Human Compatible

Artificial Intelligence and the Problem of Control

The book investigates the potential risks and long term challenges associated with developing artificial intelligence. The author argues that current approaches to machine intelligence create dangerous conflicts between human survival and machine objectives. A new foundational framework is proposed to design artificial intelligence that remains provably beneficial and deferential to human values. The discussion covers the theoretical evolution of machine capabilities and practical strategies for ensuring future technological safety.

by Stuart J. Russell

Viking · First published 2019 · 336 pages