There is a LOT of AI literature out there. Amazon.com has over 7000 AI non-fiction books in their catalogue. I've read a bunch of them (and have a bunch more on my reading list) but there are five that stand out to me as mandatory reading. The books I've listed here are by leading computer scientists and thinkers and really expanded my understanding and even my worldview. The first two get into technical details which really require reading the text, but the last three are well suited to the Audiobook format if you don't have the time to curl up on the couch with a paperback.

The Book of Why - Judea Pearl¶

Judea Pearl is big name in computer science, having played a major role in the development of Bayesian networks, even receiving the Turing Award, the highest distinction in computer science, "for fundamental contributions to artificial intelligence through the development of a calculus for probabilistic and causal reasoning". In this book Pearl argues that understanding causality, and adopting his framework for causal modelling, will revolutionize the field of AI. The book is compelling, stepping up the 'ladder of causation' (an abstraction of his model) in systematic way which is easy enough for the mathematically minded general audience to follow. I was certainly convinced of his core thesis and I'm eager to try out his Causal Diagram approach the next time I'm wrangling with a prediction problem.

The Master Algorithm - Pedro Domingos¶

Pedro Domingo's book is the best overall summary of Machine Learning approaches I've read so far. He categorizes 'five tribes of Machine Learning': Symbolists, Connectionists, Evolutionaries, Bayesians and Analogizers, describing the history, approaches, use cases and benefits of each. He then explores the future of ML and outlines an approach to unifying the "tribes" in a single universal learner, or Master Algorithm, which may be the key to unlocking Artificial General Intelligence. For a crash course in ML concepts, or a new way of looking at old approaches, look no further than this great read.

Superintelligence - Nick Bostrom¶

AI safety, AI Alignment, Responsible AI; these increasingly important topics are now getting the attention they warrant, and this can be partly credited to Nick Bostrom's popular and compelling book Superintelligence. Written for a general, even non-technical audience, Bostrom makes a logical, at times philosophical argument for taking the threat of 'unaligned' AI seriously and concentrating on engineering human preferences and control into our machines from the outset. On reading the book it's hard not to buy into his thesis of the risks of AI (to a greater or lesser extent), but even if you don't this book is a fascinating exploration of the themes surrounding AI safety.

Prediction Machines, The Simple Economics of Artificial Intelligence - Ajay Agrawal, Avi Goldfarb, and Joshua Gans¶

Prediction Machine takes some of the theoretical collateral of AI and re-frames it in economic terms. How should businesses think about the costs and benefits of AI? The book guides the reader through an approach, distilling AI down to a prediction capability then discussing the implications of improvements in prediction on business decision making and strategy There's also a useful "AI Canvas" which businesses can use to understand the component parts of an AI solution and the implications for return on investment. This is book is perfect for business leaders who need to understand and make decisions influenced by AI, or for data scientists who would benefit from some bigger picture thinking.

Homo Deus: A Brief History of Tomorrow - Yuval Noah Harari¶

This is not really a AI or ML book, but Yuval Noah Harari's credentials make his perspective on the future of digital humanity uniquely compelling among futurists. A quote from this book also inspired the title of this blog. Harari understands better than the masses of writers looking into the AI crystal ball what the motivations and tendencies of human beings are throughout history, and if you could distill all of human history into a data set then I expect his thesis to be the best fit model for prediction. In this book Harari explores the essence of humanity, our current 'Dataist' religion, and projects these forward to our future as digitally empowered human gods. This may not impact your day job today, but it's about as 'big-picture' as the literature gets.