A robot that does not understand the human mind may become a threat to humans?!

I started my study in robotics around 1980 when microcomputers, which were small enough to be put on a desk, started to become available. I was working on developing cameras to be used as robot eyes. All of a sudden, one question came into my mind: “Do robots see?” Certainly, robots moved around, avoiding obstacles with information from the camera. However, is it possible to say that robots did that by “seeing” obstacles around them? What does it mean that humans “see” things? I became aware of the importance of human consciousness and began to think about developing a robot with human consciousness. At that time, nobody else had an idea like mine. Development of artificial intelligence that exceeds human intelligence was becoming increasingly popular and gaining momentum. My research was ignored and even criticized and condemned. In 2004, we publicly announced results of experiments on “mirror image cognition using a robot.” It was first broadcasted on the Discovery Channel in the U.S.A. and reported by Reuters and the Associated Press. Subsequently, it was covered by the German science program “Planetopia,” as well as by some Japanese TV programs, receiving attention through mass media around the world.

Why did my research start receiving attention? With modern AI and robot technologies, humans are now able to produce a super robot that could outperform humans intellectually as well as physically. We are now entering an era when humans are not capable of judging if advanced calculation done by AI is correct or not. So, what will happen to humans in the future? It is a possible scenario that a “reasonable” judgement of AI will suggest eliminating human beings. The other day, during an interview with a German national newspaper “Frankfurter Allgemeine Zeitung,” they expressed concern that robots could become a threat to humans. If so, should we create a program that commands robots not to harm humans, not to do this, or do that? However, in such a case, a robot would not be able to react to situations flexibly that are not pre-programed. As a result, a robot would remain merely a high-performance machine. To make robots a partner to humans, we need to create “a robot with human consciousness.” The reason why mass media around the world is paying attention to my research is because, I think, they agree with my idea.

A simple program cannot be structurally comparable with the human brain. The neural network of humans is unique. It can detect its own actions. In other words, it is self-aware. There is an experiment is called “mirror image cognition using a robot.” Babies don’t recognize their own image in a mirror until they become two years old. Babies learn by watching and imitating others. In addition, they start distinguishing self (the one watching and imitating others) from other people (objects to be watched and imitated) through recognizing their mother or the people who are taking care of them. A robot perceives its own image in a mirror as others because it is not connected with its own body. A robot will never change this perception. In light of this, we created a program in which it recognizes as special information that it can’t distinguish movement of self from that of the image in a mirror. This program was installed in the robot’s system in addition to another program that processes information about movement of the image in a mirror. As a result, the robot was able to judge its own image in a mirror as something special although it still recognized it as others. Humans are self-aware because of functions of the neural network, which controls human actions and emotions. It is a self-referencing program. Logic that studies the characteristics of logical systems is called metalogic or super logic. Nobody has created a program of human concept that includes metalogic. Aiming to create such a program, I have developed a self-referencing program. It identifies (becomes conscious of) specific information (meaning) while an upper-level function of the program receives information related to actions of a lower-level function.

With this program, it has become possible to explain most of the functions of the human brain such as emotion, creative act. Surely, not all of them have been clarified. For instance, S. Freud indicated in his theory of libido that humans have desire for their own survival. Thus, a robot with human consciousness would have the same desire as humans. On the other hand, a person does not mind making sacrifices to protect something important to him/her. Why so? It is a very difficult question to answer. If we manage to elucidate the mechanism of self-sacrifice, we may be able to produce a robot that protects humans by all possible means.