Human Compatible

What sets this book apart for us is its focus on rethinking the very foundations of AI. The author, Russell, thoughtfully examines the concept of intelligence itself, comparing humans and machines and outlining the necessary milestones for reaching superhuman AI. 

He also doesn't shy away from the current dangers of AI misuse, providing some really great examples.

We found the book very hard to put down, and it raised so many ‘What if?” questions for us!

As ever more powerful AI-based tools are created, Russell asks the question (and provides some answers) as to how we can ensure that we stay in the control of our creations. In particular, what safeguards are needed to protect us from something that will potentially be more intelligent than ourselves? Some might argue that this is all just science fiction and, even if it’s possible to build machines that are more intelligent than we are, it’s a problem for the distant future. However, there are many areas where AI is already making the key decisions about how we are treated.… show more.

This book lays out the security risks of fast-moving AI research better than any I have read. Stuart Russell believes artificial intelligence will soon out-think human beings in scale, scope, and speed. And he knows what he’s talking about: he’s been a leading AI researcher since the 1980s. Russell’s not concerned about superhuman robots becoming conscious and turning on us. His worry is intelligent AI algorithms doing exactly what we’ve designed them to do, including objectives that incentivize manipulating or harming humans. And that’s already happening–think social media search algorithms, driving humans into violent, polarized groups or future AI used… show more.