This is a fantastic, well-researched and well-thought-through book. I have a lot of fun reading it. The basic premise of the book is saying that we need to start paying attention to how Artificial Intelligence is trained and used in our lives, the problem of control.

In the book, Stuart Russell discusses many areas involving moral philosophy, arguments for and against Artificial Intelligence, how humans behave and how reflective the behavior is to infer human's preference, etc.

In the book, he discusses a new way to build Artificial Intelligence so that we can get them to work with us (notice the word is "with" not "for"). The three rules are

1) Altruistic AI - Helping humans to build the lives that humans prefer, the values that they hold.

2) Humble AI - The AI does not know the values, which is not well-defined in humans too.

3) Empathy AI - The AI observes the behavior of humans to infer the values and also get feedback from humans where possible to improve it.

Although, it will be great if there are more diagrams for some of his explanations but overall it broadened my horizon on artificial intelligence. :)

Strongly recommended read and if you want a VERY condensed version of it, check out the TED talk he's given.