||[Nov. 10th, 2004|01:06 pm]
I just read a review of the upcoming Christmas movie, The Polar Express,, directed by Robert Zemeckis, who adapted one of my favorite books, Contact, into one of my favorite movies (coincidentally also called Contact).|
In this CNN review, the author applauds the technical achievement of the computer generated special effects, but repeatedly mentions how the characters have a tendency to look "creepy" or even downright "horrifying." That immediately brought to mine an article I had read over ten years ago, about a roboticist named Masahiro Mori, who articulated the idea that as machines become more and more human, the emotional response, or empathy, we feel towards them naturally increases. However, this trend is not linear. Mori describes a point, very close to "fully human," when the emotional response actually reverses, and plummets down into revulsion, before resuming a steep climb that ends at total, or near-total acceptance. This sharp dropoff is what Mori called "The Uncanny Valley," and the principle goes a long way to explaining why Honda, Sony, and until recently, Pixar have not attempted to make truly human-looking robots and characters.
The gist of the story is that something that is clearly not human, but has human-like features, say smiley faces doodled on paper, anthropomorphic Disney animals, and clearly artificial robots, are seen as the foreign entities that they are, and as they become more human, we are able to relate to them more easily. However, once they begin to really enter the realm of human verisimilitude, we start to see them AS human. But with that shift in our frame of reference, small discrepancies suddenly become very noticable - the eyes that blink a little too slowly, the ever-so-slight woodenness of unnatural facial expressions, the ever-so-slightly stilted gait of bipedal motion. As utterly familiar as even the most antisocial of us are with what the idea of a "person" should be, this should come to no surprise.
The idea of the Uncanny Valley is demonstrated below, in this image I ganked from Wikipedia:
The vertical axis is emotional response, with the bottom being absolute revulsion and the top being absolute acceptance. The horizontal axis is the increasing level of anthropomorphism, or human-like attributes, with the left being a chunk of limestone and the right being something like an archetypal human being. You can see where the Uncanny Valley lies, where the emotional response shifts drastically and quickly falls to a negative response - this is the territory of monsters - the zombies, the pod people, the walking corpse.
So when The Polar Express comes along, bringing with it an utterly new and exciting concept of motion "performance capture," we suddenly have concededly cartoon-looking computer-generated characters that nonetheless display an utterly amazing range of life-like body movements and facial expressions. Except, as the reviewer deftly notes, where the performance capture system doesn't have any sensors - in the eyes and inside the mouth. And here lies the Uncanny Valley of Masahiro Mori. With otherwise fully expressive and visually believable characters, the manually animated eyes and especially the tongues (likened in the review to slabs of meat) stand out in stark and unnatural contrast. The effect has such a strong impact on the reviewer that he falls right off the cliff of the emotional response curve. "Dawn of the Dead. Children of the Corn." That's pretty harsh for a G-rated children's movie, any way you cut it.
One might wonder why the movie makers have not noticed or chose not to do anything about this freakish effect. Have they tried their best? Is this level of convincing humanity beyond the technology and vision of our engineers and artists? Or are they so intimately familiar with their creation that they are too close to see the ugly baby, the Frankenstein's monster that a casual movie viewer would instantly recoil from? My guess it's a little of all of this. But I'm a firm believer in technology, and the imperfect techniques that have the multimillion dollar funding to be pioneered in triple-A movie titles today will not only be perfected eventually, but will be achievable in real-time on home computers tomorrow.
And what about me? I'll definitely see the movie. I'll most likely notice the less-than-perfect animation, but I'm not entirely convinced that I'll fall into the Valley. While I've loved most of the CG movies I've seen (both Pixar and PDI make fantastic movies, CG or otherwise), I usually walk in and out with a technical eye - for me the delivery of the animation and rendering technology stands apart from the story and acting, and I tend to judge each separately. That in effect tends to subtract from any serious emotional attachment I would normally develop for the characters; I guess I'm just not ready to completely suspend disbelief just yet.
I've just found out that Roger Ebert apparently uses the concept of the Uncanny Valley in describing the believability of artificial characters. I guess I shouldn't be surprised, considering how apt the notion is. Oh well, so much for originality.