AI Safety, Alignment & Existential Risk
Human Compatible: Artificial Intelligence and the Problem of Control
Stuart Russell · 2019 · Viking
A founding AI researcher's case for redesigning how machines are built to pursue human objectives, rather than fixed goals that might be specified wrong.
Cover image: Open Library.
Summary
This book explores the potential for artificial intelligence to become superintelligent and the challenges this poses for humanity. The author, an AI researcher, outlines near-term benefits of AI and also identifies ways in which AI is already being misused. The authors argue that a conflict between humans and machines is not inevitable if AI is fundamentally rethought. They propose redesigning AI with a foundation where machines are uncertain about human preferences, aiming to create systems that are deferential and beneficial.
Summary based on the publisher's page ↗
Read it
This book isn't free to read online. You can borrow it free with a library card, or buy it from the publisher.
About the author
Stuart Russell
Stuart Russell is a professor of Computer Science at the University of California, Berkeley. He is also the co-author of a textbook on artificial intelligence and has advised the United Nations on arms control.
Source: Publisher page ↗
Full Who's Who profile →