Artificial Intelligence

Transcription

1

2

3 Artificial Intelligence I]

4

5 Artificial Intelligence A Philosophical Introduction Jack Copeland

6 Copyright B. 1. Copeland, 1993 The right of B. J. Copeland to be identified as author of this work has been asserted in accordance with the Copyright, Designs and Patents Act First published 1993 Reprinted 1994, 1995, 1997 (twice), 1998 Blackwell Publishers Ltd 108 Cowley Road Oxford OX4 IJF, UK Blackwell Publishers Inc 350 Main Street Malden, Massachusetts 02148, USA All rights reserved. Except for the quotation of short passages for the purposes of criticism and review, no part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. Except in the United States of America, this book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated without the publisher's prior consent in any form of binding or cover other than that in which it is published and without a similar condition including this condition being imposed on the subsequent purchaser. British Library Cataloguing in Publication Data A CIP catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Copeland, Brian Jack Artificial intelligence: a philosophical introduction / B. 1. Copeland p. cm. Includes bibliographical references and index. ISBN X (pbk: alk. paper) I. Artificial intelligence - Philosophy. I. Title. Q335.C dc20 CIP Typeset in 10.5 on 12pt Ehrhardt by Graphicraft Typesetters Limited, Hong Kong Printed and bound in Great Britain by Athenreum Press Ltd, Gateshead, Tyne & Wear This book is printed on acid-free paper

7 For Jean and Reg

8

9 Contents List of figures x Acknowledgements Xl Introduction 1 In outline 2 1 The beginnings of Artificial Intelligence: a historical sketch The arrival of the computer The Logic Theorist The Dartmouth Conference Alan Turing and the philosophy of AI 9 2 Some dazzling exhibits Inside the machine Parry the paranoid program Eliza the psychotherapist Shrdlu the robot Hacker the program-writing program Programs that play games The General Problem Solver Sam and the Frump Expert systems 30 3 Can a machine think? Is consciousness necessary for thought? The Turing Test Has the Test been passed already? Four objections to the Turing Test Assessing the Test Decision time S2

10 viii Contents 4 The symbol system hypothesis Symbol manipulation Binary symbols Programs as symbols A full-scale program The definition of a computer The hypothesis Multiple realizability 81 5 A hard look at the facts The evidence for the hypothesis Getting the evidence into perspective Hype Programming common sense Data versus know-how The CYC project The complexity barrier The curious case of the Chinese room The Chinese room argument What's wrong with the argument? Deciding about understanding Turing machines and the biological objection to AI Freedom Turbo Sam makes a choice Is freedom of the will an illusion? Two kinds of freedom Kleptomania and other compulsions Libertarianism Predictivism and chaos The inevitable Consciousness Neglect and disa"ay The fuzzy baseline Consciousness as a type of internal monitoring The ineffable 'feel' of it all Into the heart of the mystery What is it like to be a bat? What Mary doesn't know Drawing the threads together 179

11 Contents ix 9 Are we computers? The strong symbol system hypothesis Hardware versus wetware Goodbye, von Neumann Putting meaning into meat Believing what you don't believe Productivity and systematicity Evaluating the arguments The meaning of 'computer' AI's fresh start: parallel distributed processing The basic ideas English lessons Escape from a nightmare? The contrast with computers Comparisons with wetware Searle's Chinese gym The Church-Turing thesis Are our cognitive processes algorithmically calculable? Simulating networks by computer The battle for the brain Concluding remarks 247 Epilogue 249 Notes 250 Bibliography 283 Index 299

12 List of figures Figure 2.1 Shrdlu's world of coloured blocks. Figure 2.2 Simple block-moving tasks. Figure 2.3 The tower of contrition. Figure 3.1 Can you read this? Figure 3.2 The Turing Test. Figure 4.1 A register. Figure 4.2 Recursion. Figure 4.3 Think of each binary digit as pointing to a box. Figure 4.4 Each box contains a base 10 number. Figure 4.5 How the compiled program is stored in memory. Figure 5.1 Shadows are a problem for visual recognition software. Figure 6.1 } T ' h" Figure 6.2 unng mac me mput. Figure 6.3 } T' h' F 6 4 unng mac me output. 19ure. Figure 8.1 Each hemisphere 'sees' only one half of the screen. Figure 8.2 The two hemispheres pick different objects when the split-brain subject is asked which pictured items he associates with what he can see on the screen. Figure 9.1 Human neuron. Figure 10.1 The neuron fires a discharge along its output fibres when the weighted sum of its inputs exceeds its threshold. Figure 10.2 Connections between the layers of artificial neurons in a PDP network. Figure 10.3 Inverting an input pattern. Figure 10.4 A few of the variations on the letter 'A' to be found in the Letraset catalogue. Figure 10.5 The curse of context. Figure 10.6 The level of output depends on the amount of input the unit is receiving from other units

13 Acknowledgements Many people have helped me with this book in many different ways, large and small. I thank you all. David Anderson, John Andreae, Derek Browne, Peter Carruthers, Stephan Chambers, John Cottingham, Hubert Dreyfus, Flip Ketch, Carmel Kokshoorn, Justin Leiber, David Lewis, Michael Lipton, Bill Lycan, Michael Maclean, Pamela McCorduck,James McGahey, Jack Messenger, Alison Mudditt, Gill Rhodes, Michael Smallman, Steve Smith, Kerry Stewart, Bob Stoothoff, Stephen Viles. Particular thanks to Jenny Arkell for commenting extensively on early drafts and suggesting many improvements; to Ann Witbrock for the computer-generated illustrations; and above all to Diane Proudfoot for support, criticism, help and encouragement. Figure 2.1 is based on figure 3 of Terry Winograd, Understanding Natural Language, with kind permission of Academic Press and Terry Winograd. Figure 3.1 is redrawn from figure 15 of Douglas Hofstadter, Codel, Escher, Bach: An Eternal Colden Braid, with kind permission of Basic Books Inc. (copyright 1979 by Basic Books, Inc.). Figure 5.1 is redrawn from figures 2.1 and 2.2 of P.H. Winston (ed.), The Psychology of Computer Vision, with kind permission of McGraw-Hill, Inc. Figure 8.2 is redrawn from figure 42 of M.S. Gazzaniga and ].E. LeDoux, The Integrated Mind, with kind permission of Plenum and Michael Gazzaniga. Figure 10.4 is a reproduction of parts of the Letraset catalogue and appears by the kind permission of Esselte Letraset Ltd.

14

15 Introduction Artificial Intelligence is the science of making machines do things that would require intelligence if done by men. Marvin Minsky, founder of the MIT Artificial Intelligence Laboratoryl Not long ago I watched a TV interview with Edward Fredkin, a specialist in electronic engineering and manager of the MIT Artificial Intelligence Laboratory. Fredkin is an earnest man with a serious, authoritative manner. What he had to say was startling. 2 There are three great events in history. One, the creation of the universe. Two, the appearance of life. The third one, which I think is equal in importance, is the appearance of artificial intelligence. This is a form of life that is very different, and that has possibilities for intellectual achievement that are hard for us to imagine. These machines will evolve: some intelligent computers will design others, and they'll get to be smarter and smarter. The question is, where will that leave us? It is fairly difficult to imagine how you can have a machine that's millions of times smarter than the smartest person and yet is really our slave, doing what we want. They may condescend to talk to us, they may play games that we like to play, and in some sense, they might keep us as pets. 3 Has Professor Fredkin missed his vocation as a science fiction writer? Or is this a realistic view of the future, a sober prediction from a man who is better placed than most to understand the implications of current developments in AI research? Are computers that think really a technological possibility? Indeed, does it even make sense to speak of a machine thinking - or is a thinking computer a conceptual absurdity, like a grinning electron or a married bachelor? For the philosophically curious, Fredkin's words raise a tumult of intriguing questions. Is thought a biological phenomenon, and thus as far beyond the compass of a silicon-and-metal machine as photosynthesis, lactation, and every other biology-dependent process? Or are thinking and perceiving more like flying - something that both living creatures and metallic artefacts can do?