Slashdot Mirror
Why the Uncanny Valley Doesn't Really Matter
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."
I think it's weird that some people have a fascination with humanoid robots in the first place. seems like most Japanese robot efforts (at least those that make the press here) are in that vein. sure, there's a golden place in the future for replicants and sex slaves, but to me those seem like fairly narrow niches. if I'm designing robots with the goal of getting useful stuff done, I certainly wouldn't start with a humanoid layout, with all respect to evolution ;)
I admit it, all the Japanese robot coverage I see is either kawai-oriented or thinly-veiled sex-slave oriented (or both). no doubt that only reflects my taste in paper an online media...
there's no Uncanny Valley for Roombas.
Many biologists think that dog attack cats and dolphins attack sharks for the reason that the latter of each pairing is too similar to the former of each, that the former might draw the comfort of familiarity until the revulsion of what appears to be an abomination of one's own species at closer inspection -- an "Uncanny Valley in the wild" so to speak. Are dogs and cats friendly once they've become acquainted? Oftentimes. Are sharks and dolphins friendly after becoming acquained in a controlled environment? I'll leave that as an experiment up to the user.
I am the richest astronaut ever to win the superbowl.
TFA says that
- it may be more nuanced than people originally thought [i.e. the "absolute level of human-likeness" may not be the problem, but mismatched levels [great skin, awful eyes don't go together and are jarring]
- may have gender bias
- seems to depend on you viewing something remotely in 2d vs interacting with something real in the same room [the latter didn't seem to engender the same creepyness in those tested]
Since I don't live in japan nor do I visit robotics labs, I don't have much occasion to interact with near-humanoid robots. So my UV experiences are limited to movies and video games.
I remember seeing the Final Fantasy: Spirits Within movie in the theater and just minutes into the movie I was convinced I was looking at real humans. Or rather, there was nothing in the film that made me dissociate with the characters; they were as "real" to me as watching actors. I kept trying to "zoom out" of the movie/picture and try to critically evaluate the job they did rendering the characters, but I kept defaulting to treating them as humans and getting sucked back into the movie. Mission accomplished on their part, i guess.
I think the UV effect is definitely apparent in 2D matter -- as a fan of anime I am more inclined to "accept" characters that are absolutely impossible.. both physically and emotionally.. but which do not attempt to persuade me they are more than they are. Yet when video game makers get something slightly wrong it _is_ a jarring experience. I've seen video game cutscenes where there are clearly a lot of polygons and textures and art time involved...but something just seems off and instead of you being wowed [or ideally, _not wowed_] you are left feeling disappointed. You know everyone worked hard to try and make the scene but they absolutely did not pull it off.. and the game experience is worse as a result. Mistakes that land your artwork into the "UV" category turn people into videogame/art critics instead of people enjoying an interactive experience.
My opinions are my own, and do not necessarily represent those of my employer.
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.
And, anyway, robot technology is improving every year, and as such they're doing their best to cross the uncanny valley and getting better all the time. Meanwhile, on this side, we're doing our best to cross to their side, led by Michael Jackson, Cher, Tila Tequila, and Jocelyn Wildenstein.
Nostalgia's not what it used to be.
I'm white and speak Japanese without an accent (Dave Specter has me beat on vocabulary, but he sounds horrible). I get the same initial shock when I open my mouth. If they're under 15, they just tend to stare for a minute. After a couple sentences, most people calm down and everything is normal. However, when I leave I often hear whispered comments about how much of a shock the experience was.
A robot needs to do more than sit and place and perform a task (even if it is engineered to do it well). Such a loose definition does to the term "robots" what has been done to "cloud (computing)" and "nano (machines / structures)" that everyone wants to slap or back-date on their project to make it sound important / relevant.
As I see it, part of the definition of a robot includes movement. As with the roomba, it has the ability to move around to perform it's task and not be restricted to performing it's task in a single place (which would be a useless vacuum cleaner). In factories, robotic arms may stay mounted in place, but the arms themselves move themselves or parts to assemble cars.
I think the phenomenon labeled Uncanny Valley is perfectly valid and perfectly irrelevant. At least, as long as it's framed solely in terms of appearance. It's trivially easy for people to relate to Wall-E. It doesn't matter in the slightest that he looks only very vaguely cubically humanoid. He could be, in the great Disney test of yesteryear, an animated flour sack. As long as he appears to express emotions, the machine instantly becomes "he" to us. (Or she, depending on the mannerisms.) A walking talking RealDoll will still be a creepy failure as long as it doesn't move right. A box with treads will succeed, as long as it can act human (or possibly canine).
In desktop computers, it's the software that's inadequate, as every attempt at game AI demonstrates. In robots, there are still a few things that are inadequate in the hardware, but truly it's still the software. Roombas have zero personality.
Slashdot carried the story about the little robot let loose in Central Park a while back. It was nothing but a bump and go car with a flag on it and a sign saying "help me get to point X" and people actually did help it. Now consider what would have happened if it had been a Wall-E bot. I'd bet money that if a little robot hunched down, tapped his manipulator tips together, tipped his cameras into a configuration vaguely reminiscent of a worried expression and shuffled his treads, and held up a sign asking for help to reach point X that someone would have literally stopped what they were doing, taken his hand, and led him the entire way there.
It's the personality, stupid. (To coin a phrase...)
GP is saying he disagrees with uncanny valley theory, and you're saying that he's mistaken because UV-theory is true, he just doesn't get it. You may be right, but you're also engaging in circular reasoning. It would be better to say, "UV-theory contends that the reason all the things you listed are easy-to-take..."
Anyway, first of all, I would dispute that all of those things are "easy-to-take"... I never got into Farscape because I found it kind of...creepy. Perhaps I'm unique in that respect but nevertheless it seems to me that many people are creeped out by certain types of puppets (Chucky?) (Clown dolls?), by industrial robots, even by fake bugs and snakes.
Second, as we're putting up hypotheses, I would guess that the reason we like Muppets is because virtually every single person under the age of 40 has grown up with them since infancy, so they become part of the brain's notion of what is an acceptable deviation from the norm. I'd argue that the same also applies to traditional cartoons v GC. We grew up with traditional animation techniques as the default, so we quickly learned to accept Bugs Bunny et al. as acceptable. Yet I seem to recall being vaguely creeped the first few times I saw pre-WWII style animation with its weird puffing and undulating movements. I'd guess that a 21st century child is much less likely to find anything odd about CG animation than her parents do.
Third, (this is a criticism of UV-theory as a whole) I don't know that it makes sense to place the revulsion behind a poorly designed CG model under the same umbrella as one's potential revulsion over actual metal-and-silicone robots. CG is fake, an illusion, and while we may allow our disbelieve to be suspended, we still know that the animated character is safely detached from reality. On the other hand, a robot, a realistic one such as CB2, seems to (or perhaps does) possess the quality of agency. It interacts with the real world and can engage in unpredictable behaviour. So it might potentially get OUT! OF! CONTROL!, and harm us, and there's something possibly off-putting about that. Or, until I've accustomed myself to its movements and its failure modes, my limbic system will tell me that that huge industrial assembly line robot might malfunction and squish me like a bug, Whether it's "uncanny" looking or not. Or my future attack-dog-droid might mistakenly view me as a target, and make short work of me without the slightest twinge of remorse. That's a much more visceral concern to me than whether or not the attack dog's skin is too close in texture to real dog skin.
So I guess I agree with TFA, that UV doesn't really matter, at least when it comes to real bots.
There are two kinds of people: 1) those who start arrays with one and 1) those who start them with zero.