Why Machines Learn: The Elegant Math Behind Modern AI

A science writer who has spent years covering physics gives AI the mathematical treatment it deserves: honest about the complexity, committed to making it genuinely understandable. This is the book for colleagues who want to understand why these systems work, not just that they do.

Anil Ananthaswamy's previous books made quantum mechanics and cosmology accessible to general readers without softening the ideas. He brings the same approach to machine learning here. Starting from the basics of gradient descent and linear regression and working through neural networks, backpropagation, and transformer architecture, he traces the mathematical logic that makes modern AI possible. The prose never condescends, and the explanations never blur into metaphor when precision matters.

This is not a how-to guide for building AI systems. It is something more durable: a conceptual map of why they behave the way they do. Assign it in a course on algorithms, data science, or the history of computation. Give it to a colleague in the humanities who is trying to form an informed opinion about AI before the next faculty committee meeting. The updated paperback edition includes a new afterword on transformer architecture and large language models.

Dutton, 2024. Paperback.