I'm thrilled to announce a massive update to my book, Mostly Harmless AI. What started as a collection of essays exploring various facets of artificial intelligence has been completely reimagined and restructured into a comprehensive, integrated learning path. This new version, with over 180 pages of principled advice, is designed to guide you systematically through the world of AI, from its foundational principles to its real-world applications and the critical challenges we face.
To celebrate this major update and to thank you for being a part of this journey, I’m offering a special discount. For a limited time, you can get your copy of Mostly Harmless AI for 50% off. This isn't just for the current draft; your purchase will grant you access to all future updates, new chapters, and revisions in perpetuity. It's a living book, and I want you to be a part of its evolution.
The most significant change in this new edition of Mostly Harmless AI is its structure. I’ve moved away from the original collection of disparate essays and have, instead, created a cohesive journey for the reader. The book is now organized into three distinct parts, each building upon the last:
Part I: Foundations of AI: We start with the fundamentals, demystifying what AI is and how core technologies like large language models work. This section covers everything from the history of AI to the practical techniques for deploying these powerful models.
Part II: Applications of AI: Here, we shift from theory to practice. You’ll explore how AI is transforming various fields, serving as a tool for critical thinkers, knowledge workers, software developers, educators, and creative professionals.
Part III: Dangers of AI: The final part of the book confronts the significant challenges and risks associated with AI. We’ll delve into the alignment problem, the inherent limitations of language models, and the real-world dangers of this technology, from autonomous weaponry to societal biases.
What's Inside?
This new structure is designed to provide a thorough understanding of AI's landscape. Here’s a glimpse of what you'll find in each chapter:
A Brief History of Artificial Intelligence: We trace the evolution of AI through its key eras, from its philosophical origins to the current generative age.
Chapter 1: Foundations of Artificial Intelligence: This chapter demystifies the core concepts of the field, including search, optimization, and machine learning.
Chapter 2: Generative AI and Large Language Models: We dive into the mechanics of LLMs, explaining how they learn to understand and generate human-like text.
Chapter 3: Techniques for LLM Deployment: Learn the essential strategies for using LLMs effectively, including prompt engineering, context augmentation, and agentic AI.
Chapter 4: AI for Critical Thinkers: A foundational guide to interacting with AI as a "cognitive partner" in your daily life.
Chapter 5: AI for Knowledge Workers: A structured workflow for professionals to leverage AI as an "exocortex" for deep research and analysis.
Chapter 6: AI for Software Development: Discover how AI is transforming the coding landscape, from code generation to agentic development workflows.
Chapter 7: AI for Educators and Learners: We examine the role of AI in education, proposing ethical and effective strategies for both teachers and students.
Chapter 10: Why Artificial Intelligence is Hard: An exploration of the fundamental challenges in AI, including the difficulties of machine learning and the formidable Alignment Problem.
Chapter 11: The Limits of Language Models: A close look at the inherent weaknesses of LLMs, such as hallucinations, biases, and the illusion of true reasoning.
Chapter 12: The Actual Risks of AI: We move beyond hypotheticals to discuss the pressing dangers of AI, from autonomous weaponry to societal disruptions.
Epilogue: The Road Ahead: A look to the future, discussing the path to General-Purpose AI and our collective role in shaping it.
The Current Status
I’m excited to share that the majority of the book is now in a very solid state. Most chapters are not just in their first draft, but have already undergone several rounds of editing. However, in the spirit of transparency that I advocate for throughout the book, I want to be clear that it is still a work in progress.
Specifically, Chapter 8: AI for Creative Professionals and Chapter 9: AI for Policy-Makers are still unfinished, and some sections in Part I still need to be fully fleshed-out. I'm actively working on all these sections and hope to have them ready for you by the end of this summer!
The rest of the book, however, is ready for you to dive into.
A Special Offer for Early Readers
As a reminder, to celebrate this major update, you can get your copy of Mostly Harmless AI for 50% off. Your purchase includes not only the current draft but also all future updates, new chapters, and revisions in perpetuity.
My goal with Mostly Harmless AI has always been to provide a clear-eyed, pragmatic perspective on this transformative technology. This new, structured version is a step closer that vision. I hope you find it as fulfilling to read as I did in writing it.
Congratulations—I know how big of a milestone that is and am looking forward to the book!