BOOK · [2801]

Human Compatible

Technology

Berkeley AI professor Stuart Russell argues that the standard model of AI—build a system that optimizes a fixed objective—is fundamentally broken, and proposes replacing it with machines that are uncertain about human preferences and defer to humans under that uncertainty. Rigorous and accessible, it bridges technical AI research and policy in a way that almost no other book does. Sits at the intersection of the AI Canon's technical deep dives and its market analysis of what safe, deployable AI actually requires.

AI Study Tools

Ways to absorb this book faster

—

Key points

Not available for this book

20 min read

Slides

Flip through the big ideas

Open →

25 min read

Chapter-by-chapter companion

A guided walk through every chapter

Open →

Endorsed By

4 People

Elon Musk

twitter.com

“Worth reading 'Human Compatible' by Stuart Russell about future AI risks.”

Page cites a tweet by Elon Musk.
a16z

a16z AI Canon

Russell's technical and philosophical case for preference-uncertain AI; the serious computer scientist's companion to the AI Canon's safety-adjacent papers.
Judea Pearl

people.eecs.berkeley.edu

“Human Compatible made me a convert to Russell's concerns with our ability to control our upcoming creation -- super-intelligent machines. Unlike outside alarmists and futurists, Russell is a leading authority on AI. His new book will educate the public about AI more than any book I can think of, and is a delightful and uplifting read.”

Jacket endorsement Pearl wrote for Russell's book.
Yoshua Bengio

people.eecs.berkeley.edu

“This beautifully written book addresses a fundamental challenge for humanity: increasingly intelligent machines that do what we ask but not what we really intend. Essential reading if you care about our future.”

Bengio's published cover endorsement for Russell's book.