Artificial Intelligence: A Guide for Thinking Humans by Melanie Mitchell

Melanie Mitchell has worked in AI research since the 1980s and wrote this because the public conversation had detached entirely from what the systems do. She explains the technology and then asks how far it actually is from understanding anything.

Neural networks and how they learn, the ImageNet moment, why deep learning works better than anyone expected, adversarial examples that fool a classifier with a few pixels, and the barrier of common sense — the everyday physical and social knowledge that humans acquire without being taught and that no system has yet reproduced.

One of the clearest and most level-headed books on AI for general readers, from an actual resear