I know i’m wandering a bit off-topic here, but this post over at Mind Hacks was too fun to let it pass unnoticed: according to recent news, Hiroshi Ishiguro, director of the Intelligent Robotics Lab (IRL) at Osaka University in Japan, has created an android double of himself. You can see Hiroshi in the figure on the left (stolen from this recent Scientific American article). If you didn’t recognise the android at first, don’t worry: Ishiguro’s team have been conducting a variation of Turing’s test where androids are shown to humans for around two seconds; 70% of times people take the androids for humans. The experiment is part of a study trying to explain the so-called uncanny valley effect: the emotional response of humans to robots increases as the robot is made more human-like, up to a point where a too close resemblance causes the empathy to abruptly fall. A sort of too good to be true effect. According to the IRL experiments, this may be caused by the stillness of androids: it’s impossible for humans to stand perfectly still, and that’d be what makes robots uncanny. To prove their hypothesis, Ishiguro’s team has added micro movements to their toys. As the result of the experiment with 20 subjects, 70% of the subjects did not become aware they were dealing with an android when the android had micro movements, but 70% became aware with a static android.
Of course, there’s much more to Android Science (as they call their interdisciplinary endeavours) than micro movements. This beautiful paper gives an introduction to some of the challenges faced by these intrepid researchers, ranging from an understanding of cognition and intelligent behaviour in humans to the technical challenges of creating an adequate sensor system for the robot: besides the piezo-electric films under their silicon skin, the androids use a sensor network distributed over their environment that includes video, sound and infrared motion sensors. (Yes, that’s a bit like cheating, but the results are impressive: you can get a glimpse of the science and technology behind sensors and data processing in these posters by the IRL researchers) The paper also discusses in length the Turing tests mentioned above.
I just wonder, why do the androids look so angry?