Thursday, June 14, 2012

Updating the Uncanny Valley (for English speakers, at least)

The uncanny valley is a strange and scary place. Hatsune Miku took a trip there recently and the results were nothing short of disastrous (to put it mildly).

But what is the uncanny valley? It's a theory put forward by Japanese roboticist Masahiro Mori over 40 years ago that explains why humans are revolted by near-perfect robotic imitations of themselves but have no problems labeling non-humanoid robots like R2-D2 as "cute".

Mori's essay was hurriedly translated into English back in 2005 and contained numerous linguistic errors. This lead to a fundamental misunderstanding of the finer points behind the concept of the uncanny valley by non-Japanese speakers.

Many of us recognize the underlying concept of the uncanny valley even if we're not familiar with the term itself thanks to animes like Ghost in the Shell (particularly Innocence). Further, the negative reactions of audiences to the CG "breakthroughs" of movies like Polar Expressand Beowulf show that Mori was right on the money when he claimed that our acceptance of an artificial human representation decreases sharply when it is close to being accurate save for a few fundamental flaws.

A new translation of Mori's seminal paper from 1970 has been released, and it looks to clear up some of this misunderstandings.

One of these is a robot, the other can do a mean version of The Robot after a few sake bombs

The major misunderstanding of the uncanny valley comes from the inability to translate Japanese into English neatly. For example, Mori created a graph to illustrate the phenomenon of the uncanny valley, and used "shinwakan" (familiarity) on the Y-Axis to plot the reaction of human beings to near-human robots.

The problem with the 2005 translation is that shinwakan is more of a concept and less of a word (anthropology nerd alert: all words are actually concepts. Moving on!) While it can translate to easy phrases like "familiarity", "comfort level", or "likability", those phrases do not encapsulate the full nature of shinwakan.

Karl MacDorman, one of the translators who worked on the new 2012 revision of Mori's paper, describes shinwakan as such:

"I think it is that feeling of being in the presence of another human being — the moment when you feel in synchrony with someone other than yourself and experience a 'meeting of minds'... Negative 'shinwakan,' the uncanny, is when that sense of synchrony falls apart, the moment you discover that the one you thought was your soul mate was nothing more than smoke and mirrors."

The full text of the updated 2012 translation is available here, and while it's a dense read it is also a fundamentally fascinating one as it helps to explain why we find robots like this to be simultaneously fascinating and horrifying...