By Toshi Doi
President
Sony’s Intelligence Dynamics

Sony's Aibo and Qrio have been widely ack­nowledged as unique products with outstanding technology in the field of personal entertainment since their respective introductions in 1999 and 2000. They not only behave autonomously in a human environment, but also can sometimes interact with humans. The fact that more than 140,000 Aibo units have been sold in six years proves the degree of this robot's success.

However, having been responsible for the development of these entertainment robots, I see a fatal limitation in the current technology. Since the robots' behavior is strictly and solely based on the program that engineers have defined for them in advance, they lack flexibility. Such a robot can easily get into trouble if exposed to an unknown environment, for example.

Al­though Sony scientists implemented functions that simulated growth or development in the robots, that helped only a little. The "fully de­signed" robot, I strongly believe, will remain worlds apart from the auton­o­mous machine that people expect when they hear the word "robot."

This problem doesn't exist in robotics only but applies to all kinds of other systems. In the near future, very high-performance computers will be em­bed­ded in electrical appliances, cars and networks. I am sure we will ex­pect these systems to show flexible, intelligent and human-friendly behaviors. The "fully de­signed" approach will no longer do. The time has come to reconsider machine intelligence.

To reach beyond the current state of the art, robots should autonomously learn and adapt through interaction with the environment. Several other research fields, such as cognitive science and brain science, are facing similar challenges. Since the integration of work in all these areas will be necessary to define this new type of machine intelligence, I have proposed a new research field I call intelligence dynamics, a discipline aimed at the integration of cognitive developmental robotics, embodied cognitive science and computational neural science. I hope that talented researchers will come together in this new field.

I am now leading Sony on a quest to implement an "engaging agent" that will be available for long-term interaction with humans. Equipped with our latest research results, Qrio, a sophisticated bipedal humanoid robot, is able to learn how to ring a bell and roll a ball—a small step forward.

Today, high-speed networks gather sensor information, and machines can develop themselves and work for humans in flexible ways. I'm confident that technologies developed through these activities will enrich our future.