Superintelligence : paths, dangers, strategies / Nick Bostrom.

Format:

Book

Author/Creator:

Bostrom, Nick, 1973- author.

Language:

English

Subjects (All):

Artificial intelligence--Philosophy.

Artificial intelligence.

Cognitive science.

Physical Description:

xvi, 328 pages ; 25 cm

Edition:

First edition.

Place of Publication:

Oxford : Oxford University Press, 2014.

Summary:

The human brain has some capabilities that the brains of other animals lack. It is to these distinctive capabilities that our species owes its dominant position. Other animals have stronger muscles or sharper claws, but we have cleverer brains. It machine brains one day come to surpass human brains in general intelligence, then this new superintelligence could become very powerful. As the fate of the gorillas now depends more on us humans than on the gorillas themselves, so the fate of our species then would come to depend on the actions of the machine superintelligence. But we have one advantage: we get to make the first move. Will it be possible to construct a seed AI or otherwise to engineer initial conditions so as to make an intelligence explosion survivable? How could one achieve a controlled detonation? This profoundly ambitious and original book picks its way carefully through a vast tract of forbiddingly difficult intellectual terrain. Yet the writing is so lucid that it somehow makes it all seem easy. After an utterly engrossing journey that takes us to the frontiers of thinking about the human condition and the future of intelligent life, we find in Nick Bostroms work nothing less than a reconceptualization of the essential task of our time. Book jacket.

Contents:

1 Past developments and present capabilities 1

Growth modes and big history 1

Great expectations 3

Seasons of hope and despair 5

State of the art 11

Opinions about the future of machine intelligence 18

2 Paths to superintelligence 22

Artificial intelligence 23

Whole brain emulation 30

Biological cognition 36

Brain-computer interfaces 44

Networks and organizations 48

Summary 50

3 Forms of superintelligence 52

Speed superintelligence 53

Collective superintelligence 54

Quality superintelligence 56

Direct and indirect reach 58

Sources of advantage for digital intelligence 59

4 The kinetics of an intelligence explosion 62

Timing and speed of the takeoff 62

Recalcitrance 66

Non-machine intelligence paths 66

Emulation and AI paths 68

Optimization power and explosivity 73

5 Decisive strategic advantage 78

Will the frontrunner get a decisive strategic advantage? 79

How large will the successful project be? 83

Monitoring 84

International collaboration 86

From decisive strategic advantage to singleton 87

6 Cognitive superpowers 91

Functionalities and superpowers 92

An AI takeover scenario 95

Power over nature and agents 99

7 The superintendent will 105

The relation between intelligence and motivation 105

Instrumental convergence 109

Self-preservation 109

Goal-content integrity 109

Cognitive enhancement 111

Technological perfection 112

Resource acquisition 113

8 Is the default outcome doom? 115

Existential catastrophe as the default outcome of an intelligence explosion? 115

The treacherous turn 116

Malignant failure modes 119

Perverse instantiation 120

Infra structure profusion 122

Mind crime 125

9 The control problem 127

Two agency problems 127

Capability control methods 129

Boxing methods 129

Incentive methods 131

Stunting 135

Tripwires 137

Motivation selection methods 138

Direct specification 139

Domesticity 140

Indirect normativity 141

Augmentation 142

Synopsis 143

10 Oracles, genies, sovereigns, tools 145

Oracles 145

Genies and sovereigns 148

Tool-AIs 151

Comparison 155

11 Multipolar scenarios 159

Of horses and men 160

Wages and unemployment 160

Capital and welfare 161

The Malthusian principle in a historical perspective 163

Population growth and investment 164

Life in an algorithmic economy 166

Voluntary slavery, casual death 167

Would maximally efficient work be fun? 169

Unconscious outsourcers? 172

Evolution is not necessarily up 173

Post-transition formation of a singleton? 176

A second transition 177

Superorganisms and scale economies 178

Unification by treaty 180

12 Acquiring values 185

The value-loading problem 185

Evolutionary selection 187

Reinforcement learning 188

Associative value accretion 189

Motivational scaffolding 191

Value learning 192

Emulation modulation 201

Institution design 202

Synopsis 207

13 Choosing the criteria for choosing 209

The need for indirect normativity 209

Coherent extrapolated volition 211

Some explications 212

Rationales for CEV 213

Further remarks 216

Morality models 217

Do What I Mean 220

Component list 221

Goal content 222

Decision theory 223

Epistemology 224

Ratification 225

Getting close enough 227

14 The strategic picture 228

Science and technology strategy 228

Differential technological development 229

Prefer red order of arrival 230

Rates of change and cognitive enhancement 233

Technology couplings 236

Second-guessing 238

Pathways and enablers 240

Effects of hardware progress 240

Should whole brain emulation research be promoted? 242

The person-affecting perspective favors speed 245

Collaboration 246

The race dynamic and its perils 246

On the benefits of collaboration 249

Working together 253

15 Crunch time 255

Philosophy with a deadline 255

What is to be done? 256

Seeking the strategic light 257

Building good capacity 258

Particular measures 258

Will the best in human nature please stand up 259.

Notes:

Includes bibliographical references and index.

ISBN:

9780199678112

0199678111

OCLC:

881706835