The Alignment Problem - By Brian Christian
Although written to be accessible for a non-technical audience, this book provides an excellent tour through the history of technical breakthroughs in mathematics, computer science, and cognitive science that brought us to the current moment in AI/ML. This history serves as a foundation to explain why the critical unsolved problem for AI/ML is the “alignment problem,” and why solving this problem is so difficult. For those unfamiliar with the AI safety field and willing to invest the time, this book provides a persuasive explanation for why creating a powerful and yet controllable and safe AI is such a risky undertaking. For those already familiar with the field, the main value here is in tracing the history of the field from its early days.
if anyone builds it, everyone dies - by eliezer Yudkowsky & Nate Soares
Not for the faint of heart, this is as stark and unequivocal a warning as you will find about the existential dangers of AI. Although excellent for clarifying the mind on the significance of the threats humanity faces, I left this feeling unsatisfied at the proposed solutions.