Library · book

Artificial Intelligence: A Modern Approach

Stuart Russell & Peter Norvig
2021·Pearson (4th edition)

Source: https://aima.cs.berkeley.edu/

The map of the entire field.

Russell and Norvig's book is the one nearly everyone in AI was taught from, and it does something the current wave badly needs: it shows that intelligence-in-machines is search, logic, planning and reasoning under uncertainty as much as it is learning from data — all under one frame, the rational agent.

The fourth edition brings in modern deep learning, but its enduring gift is perspective: today's LLMs are one technique inside a much older, larger discipline.

For directing AI, it's the reference that stops 'AI' from silently shrinking to 'LLM'.

Central argument

The field's standard, comprehensive textbook, organized around a single unifying idea — the rational agent — and spanning problem-solving and search, logic and knowledge representation, planning, reasoning under uncertainty (probability, Bayesian networks, decision theory), machine learning, and perception and action. The fourth edition integrates modern statistical and deep learning without abandoning the classical breadth.

Critique

Its breadth is also its cost: it is long, encyclopedic and necessarily shallow on any one fast-moving topic, so it is not where you learn the latest on transformers or LLMs. It is a map and a foundation, not a current-events guide — read it for orientation and the parts that don't change.

Why it matters for product

For a CTO/CPO, this is the antidote to AI amnesia: it situates today's models inside the full discipline, so that when an agent needs search, planning, or reasoning under uncertainty you recognize the problem has a fifty-year literature. It keeps 'AI' from collapsing into 'LLM', which is exactly the wider perspective good technology decisions need.