The Alignment Problem
by Brian Christian
How machine learning systems acquire human values, and fail to: bias in training data, reward hacking, inverse reinforcement learning, and the researchers trying to fix it. The best general-reader book on AI safety.
Read it if…
you want to understand AI alignment without the hype.
See the book →
Skip it if…
you want a business guide to AI.
Try instead: Yes or No →
Three things you will actually use
- Systems optimise what you measure, not what you mean.
- Bias in, bias out.
- Specify values carefully; it is harder than it looks.
Do this first
Read the chapters on reward hacking.
Our verdict
AI and human values.
See the book →
Compare with other Focus books
The Champion's Mind
Jim Afremow · Focus
Comprehensive and practical; the mental-skills reference.
CLICK TO SEEBrowse all Focus book reviews or get the free Reality Code Cheat Sheet.
As an Amazon Associate, WisdomDeck earns from qualifying purchases. Links marked “Click to see” go to Amazon. We are not paid for opinions, prices and availability are not shown here and change on Amazon, and we never see your purchase details. Nothing here is medical, legal or financial advice.