Reading list

A short list, kept short on purpose. Each entry says who it is for.

Books

• Nick Bostrom, Superintelligence: Paths, Dangers, Strategies (2014). The book that put the word into general use. Dense; read the first three chapters if nothing else.
• Stuart Russell, Human Compatible (2019). The most readable account of why controlling a very capable AI is hard, by one of the field's senior figures.
• Brian Christian, The Alignment Problem (2020). Reportage rather than argument; good if you want the people and the history.

Reports

• International AI Safety Report 2026: chaired by Yoshua Bengio, written by more than 100 experts, backed by over 30 countries. The closest thing to a consensus statement of the evidence on general-purpose AI.

Essays and papers

• I. J. Good, "Speculations Concerning the First Ultraintelligent Machine" (1965). Where the idea of an intelligence explosion starts. Short, and still the clearest statement of it.
• Vernor Vinge, "The Coming Technological Singularity" (1993). The origin of "the singularity" as a term. A few pages.
• Nick Bostrom, "How Long Before Superintelligence?" (1998). The essay that preceded the book; a compact version of the argument.

This list is kept deliberately short and is added to as pages on the site are written.