A personal bookshelf. Books that have shaped my thinking on AI ethics, technology, philosophy, and the spaces between.
This book is a dialogue, like most of Plato’s works. But there’s a point in it where his conjured version of Socrates uses the road to Larissa to explain the difference between knowledge and a belief that happens to be true. His surprising point is that either can guide you successfully to your destination. The discussion is a major focal point in episode 15 of my podcast, Machines & Meaning, that looks at how AI’s easy access to answers affects our ability to obtain knowledge.
Just started reading this book after a recommendation from a client. It seems to repackage a lot of familiar themes from pyschology and philsoophy. For example, becoming aware and overcoming cognitive bias, or how do we know what we know—epistemology. What’s perhaps unique is how these skills are being packaged/repackaged as potential meta-skills, the skills that will come to differentiate people in the future.
I found the book somewhat difficult to read. Not because it’s poorly written, but because almost every sentence seemed so profound that I found myself re-reading passages several times to make sure I understood the points. To Rawls’s credit, he does instruct the reader to focus on a few key sections to understand his core argument.
Crawford maps the material infrastructure of AI — from lithium mines to data centers to the labor of content moderators. Changed how I think about “the cloud” as a physical, political, and ecological reality.
One of the most important books on AI in the last few years, I couldn’t stop reading it. Not only does it serve as a historical portrait of the rise of Sam Altman and the formation of Open AI, it explains, in detail, how modern AI systems came to be and what is required to maintain them. Thus, pulling back the curtain back on topics such as the suffering that arises from farming out human labeling to foreign countries, the toxic content purposefully pumped into training data, and blatant disregard for human safety in an effort to chase the holy grail of AGI.
Russell’s reframing of the AI alignment problem through the lens of uncertainty about human preferences. The technical arguments are clear enough for non-specialists, and the philosophical implications are profound.
Arguably one of, if not the, most seminal works in philosophy. Aristotle distinguishes two broad types of virtues in the Nicomachean Ethics: intellectual virtues and moral virtues.
The book is also the foundation for Virtue Ethics, considered one of the “Big Three” Normative ethical approaches in Western culture along with Consequentialism and Deontology.
Benjamin’s concept of the “New Jim Code” — technology that reinforces racial hierarchies under the guise of neutrality — fundamentally shaped how I approach questions of bias in AI systems.
I have a soft spot for this book, as it’s written by the director of my AI Ethics program at Edinburgh. Shannon is a juggernaut of a thinker and helped me codify much of my approach to the role of the AI Practitioner.
Shannon’s follow-up book is unsuprisingly fantastic. While the topics are sophisticated, her writing is incredibly approachable. But it’s her real-life observations that really stand out, as she’s able to articulate what only registers subconsciously for many of us. Example: recommendation systems don’t merely learn our preferences—they can quietly start defining them for us.
The book had a strong influence on how I approach my role as an AI ethicist, making me realize that my chief value lies in ensuring that the technology we build aligns with human values. Then, of course, you have to deal with the invariable question of whose values?
A well-articulated view of how AI would really take over. Not through cyborg-like killing machines, but through AI agents infiltrating the internet.
O’Neil’s argument that algorithms encode bias at scale remains essential reading. The case studies on recidivism scoring and teacher evaluations are devastating. A foundational text for anyone working in AI ethics.
Eubanks documents how automated systems in welfare, housing, and public services disproportionately harm the poor. On my list alongside O’Neil and Benjamin.
A follow-up to Prediction Machines, examining how cheap prediction reshapes decision-making, power structures, and institutions. Recommended for the economic governance angle.
Suleyman’s argument about the convergence of AI and synthetic biology. On my list for the governance perspective.
A technical but accessible account of how fairness, privacy, and other ethical concepts can be formalized in algorithmic terms. The authors are honest about the limitations.