The Quest for Artificial Intelligence
A comprehensive technical history of artificial intelligence written from the inside by a Stanford pioneer who was there from the 1960s onward.
Nilsson covers the full arc — from early cybernetics and logic through search, knowledge representation, machine learning, and robotics — with the rigour of someone who contributed to the foundational work.
Unlike journalistic histories, this book explains the ideas themselves, not just the personalities.
The full text is freely available on Nilsson's Stanford page, making it an unusually accessible primary source.
For anyone who wants to understand AI's recurring patterns of promise and disappointment through the actual technical choices that were made, this is the reference.
Central argument
Nilsson argues that the history of AI is best understood as a sequence of technical choices — about representation, search, learning, and reasoning — each of which unlocked certain capabilities while foreclosing others, producing the field's recurring cycles of optimism and stagnation. Rather than attributing AI winters to hype or funding politics, he locates the causes in the internal logic of competing paradigms: symbolic vs. subsymbolic, logic-based vs. statistical, deliberative vs. reactive. The central claim is that understanding these architectural trade-offs, not just the outcomes, is what allows a reader to reason about AI's trajectory rather than merely observe it.
Critique
Because Nilsson writes as a Stanford insider who contributed primarily to symbolic and logic-based AI, the book's treatment of connectionism and statistical learning — the paradigms that ultimately came to dominate — is comparatively thinner and carries an implicit framing shaped by his own intellectual commitments. A reader relying on this history to understand why deep learning succeeded where earlier approaches failed may find the account structurally incomplete: the very transitions that matter most to contemporary practice are those Nilsson observed from the outside. The book was also published in 2010, placing it just before the empirical breakthroughs that redrew the field entirely.
Why it matters for product
For a CPO evaluating AI-powered product investments, Nilsson's framework of recurring paradigm cycles is directly applicable to avoiding capability inflation in roadmaps — the pattern of overcommitting to a technical approach before its failure modes are understood is as common in product strategy as it was in AI research. More concretely, understanding why knowledge-representation approaches stalled — the brittleness of hand-curated rules at scale — gives product leaders a sharper vocabulary for interrogating whether a proposed AI feature is encoding the same fragility at the data or prompt level. This is the kind of structural literacy that separates a CPO who can challenge an AI engineering team from one who simply approves their estimates.