Human Compatible, Artificial Intelligence and the Problem of Control by Russell, Stuart -- Read -- Imperial Library of Trantor

Index

Cover Also by Stuart Russell Title Page Copyright Dedication Contents Preface

Why This Book? Why Now? Overview of the Book

1. If We Succeed

How Did We Get Here? What Happens Next? What Went Wrong? Can We Fix It?

2. Intelligence in Humans and Machines

Intelligence Computers Intelligent Computers

3. How Might AI Progress in the Future?

The Near Future When Will Superintelligent AI Arrive? Conceptual Breakthroughs to Come Imagining a Superintelligent Machine The Limits of Superintelligence How Will AI Benefit Humans?

4. Misuses of AI

Surveillance, Persuasion, and Control Lethal Autonomous Weapons Eliminating Work as We Know It Usurping Other Human Roles

5. Overly Intelligent AI

The Gorilla Problem The King Midas Problem Fear and Greed: Instrumental Goals Intelligence Explosions

6. The Not-So-Great AI Debate

Denial Deflection Tribalism Can’t We Just . . . The Debate, Restarted

7. AI: A Different Approach

Principles for Beneficial Machines Reasons for Optimism Reasons for Caution

8. Provably Beneficial AI

Mathematical Guarantees Learning Preferences from Behavior Assistance Games Requests and Instructions Wireheading Recursive Self-Improvement

9. Complications: Us

Different Humans Many Humans Nice, Nasty, and Envious Humans Stupid, Emotional Humans Do Humans Really Have Preferences?

10. Problem Solved?

Beneficial Machines Governance of AI Misuse Enfeeblement and Human Autonomy

Appendix A: Searching for Solutions Appendix B: Knowledge and Logic Appendix C: Uncertainty and Probability Appendix D: Learning from Experience Acknowledgments Notes Image Credits Index About the Author