Books and Arts

Nature 447, 381-382 (24 May 2007) | doi:10.1038/447381a; Published online 23 May 2007

Intelligence in a changing world

Hiroaki Kitano1

BOOK REVIEWEDHow the Body Shapes the Way We Think: A New View of Intelligence

by Rolf Pfeifer & & Josh Bongard


Bradford Books: 2006. 394 pp. $39.95, £25.95

The study of intelligence, once dominated by biologists, has for decades been a focus for computer scientists. The question of whether a machine can be intelligent is as old as computers themselves, but was thrust into the spotlight ten years ago when IBM computer Deep Blue beat world champion Garry Kasparov at chess. Whether it really showed intelligence as we know it is still a matter for debate, but researchers have identified that 'intelligence' in the context of chess depends on having a huge database, computing power to search for moves and the ability to learn from past games to obtain a 'goodness score' for each possible move.

Intelligence in a changing world

MESSE BREMEN

Playing ball: robots must collaborate if they are to succeed at team sports.

However, there are clear differences between the way humans and computers play chess: a chess computer, unlike a human, does not have a body to enable it to interact with its environment, for example. This distinction differentiates two views on intelligence. One view is that intelligence is independent of the body and is unaffected by its existence, shape and function. The other view is that intelligence is contained within a physical body and that the body shapes the mind, an idea often referred to as physical embodiment or the presence of a behaviour-based agent. There is increasing recognition in the artificial-intelligence and robotics communities that the nature of the body significantly affects the mind, although it does not totally control it.

How The Body Shapes The Way We Think by Rolf Pfeifer and Josh Bongard provides an excellent perspective on how artificial-intelligence and robotics researchers have been tackling this issue. It is full of examples and thought-provoking discussions so that readers can easily follow some of the central debates on intelligence developed over decades. It also presents a chronological development of the field where appropriate.

The major focus of this book is to discover the design principle of an intelligent agent that has a physical and mobile body, has a high degree of autonomy, interacts with its environment and exhibits a broader range of behaviours than those single-task chess computers. It is not a book about how the body of an existing life form shapes its own mind, so there are only limited references to biology and neuroscience. Nevertheless, there are several parallels between artificial systems and biological systems. In one of the design principles, the authors point out the importance of redundancy, which also applies to biological systems. Some of these commonalities between artificial and biological systems can be seen as system-level principles that seem fundamental to a system's ability to exhibit intelligence, at least to an observer's eye.

One salient difference between the intelligent agents discussed in this book and traditional artificial-intelligence systems, as represented by chess computers, is the contextual thickness of system behaviours. Many of the robotics systems discussed in the book can cope, at least to some extent, with changes in the expected environment, tasks and other assumed conditions, whereas chess computers and other traditional artificial-intelligence systems are usually extremely fragile when faced with even a small change in such conditions. Behaviour-based robots should be able to perform almost flawlessly if the size of road or unevenness of terrain deviates from the initial assumption. However, the results will be catastrophic if a chess computer is given a chess board with nine rows and columns, rather than eight, as they are tuned specifically for the existing rules of chess. Imagine a thought experiment on a chess game between a behaviour-based system and an existing chess computer. The chess computer would be unbeatable with the defined rules, but if the rules were modified the behaviour-based system may do better.

The authors discuss learning, development and evolution, an approach inspired by, but not identical to, the biological one. These are methods designed to cope with complex, open and dynamic environments in which predefined rules cannot properly define behaviours because of the overwhelming complexity and difficulty in describing the situation. So artificial agents have to evolve, undergo development and learn how to behave. The authors also discuss the social aspect of multiple agents, or collective intelligence. How do agents with different expertise collaboratively accomplish complex tasks of the sort seen in real society or in sports such as football? These features are considered to be critically important for a multi-agent robotics team to play football in the RoboCup (http://www.robocup.org). The book beautifully illustrates the development of ideas, why we need these ideas, and what the issues are.

The book focuses on artificial agents, but with a lot of inspiration from nature. This reflects the synthetic approach to understanding: by building intelligent systems we can reach a deeper understanding of intelligence in general. Biologists often criticize such approaches as not being faithful to the biological processes that inspired them. However, researchers in artificial intelligence and robotics are, for example, trying to find out how to design an aircraft inspired by bird flight, instead of replicating a bird itself. It was a huge step in engineering when bird flight was decomposed into thrust and lift, and re-implemented by fixed wings and engines to create modern aircraft. Memory, computing and learning have been similar elements of intelligence in chess machines. The grand question behind this book is the search for such essential ingredients for intelligence in an open, dynamic environment.

  1. Hiroaki Kitano is director of Sony Computer Science Laboratories, 3-14-13 Higashi-Gotanda, Shinagwa, Tokyo 141-0022, Japan.


Readers' Comments

If you find something abusive or inappropriate or which does not otherwise comply with our Terms and Conditions or Community Guidelines, please select the relevant 'Report this comment' link.

There are currently no comments.

Add your own comment

This is a public forum. Please keep to our Community Guidelines. You can be controversial, but please don't get personal or offensive and do keep it brief. Remember our threads are for feedback and discussion - not for publishing papers, press releases or advertisements.

You need to be registered with Nature and agree to our Community Guidelines to leave a comment. Please log in or register as a new user. You will be re-directed back to this page.