* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download How the Body Shapes the Way We Think
Survey
Document related concepts
Human-Computer Interaction Institute wikipedia , lookup
Ecological interface design wikipedia , lookup
Technological singularity wikipedia , lookup
Human–computer interaction wikipedia , lookup
History of artificial intelligence wikipedia , lookup
Existential risk from artificial general intelligence wikipedia , lookup
Intelligence explosion wikipedia , lookup
Philosophy of artificial intelligence wikipedia , lookup
Transcript
How the Body Shapes the Way We Think A New View of Intelligence Rolf Pfeifer and Josh Bongard with a contribution by Simon Grand Foreword by Rodney Brooks Illustrations by Shun Iwasawa A Bradford Book The MIT Press Cambridge, Massachusetts London, England © 2007 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. MIT Press books may be purchased at special quantity discounts for business or sales promotional use. For information, please email [email protected] or write to Special Sales Department, The MIT Press, 55 Hayward Street, Cambridge, MA 02142. This book was set in Syntax and Times Roman by SNP Best-set Typesetter Ltd., Hong Kong. Printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Pfeifer, Rolf, 1947– How the body shapes the way we think : a new view of intelligence / by Rolf Pfeifer and Josh Bongard ; with a contribution by Simon Grand ; foreword by Rodney Brooks ; illustrations by Shun Iwasawa. p. cm. Includes bibliographical references (p. ). ISBN-13: 978-0-262-16239-5 (alk. paper) ISBN-10: 0-262-16239-3 (alk. paper) 1. Artificial intelligence. 2. Cognitive science. I. Bongard, Josh. II. Grand, Simon. III. Title. Q335.P445 2006 006.3—dc22 2006044919 10 9 8 7 6 5 4 3 2 1 To my friends in Japan (R. P.) To Toby, Carol, and Ralph (J. B.) Contents Foreword by Rodney Brooks Preface xvii I 1 Intelligence, Artificial Intelligence, Embodiment, and What the Book Is About 1 Intelligence, Thinking, and Artificial Intelligence 1.1 1.2 1.3 1.4 1.5 1.6 2 xiii Artificial Intelligence: The Landscape 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 5 Thinking, Cognition, and Intelligence 7 The Mystery of Intelligence 11 Defining Intelligence 14 Artificial Intelligence 17 Embodiment and Its Implications 18 Summary 22 25 Successes of the Classical Approach 27 Problems of the Classical Approach 30 The Embodied Turn 34 The Role of Neuroscience 37 Diversification 39 Biorobotics 41 Developmental Robotics 44 Ubiquitous Computing and Interfacing Technology Artificial Life and Multiagent Systems 49 Evolutionary Robotics 53 Summary 54 47 viii II 3 Contents Toward a Theory of Intelligence Prerequisites for a Theory of Intelligence 3.1 3.2 3.3 3.4 3.5 3.6 3.7 4 57 61 Level of Generality and Form of Theory Diversity-Compliance 67 Frame of Reference 72 The Synthetic Methodology 77 Time Perspectives 82 Emergence 85 Summary 88 Intelligent Systems: Properties and Principles 62 89 4.1 4.2 4.3 Real Worlds and Virtual Worlds 90 Properties of Complete Agents 95 Agent Design Principle 1: The Three-Constituents Principle 100 4.4 Agent Design Principle 2: The Complete-Agent Principle 104 4.5 Agent Design Principle 3: Cheap Design 107 4.6 Agent Design Principle 4: Redundancy 113 4.7 Agent Design Principle 5: Sensory-Motor Coordination 117 4.8 Agent Design Principle 6: Ecological Balance 123 4.9 Agent Design Principle 7: Parallel, Loosely Coupled Processes 134 4.10 Agent Design Principle 8: Value 137 4.11 Summary and Conclusions 140 5 Development: From Locomotion to Cognition 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 141 Motivation 143 Toward Developmental Robot Design 145 From Locomotion to Cognition: A Case Study 149 From Gait Patterns to Body Image to Cognition 153 The Symbol Grounding Problem 159 Matching Brain and Body Dynamics 161 Broadening the Scope: Other Aspects of Development 164 Learning in Embodied Systems 168 Social Interaction 170 Development: Where Are We and Where Do We Go from Here? 173 Contents ix 5.11 Summary: Design Principles for Developmental Systems 175 6 Evolution: Cognition from Scratch 177 6.1 6.2 6.3 6.4 Motivation 181 The Basics of Evolutionary Computation 184 The Origins of Evolutionary Computation 187 Artificial Evolution in the Real World: On Pipes, Antennas, and Electronic Circuits 189 6.5 Evolutionary Robotics 192 6.6 Evolving Morphology and Control 194 6.7 Genetic Regulatory Networks and Developmental Plasticity 196 6.8 Self-Organization: The Powerful Ally of Mutation and Selection 204 6.9 Artificial Evolution: Where Are We and Where Do We Go from Here? 206 6.10 Summary: Design Principles for Evolutionary Systems 208 7 Collective Intelligence: Cognition from Interaction 213 7.1 7.2 7.3 7.4 7.5 7.6 7.7 Motivation 215 Agent-Based Modeling 217 Simulation versus Real Robots 221 Groups of Robots 222 A Note on Cooperation 226 Modular Robots 228 Scalability, Self-Assembly, Self-Repair, Homogeneity, and Heterogeneity 232 7.8 Self-Reproducing Machines 235 7.9 Collective Intelligence: Where Are We and Where Do We Go from Here? 238 7.10 Summary: Design Principles for Collective Systems 241 III 8 Applications and Case Studies 245 Ubiquitous Computing and Interfacing Technology 8.1 8.2 8.3 249 Ubiquitous Technology as Scaffolding 251 Ubiquitous Technology: Properties and Principles Interacting with Ubiquitous Technology 263 253 x Contents 8.4 8.5 9 Building Intelligent Companies 9.1 9.2 9.3 9.4 9.5 9.6 10 Cyborgs 264 Summary and Conclusions 270 271 Management and Entrepreneurship: Decision and Action under Uncertainty 272 Companies as Embodied Systems 274 A Synthetic Approach to Management 279 Design Principles for Building Intelligent Companies 282 Corroborating the Speculations 293 Summary and Conclusions 294 Where Is Human Memory? 295 10.1 10.2 10.3 10.4 Introduction 298 The Storehouse Metaphor and Its Problems 300 Concepts of Memory 302 The Frame-of-Reference Problem in Memory Research: Ashby’s Proposal 304 10.5 The Embodied View of Memory: Applying the Design Principles for Intelligent Systems 307 10.6 Implications for Memory Research: Summary and Speculations 318 11 Robotic Technology in Everyday Life 323 11.1 Introduction: Everyday Robots 324 11.2 Vacuum Cleaners: Roomba, Trilobite, and Similar Species 327 11.3 Entertainment Robots 328 11.4 Therapeutic, Medical, and Rescue Robots 333 11.5 Humanoid Companion Robots 335 11.6 Robots Capable of Social Communication 341 11.7 Robots Capable of Facial and Bodily Expression 344 11.8 A Theoretical Note 346 11.9 Summary and Conclusions 348 IV 12 Principles and Insights 351 How the Body Shapes the Way We Think 353 12.1 Steps Toward a Theory of Intelligence 12.2 Selected Highlights 358 354 Contents 12.3 Seeing Things Differently 12.4 Epilogue 370 Notes 373 References 375 Index 389 xi 367 Foreword The great revolutions in science come about when what was formerly thought to be true and unassailable is both assailed and shown to not be true after all. Sometimes the assaults are brutal and front on, and sometimes they are gentle over a long period of time, gradually creeping up on the soon to be discredited truth. This book is a gentle assault on some of the collateral tenets of modern rationalism; not an assault on rationalism itself, but an assault on many of the things that are commonly assumed by rationalists. Rolf Pfeifer and Josh Bongard question whether our nervous systems compute, whether they are separate control systems for our bodies, and even whether there can truly be disembodied reasoning. These three ideas are so ingrained in our computational metaphors that they usually go unquestioned— they make no sense within our normal frameworks of thinking in the fields of computer science and artificial intelligence, and even neuroscience. Beyond the mere technical these questions challenge the intellectual father of rationalism Rene Descartes and his “Je pense, donc je suis” (I think, therefore I am) from his Discourse on Method (written in French, not Latin, in 1637). While such questions can be seen as a challenge to the very underpinnings of the scientific world view, they really are not. Pfeifer and Bongard are not suggesting throwing out the scientific method and replacing it, as some might fear, with postmodern relativism. Rather they are assaulting certain metaphors that have perhaps gone haywire in their influence on how we approach the study of intelligence, the study of us. In modern times there have been two important and perhaps underestimated influences on our view of intelligence. 1. As Alan Turing described in his 1950 paper “Computing Machinery and Intelligence,” his earlier and today still dominant model of xiv Foreword computation came from considering the externally observable behavior of a human computer, a person who carried out computations with pen and paper, and “is supposed to be following fixed rules.” It is worth noting here that Turing modeled what a person does, not what a person thinks. 2. Ever since the human brain has come to be considered as the seat of our thought, desires, and dreams, it has been compared to the most advanced technology possessed by mankind. In my own lifetime I have seen popular “complexity” metaphors for the brain evolve. When I was a young child the brain was likened to an electromagnetic telephone switching network. Then it became an electronic digital computer. Then a massively parallel digital computer. And delightfully, in April 2002, someone in a lecture audience asked me whether the brain could be “just like the world wide web.” Even otherwise serious scientists have become enamored of their own complexity metaphors declaring for instance that quantum phenomena and the brain are both so complex that they must be about the same thing. Turing’s metaphor has become the very definition of computation, and he points out in his 1950 paper, using Babbage’s unrealized mechanical engine as the exemplar, that such computation is independent of the medium in which it is expressed. The metaphors for the brain (except for the quantum speculations) have entrenched it as the equivalent of Turing’s form of computation, and thus rationalism largely assumes that the human brain is a Turing machine, carrying out Turing computation, and controlling its periphery, the human body. But when we consider the evolutionary history of nervous systems we are faced with a dilemma not unlike one that is so often used to challenge evolution itself. How could evolution have incrementally produced the components of an eye—the lens, the pupil, the retina—when all are necessary, fully formed, to enable the other to carry out its function within the ensemble? When we turn that skepticism on its head we are left to ask what roles earlier versions of nervous systems played, before they became fully functional control systems, like Turing’s “control” component which he talked about along with the “executive” and the “store.” Metaphors are useful in science as a way of understanding systems we wouldn’t otherwise understand—metaphors can suggest appropriate questions to ask about a system, they can provide intuitive models about how things might work, and they can bridge gaps as a more explicit theory is being formed. But they can also lead to ways of thinking about