A trade book that argues that predictive AI is snake oil: it cannot and will never work. Artificial Intelligence is an umbrella term for a set of loosely related technologies. For instance, ChatGPT has little in common with algorithms that banks use to evaluate loan applicants. Both of these are referred to as AI, but in all of the salient ways -- how they work, what they're used for and by whom, and how they fail -- they couldn't be more different. Understanding the fundamental differences between AI technologies is critical for a technologically literate public to evaluate how AI is being used all around us. In this book, Arvind Narayanan and Sayash Kapoor explain the major strains of AI in use today: generative AI, predictive AI, and AI for content moderation. They show readers how to differentiate between them and, importantly, make a cogent argument for which types of AI can work well and which can never work, because of their inherent limitations. AI in this latter category, the authors argue, is AI snake oil: it does not and cannot work. More precisely, generative AI is imperfect but can be used for good once we learn how to apply it appropriately, whereas predictive AI can never work -- in spite of the fact that it's being sold and marketed today in products -- because we have never been able to accurately predict human behavior.