Learning from Animals, But Not Following Their Evolution

Toshitada DOI (Corporate Executive Vice President, Sony)

Although robots are becoming more and more familiar to human beings, their evolution process turns out to be entirely different from that of animals. In the early age of aviation, Lilienthal and the Wright Brothers studied how birds fly and how wings work, but ended up inventing propellers. That is also the case with tires. There are no animals with rolling tires. Engineering has produced some things that do not exist in nature.

The same thing can be said about the evolution of robots. It is almost an illusion to see a future in which human-like robots would help us in our social life. We should not maintain the idea of human-like robots, but rather think of ways to make full use of robots' potentials such as computing ability.

Two Types of Robots

When I developed the first version of "Aibo" (a dog-like robot), I was asked to make it a "seeing eye dog." That was a wrong approach. Of course, it would be wonderful if robots could help visually handicapped people. But such robots do not have to resemble seeing eye dogs. It could be a white cane with a CCD camera (to process visual objects with digital signals), which transmits such information as "red light" or "watch your step" to the palm of a person who holds the cane.

Current robotics can be classified into (1) the "practical" type that makes human life more convenient and (2) the "entertainment" type, where robots are pets or friends in essence. Robots of the practical type should take the shape that suits their respective purposes such as lifting heavy materials or making moves finer than human beings can possibly make. Health care robots do not have to stand on two feet.

On the contrary, robots of the entertainment type must surprise people and, therefore, the ability to walk on two feet or move on four feet would be quite suitable.

There has been rapid progress in robots' movement control ability for the last few years. On the other hand, little progress has been made in the area of intelligence. However, we can expect a big breakthrough in this area in the future.

A recent research topic is to improve voice recognition. Sony's most recent robot "SDR-4X" has a new system that has not been installed in previous autonomous robots. This system almost has human language processing ability.

In robotics, the importance of physical moves for their intelligence has been ignored, but new studies that focus on this aspect are now being launched.

Surpassing the PC Market

Constant access to the Internet is also a challenge for robots to make great progress. The intelligence of current robots is limited by their battery capacity or memory size. If constant net connections are made possible, then robots could acquire unlimited amounts of data.

In that case, it would be crucial to have computer systems in full control, because hacking could paralyze thousands of robots at once. Manufacturers might be required to take effective measures in this respect.

At any rate, there is a huge potential for robotics in the future. Technologies are available to make various kinds of robots such as those that could be used to clear land mines in combat zones. The market for robots will be expanding in size and eventually will surpass the PC market in the future. Robots have a better function than PCs as a device to connect human beings with the Internet.

(English translation of the original Japanese article that appeared in the May 7 issue of "Economist," published by Mainichi Shinbun)