Artificial Intelligence: A Guide for Thinking Humans

A great overview of the current state of AI - relatively accessible even if you're not in computer science, but still interesting enough if you are. It helps you understand the intuition behind the main techniques currently used in vision, gameplay, language processing (convolutional nets, reinforcement learning etc). Even more importantly, it explains the limitations (through adversarial examples) and the way results are measured in all these fields so you are able to read the media hype more critically.

The author being a student of Douglas Hofstadter, she goes into philosophical parts a bit, but kinda avoids the hard question about AGI. It's understandable to me when non-engineers make fun of the topic (and it's true that AGI has been “just 10 years away” since the 50s). But I imagine a practitioner like the author should have a firmer opinion: it's either on the horizon (meaning visible next steps and, dare I say, some kind of expected timeline) or it's completely impossible to say, because we're still not sure there's not some immortal soul in humans or whatnot. Saying “my intuition is it's going to take a really long time” is weaseling out a bit.

Nevertheless, a great read if you want to get up to speed with current AI.