You may not have heard of Hanson Robotics, but if you are reading this, you have surely seen their work. This company has built the robot Sophia, a humanoid android that has repeatedly appeared at public events. Prior to that, the same company had built a strange robot that looked like Asimo with the head of Albert Einstein - or seen BINA48, which was interviewed by the New York Times in 2010. They even made a line from legendary writer Philip Dick (you know him for sure).
In other words, Hanson Robotics, with their brand of humanoid robots, is well-versed in this field. And the more painful it is to watch their progress. Anyone who communicates with a robot at least a little understands that in front of him is essentially a chatbot, with all its limitations. Even during an interview with BINA48, author Amy Harmon described the experience as frustrating - with "rare (and exciting at times) moments of understanding." This feeling is familiar to anyone who has ever interacted with a chatbot and got some reasonable responses.
There is no real intelligence under the glossy surface; although at first glance, of course, the car seems smarter than it really is. If you pull back on this surface layer - in the case of the Hanson robot - you are pulling on Frubber. This patented substance (Flesh Rubber - literally "rubber from the flesh", creepy) - incredibly complex. Up to thirty motors are used to control a person; they manipulate fluid cells to make the skin soft, pliable and capable of displaying various emotional expressions.
A quick combinatorial look at 30 motors will reveal millions of possible combinations; the researchers identified 62 of them as "human" in Sofia, although not everyone agrees with this statement. Of course, technicians who recreate facial expressions in the image of a human in a robot have surpassed a simple chatbot engine, but they also did not hesitate to program a few questions that exaggerate the expectations of the interlocutor after meeting the robot.
Hanson Robotics believes that, ultimately, much of what humans attribute to robots will depend on their faces and voices, as well as what they say. “Perception of personality is incredibly closely related to perception of human form,” says David Hanson, founder of the company.
Anyone trying to create a robot that won't scare people has to somehow cope with the ominous valley - the strange mixture of anxiety and disgust that people get when they see something that is close to a human, but is not human. Between cartoon humanoids and real humans lies the forbidden zone of robot aesthetics.
The concept of the sinister valley was introduced by robotics engineer Masahiro Mori, who insisted that roboticists should not try to accurately reproduce humans. Since anything that is not perfect, but already very good, will make people feel strange, the only way to get rid of this effect is to give up trying to create the perfect.
From a brain perspective, the idea behind this psychological horror is pretty simple.
Promotional video:
We know how to categorize things that are uniquely human or non-human. This is easy for us, even if these things are designed to interact with people. Think of the popular robots Aibo, Kibo, or just an ordinary talking speaker. Something that tries to recreate a human, but does it wrong, elicits a startled response just like a slightly out of tune guitar or rearranged furniture in the house. This creature just doesn't suit us.
You can completely abandon the idea of a sinister valley. David Hanson is not a fan himself. He believes that great works of art have often tried to recreate humans, but the ultimate goal of android robotics will likely be to create robots that are closer to humans than works of art.
In the meantime, Hanson and other scientists are conducting experiments that either demonstrate the overestimation of the ominous valley, or confirm its existence and highlight its boundaries.
The classic experiment involves the gradual transformation of a cartoon face into a human face with robots in between - because it is in movement that the fear of the "near-human" most often lies. Hanson argues that the inclusion of cartoonish features can help and that the ominous valley will slowly dissolve as new generations grow up surrounded by bizarre robots. While Hanson may not take the sinister valley seriously at all, he tries to master it with each iteration.
Hiroshi Ishiguro is one of the last roboticists to plunge headlong into the valley.
Building on the work of pioneers like Hanson, the study of robot-human interactions is moving towards the frontiers of robotics and also making progress in the social sciences. It is usually hard to repeat what you don’t understand, and we still don’t understand much about how we interpret the constant streams of non-verbal information that flow through us as you interact with people in the flesh.
Ishiguro takes human imitation to the limit. He not only tracks and records the physical movements of people from video clips, but also creates his robots based on real people; the Repliee series began with his daughter's "replicant". I had to create a rubber replica of her entire body. Then he created Geminoid, a copy of himself.
As he grew older, he realized that it would be more effective to recreate his replica with cosmetic surgery instead of re-sculpting his face, each time with more wrinkles. “I decided not to get old anymore,” he says.
We love to throw abstract concepts and ideas into the air: machines replace people, machines take care of people, sex with machines, merging with machines. But weigh any of these ideas in your open hand - and you will find a huge gap between the expected and the actual. We are still far from a world in which day-to-day interaction with robots will be a chore and not an academic study.
Ilya Khel