I was at HRI 2009 a few weeks back and had the chance to sit down briefly with Prof. Hiroshi Ishiguro to talk about his research. Prof. Ishiguro is famous for his wide-range of research in robotics, especially in the area of androids. You may remember seeing somewhere a copy of himself - the Geminoid. A Geminoid was coined by Prof. Ishiguro and is defined as a "tele-operated android of an existing person."
Below is an edited abstract of his thoughts. (You can read a previous post about the Geminoid here.)
"When you look at history, humans have always tried to build machines that can replace labor. Moving forward, whatever machines can not eventually replace will be the essence of human beings. I want to see what that is. That's why I do robotics research - to find out about what makes humans human. When you look at an android as a reflection of a human being, you realize that the physical body of the person is meaningless and that humans are a spiritual being"
"People talk about building autonomous robots, but what does autonomous mean? From my perspective, humans are not autonomous. A human being is basically just a communication device. If you put a newborn baby in a room and not have him/her interact with anyone nor anything, the baby will not become a 'human.' Humans learn from others and use that knowledge and past experience to take action. For example, when reporters come to my lab, I ask them, 'How did you use your brain to come here?' Most of them say that they just followed directions. In most human activity, we are just using our sensors - our ears and eyes, etc. - and using that sensory information to follow what people tell us or expect. Are we discovering meaning from the sensory information and acting uniquely based upon it? The answer is - not really. Our sensors are not for extracting meaning but to work as a communication device."
"So how is that different from robots that use sensory information and follow orders? Not much. And since humans are not autonomous, how can it be possible for us to develop an autonomous robot? Just like humans, robots need someone or some kind of mechanism externally that creates and provides them with useful information."
"What is intelligence? People have answered this question saying that intelligence is a function of memory, learning and reasoning. I don't think that's wrong. But even if robots have these functions, they do not show intelligence. My belief is that memory, learning and reasoning are functions and they do exist, but that intelligence is something indefinable. Intelligence is a subjective phenomenon that appears within interaction, meaning interaction between humans and humans or humans and robots. What you think or feel is intelligent is what possesses intelligence."
"The same goes with emotion. People believe that there are such things as emotion and intelligence, just because it's more convenient to think that these things exist."
"Perceptual information processing methods continue to make progress, but I think the capability of these systems should be closer to what humans can achieve and not much higher, especially when it comes to robots that interact with people in everyday life. Rather than have a single spectacular sensor that does all the job but could make a fatal mistake, it would be better if you had a variety of sensors that are not as accurate but work pretty well with each other and can avoid critical errors. Humans will be much better off with systems that have weaknesses like themselves because they will know what to expect. I believe that developing ways to integrate and use inaccurate information collected from a variety of sensors is crucial for commercializing robots, but there are not as many researchers moving into that direction compared to those that are involved in advancing the quality of each perceptual information processing technique."
