.
South Park is animated by scanning cardboard cutouts into a computer and manipulating them there. Thing is, the cutouts that get scanned ARE life size. 4 to 6 feet tall. It's probably easier to put small details in that way.
:)
I believe the first episode of South Park was made by filming actual peices of paper in front of a camera. If you look closely, you can see the shadow underneath where the edge of the paper curls up. I'm not sure at what point they switched over to computer animation. Either it happened fairly quickly, or Matt and Trey figured out that they can get rid of the shadows by putting a sheet of glass over the scene.
Technoli
This topic reminds me of an old episode of Doctor Who. A society has android robots but people start developing "robophobia." The Doctor says it's because the robots have lifelike responses in all but one area, facial expressions. The robots all have beautifully sculpted but immobile faces, so this freaks people out on a subliminal level, it's like talking to a dead person.
Of course this was partly a sly commentary on the cheapness of the BBC's special effects on the show, of course they didn't have the budget to do really great robot effects, so they just wrote crappy effects into the storyline. But maybe they were on to something, they were ahead of their time in anticipating the social effects of lifelike robots.
That's simply amazing. That picture looks like someone just used the Healing Brush on a Real photo. If you told me that was a real photo cleaned up I'd definitely believe you at first glance. For me personally only the Forehead and of course the shirt look fake.
On the topic at hand I really would rather the people not look totally real in the types of games I play, FPS. But for Adventure/Mystery games (do they make them anymore btw) etc it could be really cool. Of course real looking people with Brain Dead AI will ruin things. I think the graphic component will arrive well before the AI does. I mean if your playing a game and start acting stupid most AI characters just stare blankly into space not commenting. I want to see games where the Characters are like "Hey jackass, stop running around me in circles" etc.
If you wanna get rich, you know that payback is a bitch
Scott McCloud covered this one too; the more iconic a human figure is, the easier it is for the reader/viewer to identify with it. Conversely, it's possible to anthropomorphize even the most iconic images; the standout example that he gives in his book is an electrical socket that (in the right context) still clearly identifies as a face. If you're interested in the design aspects of this, check out Understanding Comics for more details.
I think this is what killed the Final Fantasy film. The characters were so realistic that my brain accepted them as human--and then spent the rest of the film wondering, "What's wrong with them?" The problem is that we are very sensitive to the subtleties of human behavior. As long as you aren't actually fooled, you are impressed by the quality of the simulation. But if it is good enough for you to take it for human, then what would otherwise be a minor flaw in an excellent simulation suddenly seems like something pathological about another person. So beyond a certain point, if the simulation is not perfect, it starts to seem disturbingly wrong in some undefinable way.
I'd love to see a remake of the Matrix films, in which all of the "in the Matrix" sequences were done with computer animation, like the excellent "Flight of the Osiris" short by the Final Fantasy team. In that context, I think this "problem" would become an asset.
The most effective games for me were the ones that were not trying to be photorealistic. Early games developers really were on to something employing anime for graphical sequences and character charts.
Robotech was one of the most realistic playstation 2 games I've played. Not because the planes and robots looked like actual real-world weapons. It was because they looked and acted like the weapons from the cartoon series I remembered as a kid.
The animation sequences in Dungeon keeper 2 were absolutely believable. The same animation quality applied to Blizzard's Starcraft was not. Why? Dungeon keeper didn't try to look real, and employed a lot of tongue in cheek cartoony elements. Starcraft tried to be entirely too serious.
And don't get me started on Squaresoft...
"Learning is not compulsory... neither is survival."
--Dr.W.Edwards Deming
Now [i]that's[/i] creepy.
Alias has a prett neat quiz bettween real pictures and computer generated ones. If you've never seen it it's availible here. I first recall seing it more than a year ago, so it's not exactly still state of the art, but I don't think I did that well on the quiz.
Degaussing scares the bad magnetism out of the monitor and fills it with good karma.
But I'm probably just nitpicking semantics....
No, that's actually the whole point of the Uncanny Valley.
To take the red pants, red shirt analogy:
The point is: we're trying to get the pants ever closer to matching the shirt, but it's proving really difficult to do. As we get closer and closer to matching, we're finding out that human perception has a finer resolution than we previously though. And the closer the shirt and pants get to matching, the more distracting it becomes, and the more the detail and not the whole becomes the focus of our attention.
The previous poster was trying to suggest that as technology improves, we will indeed be able to make the pants match the shirt perfectly, and there will be no divide.
Many people tend to doubt that. Most feel we're working with two entirely different fabrics, and human perception is so good that exact matching just isn't possible. Therefore, they feel it's better to focus on complementary style.
// "Can't clowns and pirates just -try- to get along?"
In an ep of Doctor Who (Robots of Death) was bandied about the term "Grimwade's Syndrome", a made-up name for a made-up condition where people go crazy in close quarters with robots, because they lack the usual body language to let humans know there are humans in the room.
Not much in Doctor Who turns out to be startlingly prescient, but that certainly did. Grimwade's Syndrome is the best way to describe what the article is talking about - the discomfort of interacting (even one-way, via movie screen) with a "thing" that looks human while every intuitive sense in your brain screams not human.
There's a lot that can be talked about here. I watch our pet bunnies interact with our cat - the cat doesn't try to eat them, which is interesting in itself, but more interesting is how the bunnies respond to the cat. They are confused by her. She is, to them, an only slightly funny-looking bunny, but frustratingly she does not "speak" their language. She doesn't make bunny body language, nor does she respond to it when the bunnies try to communicate with her via body language. I imagine what the bunnies are experiencing is similar to our notional Grimwade's Syndrome, they're interacting with a creepy simulacrum of a bunny that doesn't act quite right.
Or consider this. Because we actually have an "FPU" (Face Processing Unit) in our brains, we pick up on degrees of subtlety in faces - we have perhaps a too-strict sense of beauty, in terms of which faces we find pleasing (ever stop to think how important symmetry is in a face?) - and we see faces anywhere there is even a remotely facelike shape, including the Moon. (I suspect it will come to be the defining characteristic of the human species that we can see a human face there - machine vision systems and alien intelligences will both stare at it and say "I still don't see it".)
Humans therefore tend to react very strongly (understatement) to anything that makes the "FPU" work too hard. If it's sorta like a face but has big things wrong with it, it's "ugly" - maybe even to the point of being a "monster", be it an eyeless skull, a Grey Alien, or a person with a deformity or disfigurement. What IS an ideal, simple thing for the FPU to play with? We may describe an attractive person as "easy on the eyes" but I'd also make a case that the face detector also has an easy time with Hello Kitty, and Hello Kitty looks nothing like Jennifer Connelly. And people tend not to be scared of the "faces" found on the fronts of some cars (unless the driver is a maniac or the car is a Cuda) or of the man in the Moon for that matter, who is greatly distorted and asymmetrical at that. But hey, it's a complex and poorly understood system.
What's interesting is what happens to people who've had damage to that part of the brain. Did anyone else catch the show - mighta been Scientific American Frontiers - where they profiled a guy who had a head injury and now believes his family have been replaced by clones? The kicker was that when they CALLED him and spoke to him over the phone, he believed it was really them, but in person he was certain, despite all better knowledge, that these were not his parents, these were replicants of some kind. Something to do with the part of his brain that considers a person familiar, was malfunctioning, and something at a higher level in his brain was getting uncomfortably confused between people who LOOK like his parents but do not register lower-level feelings of recognition like his parents would. The compulsion to believe this overrode all his better sense: he KNEW these were his real parents, but couldn't make it real in his head.
We're ALL in that boat now with CGI. Our brains are confused: our FPUs are satisfied that the faces look real, but everything else is wrong, the movement is wrong, the behavior is wrong. We process what we're seeing as some kind of weird painting or a reanimated corpse. (And yes, Michael Jackson does trigger this response now that much of his face doesn't move normally when he speaks.) That creepy
~ radiographite: art by john shepard
The animators didn't have a good grasp on (and probably didn't have the technology to model) realistic facial movements. They didn't convey a great deal of emotion.
IAADA (I am a digital animator), and I'd like to point out that you're batting 0.500 in regards to this snippet. While it's true that the facial animation and a sizeable portion of the body language in the final product (no pun intended) both fell short of accurately conveying human-like emotional dynamics, declaring that the animators didn't know part of their craft (referring only to facial animation/body language), greatly oversimplifies the issue; what it mostly boils down to is Time.
Firstly, the entire FF:TSW staff was comprised of some of the best in the business, bar none. Without getting too technical, the modeling techniques available to the modeling staff at the time of production could've indeed sufficed. In regards to complex organic surface modeling, I'm referring to NURBS ("Non-Uniform Rational B-Splines" - a type of curve) Patch Modeling: simply put, a technique involving only curves, where a modeler produces "patches" of any size one-by-one, and then "stitches" them together. Think of it roughly as tailoring a dress made only out of patches and with no visible seams, with as many patches as you'd like. The result: any desired surface, composed of any number of individual regions, each one able to infinitely deform itself while able to affect other nearby patches accordingly. However, patch modeling is painstakingly time-intensive, increasing nigh-exponentially when the amount of patches comprising a surface is increased. This in turn affects the animators' control, giving them near-infinite possibilities for motion, but also increasing the workload tremendously.
IIRC, the characters in FF:TSW were in fact modeled using NURBS patches, but as I mentionned before, the number of patches used could've been two, five, or even tenfold, had the staff been given that much time to work on them. For example, compare the facial animation and body language in FF:TSW to the animation in The Animatrix's "Final Flight Of The Osiris", which was largely produced by the same team. If I may indulge, note the sly glance that the female lead throws to her male counterpart, when they find themselves on the Osiris' bridge just after their sword fight. That quarter-second of motion is nothing short of stunning, as was the rest! Why? Consider the running time of this short film compared to the full-length FF:TSW. Granted, the production schedule was shorter yet not proportional, while the modeling and animation technologies changed very little (granted, this is up for debate).
So there you have it! If anything, the fact that the picture was already 2 years overdue when released leads me to place the blame for the "... [lack of] light in their eyes, or any of the other subtle facial clues we look for when talking to someone" largely on the producers trying to bite off more than they could chew. I'm certain that everyone on the team could pick up a share of the blame, but IMHO the entire endeavour was a laudable effort nonetheless, which greatly upped the bar in terms of achieving life-like digital animation. Cheers!
This explains it.
mt
It doesn't take a FPS to do that. Before we had computer graphics, Generals would act out upcoming battles on massive game tables, using chance to simulate complex components of battle. Both sides in WWII used this technique with pretty good success. Well, when they used it. The Nazis stopped using simulations when Hitler got it in his head that he was some sort of Napolean.
In the Art of War, Sun Tsu writes: Now the general who wins a battle makes many calculations in his temple ere the battle is fought. The general who loses a battle makes but few calculations beforehand. Thus do many calculations lead to victory, and few calculations to defeat, how much more no calculation at all!
Not exactly on topic...
"Learning is not compulsory... neither is survival."
--Dr.W.Edwards Deming
In a game, you invariably respawn or reload from an earlier point. Sure, some people play "iron man" games where there is no saving, but that's rare, I suspect. Heck, most FPSes will currently save your game automatically before you run into a dangerous spot.
I can personally attest to the odd mindset that can leave. I was working with some electronics at one point, shortly after a long gaming session. As I was reaching for some components, I realized I'd better first check to be sure everything was turned off and unplugged. THe thought right afterwards of, "Eh, I can always restore a save point" caused me sober up immediately and put off that work until I'd some sleep under my belt. *shrug* Or maybe I've just got a weak grasp of reality.
This sig has absolutely no significance and serves only to take up screen space and waste the time of the reader.
No kidding! I had a friend who was an airline industry worker, and every time we saw the "Aircraft Control Simulato" game in a videogame store, we always used to joke about how people sit at their desk and drag cigarettes like mad while tearing their hair out and getting no sleep.
It's like, the most stressful job in the world. Why anyone would want to simulate that is faaaar beyond me.
I can't sit through a movie with fighter jets without pissing everyone else off.
"No, you fool...you cannot launch a missile while the aircraft is on the ground!" (well...you can, but it's hard and you can't do it solely from the cockpit) (something with Michael Douglas)
"No, the runway is NEVER next to the main gate" (James Bond)
"If you shot off all your missiles, how come you now have 3 more? (All of them)
"um, no. A 'modified' F-117 is NOT big enough to hold a squad of people inside" (Air Force One)
"No, you can't outrun a missile for that long. Either it would have run out of fuel, proximity detonated, or hit you by now." (Behind Enemy Lines)
Hair and uniforms. (All of them)
"If you're going to call them Air Force planes, at least use Air Force planes. F-18's don't count." (The Rock, ID4)
and don't get me started on Iron Eagle:
"No, you fool...you cannot plug your flight helmet into your Walkman!"
Ever ghost a teammate in America's Army (before version 2.1) and panned around? Sometimes if your subject was standing next to wall your point view would get crammed into the body of the solider itself. At this point you were looking out at the world from his insides. It appeared as a crude wireframe for the most part. For the other part, the developers rendered the backs of the teeth and gums inside the head. Let me tell you, this looked so damn creepy the first time I saw it I couldn't stop staring at it. I later wondered why they bothered rendering the teeth if you can't even see them from the outside anyway? It had to be to creep people out. Had to be. *shiver*
Speak truth to power.
You're probably thinking of the CPPA (or is it the COPPA, I forget). You can find an analysis of the law and the Supreme Court's interpretation of it in Ashcroft v. Free Speech, 535 US 234.
In essence, the Court's ruling hinged on prior cases defining obscenity (and protecting pornography), and outlawing child pornography.
In NY v. Ferber, the Supreme Court said that child porn could be outlawed because of the harm to the child.
In Miller, the Supreme Court defined the test to determine whether speech was obscene such that it was not entitled to first amendment protections.
In Ashcroft, the Supreme Court held that virtual child pornography didn't fall within the Ferber exception because there was no harm to a real child. Therefore any attempt to ban it required application of the Miller test (a very high standard).
--AC