Every time I try to put away my obsession with the uncanny valley, some new robot or robot-story invades my world and reanimates that fascination all over again. Regular readers of this blog will know that I first became interested in robotocist Masahiro Mori's theory of the uncanny valley back in October of 2010, when I first speculated that Mori's theory suggested a fascinating non-human/human distinction that was structurally parallel-- but not identical-- to the way we think about the simulation/real distinction. Since then, I've written a series of posts on the uncanny valley here, this being the seventh in that series. The catalyst for this most recent uncanny valley revival is threefold, motivated by (1) the release of videos from the Geminoid project's test of their new robot "Geminoid-DK," (2) a recent NPR story on the ethical implications of "social robots," and (3) the reflections of Brian Christian, who recently served as a human participant in the contest (a variant of the Turing Test) that determines the Loebner Prize for Artificial Intelligence winner. (Christian has written a book, The Most Human Human, about that experience that I haven't finished yet.) Since I'm still reading Christian's book and since addressing all three of these topics would make for a overly lengthy read, I'll separate them into separate posts. First up: Geminoid-DK.
Geminoid is a Japanese research group that develops remote-control doppelgängers (or, as they call their products, "real-person based androids"). Geminoid describes its work as interested in both in the mechanical engineering aspects of android development (i.e., "effective tele-operation interface or generation of natural, humanlike motion") and also in the development of androids that possess what they call a "human presence" (or Sonzai-kan in Japanese). If you haven't yet seen their latest development, Geminoid-DK, here are some short videos of him/it:
The research group Geminoid (from geminus, Latin for "twin") is the brainchild of Hiroshi Ishiguro (aka, The Man Who Made A Copy of Himself). The image at the top of this post is Ishiguro and his own robot-double. "Geminoid-DK" in the videos above is the most recent and most advanced-- meaning, most "humanlike"-- creation so far. Ishiguro has created several such androids that you can see here. One of his earliest was a copy of his daughter, and it is reported to have unsettled her so badly that she refused to step foot in his lab after seeing her mechanical doppelgänger. Ishiguro's daughter's response confirms, in a way, the success of his project, which aims to manufacture robots that are humanlike enough to transmit some element of the Sonzai-kan (Japanese for "presence" or "authority") of their real-human models. But what does "humanlike" mean? What is Sonzai-kan?
What I find the most interesting about Ishiguro's approach to his project is that he understands the degree to which our attempts to re-create ourselves ("mechanically" in this case, but I think the insight applies equally to our literary, artistic, and even philosophical attempts to do the same) brings us into sober confrontation with that which is by definition impossible to copy: the idiomatic, the idiosyncratic, the ἴδιος (Greek, idios, meaning "pertaining to one's self, one's own, belonging to one's self"). My intuition is that this is what is meant by Sonzai-kan, and a similar "presence" or "authority" is intended by Western variants of Sonzai-kan like "soul," "personality," or "identity." And my intuition is that THIS is what we find at the heart of the uncanny. Ishiguro's daughter (pictured left with her doppelgänger) likely would not have had the same reaction to a copy of one of her father's other androids, but she experienced the copy of herself as uncanny. Is that because the simulation or manufactured creation presents itself to us as having no sui generis charactertistics, no Sonzai-kan, no real "presence" or self-generated "authority"? As I attempted to argue in my first iteration of uncanny valley reflections, the copy is never non-pareil. Art that aims to represent (or re-present) the human will always confront this problem. Or, at least, it will until we radically redefine what we mean by "the human"-- which is what I think arts like Ishiguro's are attempting to do.
A colleague of mine, an author and Professor of creative writing, recently recounted a story to me of one of his students whose fiction pieces left much to be desired. The student's problem, he explained, was that she seemed unable to relay her characters and situations in anything more than a reporter's narrative voice. There was no "life" to the student's prose-- and this is my interpretation of his complaint, not his words-- because the persons, things and states of affairs she was creating/recounting had no "presence." No Sonzai-kan, no redolence or affect or feeling, no familiarity. They were, like reporter's accounts characteristically are, bad copies of the real. Because she was writing fiction, her product was doubly unsatisfactory: it possessed neither the inimitable presence of the real, nor the simulated presence of art. It did not, and could not, move its reader or inspire in him anything like an experience-- not even the experience of the uncanny-- nor could it inspire any of the human affects (joy, fear, revulsion, compassion) generated by experience.
It is the generation of those affects that I think Ishiguro is attempting to equip his robots with, and he clearly believes that the only way to do this is by figuring out how to "transmit" Sonzai-kan. The problem, of course, is that Sonzai-kan-- whatever that is-- stands between the real (whatever that is) and its simulation (whatever that is) and seems to refuse us the luxury of their mutual contamination.
But more on this later... Next up in the uncanny valley series: (1) Social Robots and (2) The Most Human Human. Stay tuned!