Reading List

Natural General Intelligence by Christopher Summerfield is probably the best book on Artificial Intelligence we have ever read. This truly information-dense volume of about 300 pages captures the history of AI research, its deep connections with neuroscience, the enormous progress of deep learning, and the key questions that remain unanswered. Each page is filled with information and no fluff.

It's a book worth reading and re-reading, as its insights build upon each other in profound ways. Destined to become a classic in the field!

Human Compatible stands out as a level-headed masterpiece in the AI landscape, steering clear of hype or doom-mongering while delivering well-reasoned perspectives on the potential dangers of advanced AI. Stuart Russell explores the profound risks arising from misaligned AI systems and proposes innovative frameworks for building next-generation AI that prioritizes human values and safety. The book delves into real-world examples, from algorithmic biases to self-driving cars, illustrating how current AI challenges stem from fundamental design flaws.

A Brief History of Intelligence offers a captivating journey through the evolution of brains and cognition, tracing intelligence from lifeless matter to complex human minds via five key breakthroughs. Max Bennett masterfully unique narrative connecting biological history with modern AI. Highlights how evolutionary innovations—like the first brains in ancient worms—inform today's quest for artificial general intelligence. The narrative is rich with scientific details yet accessible, making it a bridge between neuroscience, evolutionary biology, and AI research.

The Alignment Problem brilliantly examines the ethical and technical challenges of aligning machine learning systems with human values, drawing on interviews with leading experts in the field. Brian Christian unpacks real-world issues like algorithmic bias in predictive models and the unintended consequences of AI decisions, from pneumonia diagnostics to broader societal impacts. The book serves as a cautionary yet optimistic exploration of how to ensure AI advances ethically and safely.

AI Fairness presents a groundbreaking framework for evaluating fairness metrics in AI systems, addressing critical decisions in areas like employment, lending, education, and criminal justice. Derek Leben outlines principles of algorithmic justice, using case studies such as the Apple Card and COMPAS to illustrate the trade-offs among competing fairness measures. This new book emphasizes that perfect fairness is often impossible, advocating for a theory of justice tailored to automated decision-making.

Machine Agency delivers a thought-provoking introduction to the concept of agency in AI systems, exploring what qualifies as true agency in machines and its implications for future technologies. James Mattingly and Beba Cibralic provide philosophical tools to debate ownership, action, and responsibility in automated systems, making complex ideas accessible through engaging analysis. The book equips readers to participate in ongoing discussions about intelligent machines in an era of automation.