Photo-Realistic Animation (Part Two): The Uncanny Valley
In the earlier blog dealing with the issue of photo-realistic animation, we looked at basic terminology and types of animation, which is a central storytelling tool in training. Four types of animation were identified: stick figures, cartoon characters, realistic characters, and photo-realistic characters. The last type was reserved for this blog. Photo-realistic animation injects an interesting and often unsettling element into the field of animation. In order to appreciate this element, one needs to investigate the nature of photo-realistic reproduction and the central issue, the emergence of the uncanny valley. Moreover, when assessing negative features associated with photo-realistic animation, one needs to place it in a broader cultural context.
The Work of Art in the Digital Age of Photo-Realistic Reproduction
The title chosen for this subsection is a conscious allusion to the famous essay written by Walter Benjamin, “The Work of Art in the Age of Mechanical Reproduction.” Written in 1935, Benjamin discussed how artistic works, such as painting and even theatrical plays, were devalued due the process of mechanical reproduction. He made this conclusion based on his argument that the true artistic value, the “aura,” of the work was lost in its reproduction. In short, art lost its cultural authenticity when it became a product of consumerism instead of aesthetic reflection. The culprits of his cultural descent included printmaking, photography, and cinema. This neo-Marxist critique, more importantly for Benjamin, explained how capitalist societies politically deluded and pacified the masses.
Walter Benjamin, c. 1924
Placing the issue of political analysis aside for the moment, the “artistic” goal, for some, when engaging in “mechanical reproduction” was to create duplicates as close to the original as possible. In many ways, the photograph was the model for this endeavor. We now are at a very different stage regarding the types of technology utilized in reproduction.
In the digital age, as noted in a previous blog, we see a comparable cultural lament; new forms of reproduction are replacing the older “mechanical” forms, the rise of the digital at the expense of the analog. While the photograph was viewed as the epitome of accurate reproduction, digital photographic formats have created the means for expanded content manipulation as well as the ability to generate new content devoid and separate from any external reality. This is where the world of photo-realistic animation comes into the picture.
The previous blog differentiated amongst four different forms of animation. Excluding photo-realistic representation, the other forms harbored no pretence of even trying to mimic reality in a manner that would fool an observer. Photo-realistic animation, however, does seek to replicate scenes, as a conscious goal, as one would see them in a photograph or live-action cinema. For some, this is where a problem emerges. Like in the era of mechanical reproduction, concerned parties view the issue of authenticity as the key element. For Benjamin, the loss of authenticity eclipsed the political identity of citizens; for those concerned with photo-realistic representation, that gap in authenticity generates an unsettling sensibility regarding the nature of human identity itself.
The Uncanny Valley. The concept of the uncanny valley was coined Masahiro Mori, a Japanese professor in robotics. Born in 1927, Mori published the seminal article titled “The Uncanny Valley” in the journal Energy. Simply stated, Mori’s thesis is the following:
“I have noticed that, in climbing toward the goal of making robots appear like a human, our affinity for them increases until we come to a valley (Figure 1), which I call the uncanny valley. . . . One might say that the prosthetic hand has achieved a degree of resemblance to the human form, perhaps on par with false teeth. However, once we realize that the hand that looked real at first sight is actually artificial, we experience an eerie sensation. For example, we could be started during a handshake by its limp boneless grip together with its texture and coldness. When this happens, we lose our sense of affinity, and the hand become uncanny.” (For an English translation, see: http://ieeexplore.ieee.org/abstract/document/6213238/?reload=true.)
Mori argues that the genesis of the uncanny valley is based on the human anxiety over death and the drive for self-preservation. Others have linked its genesis to the drive for mate selection, cognitive dissonance, and religious sensibilities among others.
Mori posits a strong relationship between the effect of movement and the uncanny valley. If robotic movement begins to mimic human motion, then its level of affinity is increased. However, simple replication of certain types of motion is not enough, as there is a large-range of human movements that might not or could not be programmed into a robot. As a result, Mori suggests that scientists consciously pursue a nonhuman design in order to generate a safe level of affinity.
While Mori’s concept specifically dealt with human-like qualities in robots, it is relevant in the entire range of projects that seek to replicate human qualities, which include appearance, movement, and voice.
The Range of the Uncanny Valley
To best appreciate the visual sense of the uncanny valley, one needs to be able to clearly see and interact with the range of images, so a selection of large images are included below:
Android: Hiroshi Ishiguro and Double (Hiroshi Ishiguro Laboratory), 2013
Movie: Polar Express (2004)
Video Game: Medal of Honor: Warfighter
In comparison, let’s look at a spectrum of images created, from the most-robot like to the most human:
Maya B. Mathura & David B. Reichling
Let’s look at other ways the uncanny valley phenomena has been observed in the range of evolution:
- Monkeys: Princeton researchers presented evidence arguing that monkeys also respond the uncanny valley.
- Computer chat bots have been categorized in the realm of the uncanny valley.
For many years if not decades, the ability to generate life-like creations have been limited by restrictions in technology. If we had better technology, wouldn’t we have better representations and reproductions? Well that’s true. Ironically, the issue of hyperrealism has not only impacted the digital world of computers but also the “authentic” art world of canvas-based painting.
Phillip Weber, Bless 3, Antonia (2012)
There are a number of factors that force one to reconsider the operative concept of the uncanny valley:
- Human Robot/Interaction: In a recent significant research article in the field, Jakub A. Zlotowski and his colleagues posit that the perception of the uncanny valley can be impacted by human/robot interactions. That is, many of the studies on the uncanny valley focus on experiments with subject responses to digital images and/or videos. Zlotowski determined that “repeated interactions [with the robot] was sufficient to reduce eeriness irrespective of the robot’s embodiment.” This factor, of course, does not obtain in the same way when investigating the uncanny valley in scenarios based in the context of videos itself, for example, in the field of training.
- Japan and Robot Culture: While the notion of the uncanny valley originated from Japan, it is useful, extremely useful, to highlight the cultural differences between the East and the West. It should not be a surprise, simply based on popular media coverage, that the Japanese have a strong affinity for modern robot culture. This sensibility, though, is rooted in the Karakuri tradition, which shielded technology to promote “feelings and emotions;” a similar sense existed in Eastern attitudes toward marionettes. In comparison, the Western automata tradition was based on mimicry and scientific knowledge.
- Improving Technology: In a recent article in the magazine Wired, Sandra Upton argues that advances in artificial intelligence brings into question the continuing validity of the uncanny valley. Upton argues that the algorithms in various digital tools, such as Adobe’s Sensei, are so powerful as to create products that almost are indistinguishable from reality. She states bluntly that this could be, actually, the technological basis for Fake News.
When reflecting upon these discussions, we are directed back to some of our original questions. Benjamin investigated how mechanical reproduction impacted our ability to maintain authenticity and class insight. Mori broached the same question in the context of the fabrication of robots; for Mori, though, the matter dealt more with the issue of authenticity and human consanguinity with robots.
All withstanding, one is forced to deal with the 800-pound gorilla in the room: the limits and range of artificial intelligence. Mori dealt with this issue in his 1974 book The Buddha in the Robot: A Robot Engineer’s Thoughts on Science and Religion. In this work, he claimed that robots can give us insight into our own ethics and concluded that robots are capable of cultivating a Buddha-like nature. So our initial investigation looked at the experiential pitfalls of using photo-realistic animation and ended up by asking existential questions about the nature of humanity. But how does this factor into the more discrete issue of training? In short, as suggested by Mori himself, one should tread carefully in using animation that is too accurate, as technological limitations may lead you to slip into the uncanny valley.
As always, let us know what you think!
Craig Lee Keller, Ph.D., Learning Strategist