Book Review: How AI Works: From Sorcery to Science

Title: How AI Works: From Sorcery to Science

Author: Ronald T. Kneusel

Publisher: No Starch Press

Year: 2023

Introduction

"How AI Works: From Sorcery to Science" by Ronald T. Kneusel stands out as a refreshingly clear and engaging primer on artificial intelligence, aimed at breaking down barriers in a field that often seems shrouded in complexity and mystique. This is a book that stood out to me because of how eloquently it breaks down advanced topics like Convolutional Neural Networks(CNNs) while being highly engaging.

This review looks more at the content and the value that can be gleaned from this book. These are my opinions gathered through my raw notes in my first read-through and the second for writing this review. This review is written from the perspective of a professional software engineer researching AI.

Structure and Content

Groundwork

Kneusel begins by carefully laying the groundwork—defining the distinctions between artificial intelligence, machine learning, and deep learning. He methodically traces the lineage of AI from early systems and classical approaches, through the rise of neural networks, to today’s transformative technologies such as large language models. This historical context not only enriches the reader’s understanding but also highlights the incremental advances in AI research and application. The groundwork chapters do well in building off one another but still are easy to follow for those skimming for a specific heading as often tech professionals do.

The ‘Meat’

The groundwork section ends with chapter four leading into the ‘meat’ of the book: Convolutional Neural Networks(CNNs), Generative Adversarial Networks (GANs), and Large Language Models (LLMs). The ‘meat’ of the book spans chapters 5 and 6 being the most heavy portion of the book as far as new information is concerned. This section has more thoughtfully placed diagrams to help the reader digest the content.

Wrap-up

Ending on a strong note, Kneusel takes a critical look at our current AI system's limitations. As readers descend the informational summit that is the ‘meat’ of the book this section fills in the gaps and the real work implications with examples of prompts run on current—as of the writing of the book— LLMs and noting their responses. This introspective and commentary is relevant to the AI discussion and adds value to the book.

Strengths

Clarity

One of the book’s foremost strengths is its clarity. Kneusel eschews dense technical jargon and intimidating equations in favor of relatable analogies and a narrative style that makes complex concepts digestible. By presenting neural networks as collections of simple, repeating units working in concert, he invites readers to appreciate how emergent intelligence is built from elementary processes. This approach makes the subject approachable without sacrificing depth, the hallmark of a good introductory book. This strength can be a double-edged sword for some readers as is noted in the weaknesses.

Historical Context

History is often relegated to the dusty thick volumes in reference sections but Kneusel masterfully integrates a bite-sized dose early on. Typically backstory being added isn't a noteworthy strength but rather an expectation however with a topic as nuanced as AI this delivery is impressive.

Visuals

The illustration in the book is clear and fully sprinkled in to support the content of the text surrounding it. The diagrams are easy to understand even if the reader is fuzzy on the subject. The screenshots and pictures are clear even for a black and white print book.

Balanced Perspective

While celebrating the remarkable achievements of modern AI, Kneusel is careful not to overlook the limitations and challenges that persist. The book addresses the nuanced capabilities and occasional shortcomings of LLMs, offering readers a balanced view of AI’s potential and its boundaries. This section can be a bit over the head of some novices however is still valuable in context.

Weaknesses

Depending on the reader, the limited technical depth could be a weakness. This is an introductory book with the application of AI limited to what could be fit in the pages. The later chapters make use of OpenAI’s ChatGPT chatbot interface, something most people are likely to utilize. Kneusel makes his intent know that this book is more introductory without all the complex mathematics and equations as in his other books on AI. For those who are more advanced in the subject matter and desire a more technical read, I recommend Math for Deep Learning: What You Need to Know to Understand Neural Networks by Ronald T. Kneusel and Practical Deep Learning, 2nd Edition by Ronald T. Kneusel. The second edition of Practical Deep Learning has not been available to the public, if you cannot wait to get into AI and Deep learning the first edition will still cover everything you need to know.

Overall Assessment

"How AI Works: From Sorcery to Science" is an outstanding contribution to the popular science and AI genres. Ronald T. Kneusel has crafted a work that is as informative as it is engaging, effectively bridging the gap between complex technological ideas and everyday understanding. For professionals and lay readers alike, this book provides both a solid foundation and a thought-provoking look at the past, present, and future of artificial intelligence—a must-add to any tech enthusiast or IT professional library.

This book is recommended for:

Anyone curious about AI but intimidated by the technical jargon
Professionals in non-technical fields who want to understand the impact of AI on their industries
Students seeking an accessible introduction to artificial intelligence
General readers interested in understanding one of the most transformative technologies of our time, popular science readers
Decision-makers who need to make informed choices about AI adoption and policy

For those who are in a time crunch or just need that one bit of information to finish working on their project, I have provided two recommended sets of chapters to ‘speed-run’ the content. Although Kneusel’s work is brief enough as is, with an understanding that in the tech world time is the most valuable commodity. Below the sets are listed, if at any point you struggle to understand the content I recommend going to the previous chapter and then returning.

Students and IT professionals with a general knowledge of AI fundamentals looking to study AI theory : 3-6
Students and IT professionals with a general knowledge of AI fundamentals looking to study modern AI theory and AI application: 4-7