Forgot your password?
typodupeerror
Graphics

Why the Uncanny Valley Doesn't Really Matter 214

Posted by CmdrTaco
from the says-you-dude dept.
malachiorion writes "Are humanoid bots and CGI characters still crawling their way out of the Uncanny Valley? Maybe, but maybe it doesn't matter. Here's a cold, hard look at a popular robotics theory that might have no legs to stand on, android or otherwise. It's everything that seems wrong and irrelevant about the Uncanny Valley that I wasn't able to fit into this month's Popular Mechanics cover story on social bots."
This discussion has been archived. No new comments can be posted.

Why the Uncanny Valley Doesn't Really Matter

Comments Filter:
  • by avandesande (143899) on Thursday January 21, 2010 @03:20PM (#30849558) Journal

    Let's worry about it when robots that fall into this scenario actually exist.

  • by Un pobre guey (593801) on Thursday January 21, 2010 @03:22PM (#30849594) Homepage
    with all respect to evolution

    Humans are a small niche in biological evolution. Most creatures are very well adapted for specific environments and life strategies, which I presume is the underlying point you are making about how robots should be designed. If you have a recent model new car that is midrange or higher in price, you have a robot. Roombas, appliances with computers in them, washing machines, dishwashers, robots all. We just haven't been calling them that.
  • by jameskojiro (705701) on Thursday January 21, 2010 @03:28PM (#30849674) Journal

    Humanoid robots are great as they can use the same tools as humans can and can more easily relate to humans. Why build several $5,000 domestic chore robots that need special tools when you can buy one $20,000 humanoid robot that does all of the shores, need no special tools to clean the toilets, do the dishes, and vacuum the floor except the cheap tools humans already use. Plus make it so you can shag the robot so that makes it win-win... Kinda hard to have a relationship with a Roomba....

  • by circletimessquare (444983) <circletimessquare@NoSpaM.gmail.com> on Thursday January 21, 2010 @03:30PM (#30849710) Homepage Journal

    went on their famous expedition, there was a black guy in their group, york

    http://en.wikipedia.org/wiki/York_(explorer) [wikipedia.org]

    the native americans would stand in slack jaw amazement at york, as if he were possessed of magic. they never saw a black man before. york would further dumbfound them by taking out and reinserting his false teeth

    meanwhile, consider the cantina scene in star wars: aliens of extreme forms, and humans mingling in with them as if no big deal

    both the cantina scene and york's experience are the truth: our amazement at first is profound and very real at seeing new ethnicities/ life forms. but it also wears off very quick

    we can get used to interacting with anything. the uncanny valley is real, but its also very temporary

  • by Scrameustache (459504) on Thursday January 21, 2010 @03:33PM (#30849754) Homepage Journal

    I think it's weird that some people have a fascination with humanoid robots in the first place.

    Everything we have is designed to work with our humanoid bodies, so if we want to make a device that interfaces with those things, it will work better if it shares the humanoid design.

  • by PylonHead (61401) on Thursday January 21, 2010 @03:35PM (#30849780) Homepage Journal

    Wasn't impressed with the article.

    He calls the Uncanny Valley "a groundless thought experiment", when it's really a simple description of a phenomenon that I (and presumably many other people) have experienced.

    He goes on to say that people aren't frightened by humanoid robots. My experience with the uncanny has never frightened me. It's more of a vague repulsion and an emotional disinterest.

    He then goes on to talk about a series of robots that aren't nearly human-like enough to trigger the uncanny valley phenomenon. Honestly the phenomenon seems much more relevant to the computer graphics world than it does to robotics at this moment in time.

  • by Itninja (937614) on Thursday January 21, 2010 @03:38PM (#30849828) Homepage
    Reminds me of a story my English teacher told the class in high school. He is 6'6" tall and spent some time in China after college. He told us that, even though there were other Anglos here and there, everywhere he went there would eventually be a crowd surrounding him and gawking at his height. I imagine it was the same phenomenon he are speaking of: unfamiliarity adds either novelty or revulsion....but once something is familiar it is (eventually) accepted.
  • by Rhacman (1528815) on Thursday January 21, 2010 @03:44PM (#30849896)
    I thought dogs just generally chased, attacked, and ate just about anything i.e. squirrels, rabbits, cars, postal employees, spherical objects, non-spherical objects, dirt, rocks... As for sharks, I'm under the impression that their perception of the world can be classified by "to be eaten" and "to be ignored" wheras dolphins are simply reacting to their presence on the former list. That said, I could be swayed if some sources were provided.
  • by Anonymous Coward on Thursday January 21, 2010 @03:49PM (#30849988)

    It was the first virtual world which I could see as real, which I didn't have to pretend otherwise because all previous efforts has give-aways that it was fake. It looked goood (and if you sat through the credits, the masses of names hint towards the work needed to make this so) and that's why it's so successful and a breakthrough, imo.

    No. The reason you didnt feel the uncanny valley was because it wasnt real. It was so far from real that your brain didnt find the twisted smurf creatures disturbing.

    I'm also pissed off (as a phd in graphics research) that everyone thinks its breakthrough. Gollum in LOTR was a breakthrough, theres no new tech in this movie. James Cameron needs to stop saying how he invented mocap, its stupid. You'll find that most of the amazing "breakthroughs" of the last decade you didnt actually notice because the CG was perfect and more importantly subtle.

    If you create an entirely made up world you can put anything in it and have it "fit", because you have accepted the fantasy.

  • by Deadstick (535032) on Thursday January 21, 2010 @03:51PM (#30850010)
    Perhaps there would be an uncanny valley if you knew what a ten-foot blue alien was supposed to look like.

    rj

  • by DutchUncle (826473) on Thursday January 21, 2010 @03:57PM (#30850084)
    Somehow nobody has trouble dealing with the Muppets, or the Henson-created aliens on Farscape; even little children deal with them, and my non-techie mother-in-law thinks my wife's Rygel doll is "cute". (Well, maybe it looks better than Rygel did; point is she doesn't say "it's a squishy frog".) Somehow the folks at Pixar manage to make an architect's lamp behave enough to make people think of it as a creature. Humans can accept a *lot*.
  • by khallow (566160) on Thursday January 21, 2010 @04:04PM (#30850182)
    I think a key problem here is simply that humanity has evolved elaborate behaviors and telltales for communication even without intent. For example, in a store there are distinct differences in the behavior of someone looking for something versus someone walking purposefully to a destination. Irritation is easily transmitted. What this means is that for the typical human, there are limits to how well they can deceive another human. I think that's one of the causes of the uncanny valley. If you're in the valley, then the behavioral cues either cannot be interpreted or even worse are merely a skin that can be readily changed.

    Violations of these behaviors and evolved protocols can really upset us. For example, a scene in Terminator II shows the evil robot of the movie (which can take anyone's form that it kills) has earlier (unknown to the audience) taken over the form of a woman talking on the phone to her son. The audience senses something is wrong, the dog is barking hectically outside and the woman is trying suspiciously to milk the son for information on his location. She gets the name of the dog wrong. It's only then that we know she's the killer robot. Right after the call ends, we find the robot also killed the husband of the woman while casually talking on the phone. Think about that. Someone who can chat on the phone without even a trace of emotion or extertion while killing a person at the same time. Bladerunner explores this to great extent (the opening scene is a great example). Silence of the Lambs is in part about a hideous serial killer who shows no remorse and reveals of himself only what he wishes.

    We are scared of people who can lie and kill without the deed showing in their behavior and that fear is readily milked in many movies about murderous robots and calm psychopaths. I believe this is part of the uncanny valley. We've evolved over time to share a common nonverbal system of communication. Anything which can exploit this system, be it beings that don't look quite right or can deceive us completely and effortlessly, triggers a warning in us.
  • About dogs... (Score:3, Insightful)

    by grocer (718489) on Thursday January 21, 2010 @04:07PM (#30850222)
    Dogs are domesticated wolves...who live and hunt in packs. If the cat is an accepted member of the pack, it will be tolerated by the dog. This is a vast oversimplification, of course, but what it comes down to. Man has selectively bred dogs for specific tasks since domesticating them...so we have dogs for hunting, herding, security, and companionship. Depending on which tasks the dog was bred for will determine whether it's sociable with other pets. Even then, there will be variation between individuals of the same breed and while some breeds are more cat/pet-friendly than others, each dog is still an individual and results will vary. If they dog accepts the cat or cats as part of its pack, there's no problem. If the dog doesn't have a strong prey drive, it may just ignore the cat. Either way, it has nothing to do with the Uncanny Valley.
  • by ceoyoyo (59147) on Thursday January 21, 2010 @04:09PM (#30850262)

    Got a reference for that?

    Sharks and dolphins compete for food and a lone dolphin is potential prey for a hungry member of one of the larger shark species.

    Ditto with dogs and cats - in the wild they compete for food and are potential prey for each other. Sure the domesticated variety often live happily together... as long as there's plenty of food and living space.

  • Re:So? (Score:2, Insightful)

    by poopdeville (841677) on Thursday January 21, 2010 @04:21PM (#30850476)

    There's only a handful of tasks such a machine would be optimal for

    Including replacing most of the service industry.

    , and just having a human do it will still be the better choice for quite a while.

    I think I'd rather live un-optimally, with a Maidbot 4000.

  • by hitmark (640295) on Thursday January 21, 2010 @04:35PM (#30850676) Journal

    i would not recommend building any kind of cannon into a robot designed to serve us, just in case...

  • by interkin3tic (1469267) on Thursday January 21, 2010 @04:35PM (#30850680)

    I don't think Avatar applies in this case because only the aliens were CG. All the significant shots of humans were flesh-and-blood people.

    And I think importantly the aliens were far enough from humans that the uncanny valley wasn't triggered. Their eyes were semi human, but were different enough that they looked cute, not wierdly almost human. Their skin was blue, so there wasn't any "Something isn't quite right with their skin...OMG THEY'RE DISEASED" response from your brain.

    The movements and expressions looked natural because if I've heard correctly, those were basically real movements and expressions, not artificially made. If there were humans who were CGed in Avatar, I'm guessing they required far, far more work and money than the alien sequences.

    Maybe if there were a real pandorean and they were to see it, they'd be creeped out as hell by the CG aliens.

    Basically, I think Avatar cheated out of the uncanny valley, or at least got off on a technicality.

    I remember in the matrix there were a few CG shots of people. There was a reason they were wearing sunglasses at night while kung fu fighting, it wasn't just to try to look cool or make me think of the song, it was also so they wouldn't have to spend a lot of money making the eyes look right.

  • by Blakey Rat (99501) on Thursday January 21, 2010 @04:39PM (#30850762)

    You kind of misunderstand the phenomena. The reason all the things you listed are easy-to-take is that they don't even look remotely close to humans.

    The uncanny valley refers to the emotional detachment towards CGI creations that look and behave *almost* (but not exactly) like real people. A good example would be the recent CGI Beowulf movie, or another poster mentioned Final Fantasy: The Spirits Within from a few years ago.

    The theory is that the less human a creature looks/acts, the more we can accept it. Pilot or Rigel from Farscape don't look human at all, for example. As the creature approaches realism, the viewer will have more and more trouble accepting them. Once your simulation is sufficiently complete, then you accept them as you would an actual human actor. That's why porcelain dolls are so creepy.

    So you take Rigel, easy to accept. The humans in Beowulf, much less so. Some would argue that the aliens in Avatar are close enough to human that people accept them without any problems, which partially explains why the film has been so successful.

  • by Kjella (173770) on Thursday January 21, 2010 @04:40PM (#30850794) Homepage

    Human contact has been replaced by machines lots of places like bank clerks who has been almost fully replaced by online banking and cash machines, or how about ticket machines or vending machines? I have a dishwasher and washing machine, none of those are built the way I'd wash dishes or do laundry. The point is not that machines suck at being useful, it's that they suck at being humans. I'd rather in fact not have a clippy interface to my machine if I can help it. Why does everyone seem to think a humanoid robot would be such a great solution? Would you like to piggyback on a humanoid robot to work every day? Do you honestly think it's good design to command a robot to use a remote control to tune your TV when you could command the TV to tune itself?

    Don't get me wrong, eventually we will need some sort of general robot but my home could be a lot more intelligent than it is. There's no universal "bus" that things expose themselves to, and I don't mean building a special house full of special tools that are all built to work together. I mean something that'll be pretty much as basic as electricity and everything announces itself and lets me turn on and off lamps, turn up and down the heating, tune the TV, monitor the oven (maybe not set that one), check the status of my washing machine all in one dashboard right here, without getting my ass off the chair. That would at least be a start...

  • by BetterSense (1398915) on Thursday January 21, 2010 @04:42PM (#30850838)
    I've been told that the reason people (often) fear snakes and clowns is that they cannot read any emotion from their expressions. I think a robot would have to be amazingly nuanced and advanced for people to accept its body language as human.
  • Why? (Score:3, Insightful)

    by mosb1000 (710161) <mosb1000@mac.com> on Thursday January 21, 2010 @04:47PM (#30850914)
    Humanoid robots terrify me to no end. At first, I was simply bothered by the fact that people were trying to develop this technology. I couldn't understand what new functionality they were trying to develop, and I didn't see why they would simply try to duplicate the existing functionality of human beings (since there are already billions of us around).

    Now that I've thought about it, I think the new "functionality" they want to add is compliance. They want to say to something, resembling a person, "do this" and have them do it without talking back. Basically, they want someone to serve them without ever having to consider that person's needs or feelings. They want someone to go and take car of their mother or their children for them, so that they don't have to. They want someone who will have sex with them for no reason other than their desire for sex.

    So some day, the hope is, we will be surrounded by human-looking robots who will cater to our every whim and never give us any trouble. I don't think that's good for us, and I question the mental stability of someone who would want to live that way.
  • by jollyreaper (513215) on Thursday January 21, 2010 @04:59PM (#30851168)

    Everything we have is designed to work with our humanoid bodies, so if we want to make a device that interfaces with those things, it will work better if it shares the humanoid design.

    To an extent. A humanoid form for domestic robots would seem useful but we see that a roomba does a pretty good job and it's nothing more than a flat disc. If you look at conversion kits to turn standard human-operated trucks into remote vehicles, they're admirably utilitarian with a set of stereoscoptic cameras mounted where a human head would go but with simple servo-operated levers for controlling the gas and brake and a neat little set of rubber gears for gripping and turning the steering wheel.

    If we were to ever invent a general-purpose robot, one capable of doing many tasks, it might settle on a human form. Right now our robots tend to be more designed for the purpose. A roomba whirls around the room but does not lift furniture, does not have an attachment for getting between the cushions, etc. An automatic car wash is basically robotic and looks nothing like a human while doing the same work. They're usually worse at it than a human but all you'd need to fix that deficiency is mount some cameras so they can really see the job they're doing and have an articulated pressure washer and scrubber arm to get at the dirt that's not coming off. Computer vision systems are getting to the point where they really could identify clean and dirty with cars off the street. Previous example of computer vision system were like the ones the potato chip companies use to sort bad spuds and they check the incoming potatoes against a known list of acceptable potato colors.

    There's a whole field of biomimicry that seeks to borrow nature's solutions for various engineering problems. While nature can develop some very interesting techniques, it's important to remember that the process is not guided and also has to work with the materials at hand. The common example given is wing-flapping flight. It's not very efficient but modifying limbs to flapping surfaces is about the best you can do. Same goes for terrestrial locomotion. Wheels are awesome but there's only one axle we've ever found in nature and it's on a microscopic organism. Anything bigger than that is pushing itself around with limbs.

  • by b4k3d b34nz (900066) on Thursday January 21, 2010 @05:10PM (#30851386)

    A primary reason that humanity is so successful is because we are general-purpose, but also because we have the ability to use tools. Robots can be designed to cut out the middle man and become the tool itself, or they can be more general use and still require tools or swappable arms/interfaces.

    So think about it, do humans REALLY interface with the items we build? If so, why are there eventually 2 versions of these created items: "standard" and "ergonomic"? It's because the standard ones never are actually designed for humans to a T, and ergonomic versions attempt to rectify it. Also, humans are so different from each other in terms of size and agility that most of our interfaces do not exactly match even the average human being. Robots don't have to suffer from this problem, although they surely will have to handle deviation in the environment.

    Humanoid robots are potentially successful in certain areas like teaching--although Disney's "Crush" the Turtle exhibit seems to work quite well for a lot of people--but until we get to the point where biomimicry is successful to the point where robots are commonly sexualized (plausable acceptability), we may as well continue creating robots that are fitted to a smaller number of uses in specific niches.

  • by SmallFurryCreature (593017) on Thursday January 21, 2010 @05:10PM (#30851392) Journal

    He only briefly mentions the Polar Express movie that had a really big uncanny valley problem.

    If you have seen the movie, you know what it is, it looks at once so real and at the same time is a Bugs Bunny cartoon. The two just don't match.

    There is a reason movies like Madagascar, Wall-E, Up etc despite being drop dead gorgous use clear cartoony looks ESPECIALLY for their characters. And while it is acceptable for the plane in Madagascar 2 to bounce like cartoon plane because it is clear that no matter how accurately it is rendered it is NOT a real aircraft. But the train in Polar Express DOES look real and real trains do not behave like that.

    It is the superman problem. Spiderman can swing from wall to wall because it slightly makes sense, sense enough perhaps. Superman being able to lift an Island doesn't. Stopping a train. That has nothing to do with strength anymore, but with conversion of energy. He may be strong enough, but we know the ground isn't.

    The uncanny valley is NOT something not looking right, but us knowing that the real thing looks or behaves different. It is why walking is SO hard to do in any animation. Most animation showing walking has a problem with slipping, it looks immidiatly wrong to us.

    A robot, and no robot so far looks anything like a real human, just doesn't look real enough to trigger the feeling that it is wrong.

    The uncanny valley is not restricted to animation or future robots. Ordinary humans got it as well, watch someone with a glass eye or scarring that is not obvious anymore but still large. Burn victim with a lot of corrective surgery whose face is close but at the same time a million miles removed. A burn victim with just a huge nasty scar might be shocking, but that is soon passed, but that face that is almost normal but isn't, that gets the constant stares.

    The uncanny valley is not a theory or a measurement, it is simply the observation that people accept a :) as a face but a 99% realistic rendering is instantly picked as fake NOT for missing a pixel but because the eyes are at an unnatural angle.

  • by ArsonSmith (13997) on Thursday January 21, 2010 @05:30PM (#30851808) Journal

    Most of those are just Machines. Robots should have a bit of autonomousness about them.

  • Nah... (Score:3, Insightful)

    by denzacar (181829) on Thursday January 21, 2010 @05:33PM (#30851860) Journal

    That there is clearly a robot with a head of a human looking doll.
    It is not even trying to appear human - it is trying to appear like a doll.
    So, we file it mentally under "D for doll" or "R for Robot" [wikipedia.org] and ignore the uncannyness as we are just approaching the valley.
    The valley actually starts to show with examples like this, [wikipedia.org] and Repliee. [osaka-u.ac.jp]

    And here is where TFA falls on its uncanny ass:

    If machines can trigger cognitive dissonance in the human brain, roboticists must continue to carefully tweak their creations, to avoid individual revulsion and even a society-wide blowback.
    That would be a major concern for the designers and manufacturers of the coming generation of social robots.
    It would be, if the uncanny didn't evaporate on contact.

    A Hypothetical Chasm
    David Hanson, a roboticist whose company, Hanson Robotics, specializes in ultra-realistic robotic heads, actively seeks out the uncanny.
    He keeps the motors in his rubber-skinned faces noisy and overtly robotic, and sometimes presents these lifelike talking heads mounted on a stick.
    And for better or worse, even the shock value of Hanson's buzzing, decapitated heads doesn't stick around for long.
    "In my experience, people get used to the robots very quickly," Hanson says. "As in, within minutes."

    It is really hard to argue with an article that so blindly ignores the very topic it is talking about.
    Its not the robots (of any kind - humanoid or not) that fall into the valley. We KNOW robots.
    They are just another version of all those mechanical puppets that have been around for centuries.

    It is the human simulacrum that freaks us out.

    A photo of a nearly perfect humanoid head freaks us out because it registers as a human head that is "not OK" in some way.
    We "feel" that there is something wrong with it.
    But as soon as it starts to whir and buzz and click and move rhythmically like a robot - our perspective changes and it is no longer "strange and alien and wrong".
    It "becomes" a robot.

    That is why the photos of those supposed robotic inhabitants of the valley are freaky, while people find CB2 [youtube.com] to be "cuddly" in person.
    Or why this one [youtube.com] becomes increasingly freaky once the camera zooms in to show just the face.
    Instantly, it is no longer a robotic mannequin we are looking at, but a dead human head - smiling.
    Then, as the camera zooms out, and the robot starts to move and talk - it is once again a humanoid mannequin, a moving statue, a robot.

  • by Scrameustache (459504) on Thursday January 21, 2010 @05:55PM (#30852294) Homepage Journal

    Everything we have is designed to work with our humanoid bodies, so if we want to make a device that interfaces with those things, it will work better if it shares the humanoid design.

    To an extent. A humanoid form for domestic robots would seem useful but we see that a roomba does a pretty good job and it's nothing more than a flat disc.

    Exactly, it does -one- good job.
    If someone breaks the android challenge, there's no need to redesign every aspect of human life with purpose-specific robots for every single job.

    P.S. That logic was much better explained by Asimov, from whom I was convinced.

  • by Wraithlyn (133796) on Thursday January 21, 2010 @06:04PM (#30852476)

    As another poster pointed out, Cameron never claimed he "invented" mocap.

    He has however, perfected it, via the facial capture stuff he added, to the extent that "performance capture" is now a more apt label than simply motion capture, and can be used large scale.

    Gollum required a small army of animators to complete Serkis's performance (especially Weta Digital's Bay Raitt, who crafted Gollum's facial expressions [http://www.extremetech.com/article2/0,1558,1554342,00.asp]).

    The detailed expressions on the Na'vi in Avatar, on the other hand, are almost entirely (95% is the number being thrown around) taken directly from the actor's performances, without requiring extensive intervention from animators.

    That is a breakthrough in filmmaking IMHO, and Cameron deserves kudos for it. The new 3D tech he was instrumental in developing (or at least championing), and his extensive use of a virtual camera to compose scenes after the fact, are also very impressive.

    I also disagree with your dismissal of why Avatar avoided the "uncanny valley"; the Na'vi seemed like actors in blue makeup, not "so alien you don't even associate them with humans".

  • by Patch86 (1465427) on Thursday January 21, 2010 @08:59PM (#30854904)

    It's lazy thinking, that.

    If you want your floor vacuumed by your humanoid robot then you'll also need to buy him a vacuum cleaner with which to do it. Considering the low cost of computer equipment now, why not just jam a tiny electronic brain in the vacuum which you'll have to buy either way?

    Same goes for most things. Why buy one car and one humanoid robotic chauffeur when you could just buy the car (which you have to buy either way) and stick a computer and a couple of sensors in it? Why have a shed full of gardening equipment and an android gardener when you could just bundle all your already-required tools into a handy automated bundle?

    And that's not all of it. The idea of having a machine going around and controlling other machines by way of buttons and leavers it quaint- why wouldn't all of these devices just network with each other? Why have an android turn the steering wheel and push the pedals on a car (and designing it with hands and feet for just this purpose) when it could just connect (wirelessly perhaps) and order the car what to do? Why would it even need to be there to do that?

    Far more sensible to have your big "house" computer nestled away somewhere, controlling all your devices for you over the airwaves. Like HAL, only with less psychopathy.

The test of intelligent tinkering is to save all the parts. -- Aldo Leopold

Working...