The Best Books on the History of Artificial Intelligence
Ten books, ranked — from Turing to transformers, the real story behind the AI moment
The best single book on the history of artificial intelligence is Cade Metz's Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World (Dutton, 2021) — a reported, character-driven narrative that follows Geoffrey Hinton, Yann LeCun, Yoshua Bengio, Demis Hassabis, Ilya Sutskever, and the small, once-derided deep-learning community from academic exile in the 1980s and 1990s to the center of the tech industry after AlexNet's 2012 breakthrough. It reads like a thriller, it is accurate, and it is the right place to start if you want to understand how today's AI boom actually happened, told through the people who built it rather than the algorithms alone.
But 'history of AI' is really three stories layered on top of each other: the founding one (Turing's 1950 question of whether machines can think, the 1956 Dartmouth workshop that coined the term, and the boom-bust cycles of symbolic AI through the 1970s and 1980s 'AI winters'); the deep-learning insurgency (statistical pattern-matching and neural networks displacing hand-coded logic, from backpropagation through AlexNet to the 2017 Transformer architecture that made large language models possible); and the reckoning now underway over what these systems mean — for labor, safety, and whether scaling curves lead anywhere near general intelligence. This list covers all three: the popular narrative history, the primary source (Turing's own writing), the scholarly standard textbook, the philosophical case for taking AI risk seriously, and the recent books arguing about where this all goes.
Every title below has been live-verified against Open Library's ISBN lookup API, in addition to publisher, Amazon, and AbeBooks bibliographic records. Where a book has been reissued or updated, the annotation says which edition to buy and why it matters.
The books
1. Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World — Cade Metz (2021)
Deep learning's triumph was not inevitable: it was the multi-decade, often-mocked project of a small group of researchers who kept betting on neural networks through two AI winters until the compute and data finally caught up.
The best popular entry point, full stop. Metz, a longtime New York Times technology reporter, spent years interviewing the deep-learning community directly — Geoffrey Hinton, Yann LeCun, Yoshua Bengio, Demis Hassabis, Ilya Sutskever, Andrew Ng — and turns their decades in the academic wilderness, followed by the 2012 AlexNet result that made neural networks suddenly indispensable, into a fast, human, well-sourced narrative. It is also a genuine audiobook standout: the reported dialogue and scene-setting carry well aloud, and it is widely available through major audiobook subscription services. Kirkus called it a vivid group portrait of the researchers who 'brought AI to Google, Facebook, and the world' — which is exactly what it delivers.
Pick this if: Anyone who wants the whole modern AI story — 1980s neural-net exile through the 2020s corporate AI race — told through its people. (Level: Beginner)
2. Machines Who Think: A Personal Inquiry into the History and Prospects of Artificial Intelligence — Pamela McCorduck (2004)
AI is the latest chapter in humanity's ancient dream of creating thinking machines in our own image — and its 1950s–1970s founders pursued that dream with real, if often overconfident, scientific rigor.
The classic. First published in 1979 and reissued in this expanded 25th-anniversary edition (A K Peters, 2004) with a new afterword tracking the field into the 2000s, McCorduck's book is the original definitive history of AI's founding generation — she interviewed Marvin Minsky, John McCarthy, Allen Newell, and Herbert Simon directly, and traces the field's ambition back through Pygmalion and the golem to the 1956 Dartmouth workshop. It is the primary account historians of AI still cite for how the symbolic-AI founders themselves understood their project, written before deep learning existed to reframe the story.
Pick this if: Readers who want AI's origin story from the people who coined the term, not a retrospective gloss on it. (Level: Intermediate)
3. The Essential Turing: Seminal Writings in Computing, Logic, Philosophy, Artificial Intelligence, and Artificial Life, Plus The Secrets of Enigma — Alan Turing (B. Jack Copeland, ed.) (2004)
The question of whether machines can think is best replaced by a testable, operational one — the imitation game — and Turing predicted, provisionally but seriously, that machines would eventually pass it.
The primary source. Copeland's Oxford collection gathers Turing's own papers in one place, most importantly 'Computing Machinery and Intelligence' (Mind, 1950), the paper that opened with 'Can machines think?', proposed the imitation game now known as the Turing Test, and predicted that by the year 2000 people would speak of machines thinking without being contradicted. Reading Turing's actual argument — including his rebuttals to nine objections he anticipated, from theological to Lady Lovelace's — is a different experience than reading summaries of it, and this edition situates the AI paper alongside his foundational 1936 computability work.
Pick this if: Readers who want to encounter the field's founding question in Turing's own words before reading anyone's history of it. (Level: Scholarly)
4. Alan Turing: The Enigma (Updated Edition) — Andrew Hodges (2014)
Turing's 1950 case for thinking machines cannot be separated from the rest of his intellectual life — the universal machine, wartime cryptanalysis, and his own persecuted difference all fed the same restless question of what minds are.
The definitive biography of the field's founding figure, and the book that inspired the film The Imitation Game. Hodges, a mathematician himself, reconstructs Turing's path from the 1936 universal-machine paper through Bletchley Park codebreaking to his 1950 AI paper and his prosecution and death in 1954 — and does not soften the record on his persecution. Buy this 2014 updated paperback (Princeton University Press), which adds a new preface by Hodges reflecting on the centenary; the original 1983 edition is the same core text but without it.
Pick this if: Readers who want to understand the mind and life behind AI's founding question, not just the paper it produced. (Level: Intermediate)
5. Artificial Intelligence: A Modern Approach (4th Edition) — Stuart Russell and Peter Norvig (2020)
AI is best understood as the study of rational agents that perceive and act — a unifying framework broad enough to encompass search, logic, probability, learning, and the deep-learning methods that now dominate the field.
The scholarly standard, and has been since its first edition in 1995 — used at over 1,500 universities and carrying tens of thousands of citations. This fourth edition (Pearson, 2020) is a genuine overhaul, not a refresh: expanded coverage of deep learning, transfer learning, multiagent systems, probabilistic programming, and — notably — a new chapter on AI safety and ethics. It is a textbook, not bedtime reading, but it is the reference every other book on this list is implicitly arguing with or building on.
Pick this if: Readers who want the field's actual technical map, not just its narrative history. (Level: Scholarly)
6. The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World — Pedro Domingos (2015)
Machine learning has multiple, historically rival paradigms, and the still-unsolved challenge is finding — or building — a Master Algorithm that unifies them.
The best guide to why machine learning splintered into competing schools before deep learning's recent dominance. Domingos, a University of Washington computer scientist, walks through five 'tribes' of machine learning — symbolists, connectionists, evolutionaries, Bayesians, and analogizers — and argues each captures part of a still-missing unified theory of learning. It is the book to read to understand that 'AI' was never one idea, and that the neural-network approach other books on this list focus on is one contender among several serious research programs.
Pick this if: Readers who want the intellectual taxonomy of machine learning, not just the deep-learning victory narrative. (Level: Intermediate)
7. Superintelligence: Paths, Dangers, Strategies — Nick Bostrom (2014)
A machine intelligence that surpasses human cognition across the board could rapidly gain a decisive strategic advantage, and there is no guarantee its goals would remain aligned with human interests unless that problem is solved in advance.
The book that moved AI risk from science fiction to a subject serious philosophers, technologists, and eventually governments would engage with. Bostrom, founding director of Oxford's Future of Humanity Institute, works through how a superintelligent system might arise, what strategic advantages it could gain, and why aligning its goals with human values is a hard, unsolved problem — not a footnote to be handled later. A New York Times bestseller on its release, it remains the reference point every subsequent AI-safety book (including Christian's, below) responds to or builds on.
Pick this if: Readers who want the rigorous philosophical case for why advanced AI is a serious risk, not just a research curiosity. (Level: Scholarly)
8. The Alignment Problem: Machine Learning and Human Values — Brian Christian (2020)
Making machine-learning systems do what we actually mean, rather than a narrow proxy of it, is a live technical and ethical challenge already visible in deployed systems — not a speculative future one.
The best-reported account of how the field itself started grappling with the gap between what machine-learning systems are trained to optimize and what humans actually want. Christian interviewed the researchers building reinforcement learning, fairness, and interpretability tools inside the same labs Metz profiles in Genius Makers, and the two books pair well: Genius Makers tells you how deep learning won, and The Alignment Problem tells you what its winners started worrying about next. It won the 2021 Association for Computing Machinery's Eugene L. Lawler Award and was widely and favorably reviewed on release.
Pick this if: Readers who finished the deep-learning triumph narrative and want the honest sequel about its unresolved problems. (Level: Intermediate)
9. Life 3.0: Being Human in the Age of Artificial Intelligence — Max Tegmark (2017)
Advanced AI could enable humanity's best or worst possible future, and which one we get depends on choices being made now about goals, safety, and governance — not on technology alone.
MIT physicist Tegmark's contribution to the post-2015 wave of serious popular AI-futures books, and the most readable one aimed at a general audience rather than specialists. He frames biological evolution as 'Life 1.0,' culturally-adaptive humans as 'Life 2.0,' and asks what 'Life 3.0' — a civilization that can redesign its own software and hardware — might look like, walking through scenarios from beneficial superintelligence to existential catastrophe without picking one as inevitable. It is well suited to audiobook listening for readers who want the AI-futures conversation on a commute rather than in a chair.
Pick this if: Readers who want the most accessible tour of AI's long-term stakes, written for people outside the field. (Level: Beginner)
10. AI Superpowers: China, Silicon Valley, and the New World Order — Kai-Fu Lee (2018)
AI's next phase is a race for applied deployment and data scale rather than algorithmic novelty, and China's advantages in that race make it a true superpower rival to the United States.
The geopolitical chapter the rest of this list mostly skips. Lee — a computer scientist who ran Google China and Microsoft Research Asia before becoming a venture capitalist — argues that the era of algorithmic breakthroughs (Turing through transformers) is giving way to an era of implementation, where China's scale of data, entrepreneurial intensity, and state support make it a genuine peer competitor to Silicon Valley. Written just as Chinese AI investment was accelerating, it remains the clearest explanation of why 'history of AI' from 2018 onward is inseparable from US–China competition.
Pick this if: Readers who want the deep-learning story extended into industrial and national-competition terms. (Level: Intermediate)
From Turing's question to the transformer
The history of AI has a recognizable arc, and it helps to know its beats before diving into any one book. It starts with Alan Turing's 1950 paper 'Computing Machinery and Intelligence,' which proposed the imitation game as a way to sidestep the unanswerable philosophical question of machine consciousness in favor of a testable behavioral one. Six years later, the 1956 Dartmouth Summer Research Project — organized by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon — coined the term 'artificial intelligence' and set off decades of symbolic, rule-based AI research: expert systems, logic programming, and the confident (often overconfident) predictions McCorduck's Machines Who Think documents from the inside.
That symbolic-AI program hit real limits, producing two 'AI winters' — funding and confidence collapses in the mid-1970s and again in the late 1980s — when hand-coded rules failed to scale to real-world messiness. The field's revival came from a different direction entirely: statistical, data-driven machine learning, and specifically neural networks, which had themselves been left for dead after Minsky and Seymour Papert's 1969 critique Perceptrons. Geoffrey Hinton, Yann LeCun, and Yoshua Bengio kept working on neural networks through the 1980s and 1990s largely outside the mainstream — the period Genius Makers covers in the most detail — until three things converged around 2012: much larger labeled datasets (ImageNet), much cheaper parallel compute (GPUs), and refined training techniques. AlexNet's 2012 ImageNet win, cutting the error rate roughly in half versus the runner-up, is the conventional starting gun for the deep-learning era.
From there the timeline compresses fast. Google's 2017 paper 'Attention Is All You Need' introduced the Transformer architecture that underlies essentially every major language model since; OpenAI's GPT series (2018 onward) showed that scaling that architecture with more data and compute produced increasingly general capabilities; and by the early 2020s, systems like ChatGPT had moved the conversation from 'can machines think?' back to something closer to Turing's original question, but now asked about systems already in half a billion people's hands. The scholarly and public debate current books like Superintelligence, The Alignment Problem, and Life 3.0 are having — over whether this trajectory leads toward artificial general intelligence, and how to keep it safe if it does — is the live continuation of the argument Turing opened in 1950.
The verdict
Start with Genius Makers for the modern narrative and Machines Who Think for the founding one — together they cover roughly 1950 to 2020 in readable, well-sourced prose. Add The Essential Turing if you want the field's founding argument in its own words, and Alan Turing: The Enigma if you want the life behind it. Russell and Norvig's Artificial Intelligence: A Modern Approach is the reference to own once you want the actual technical map rather than narrative history. For where the field goes next, Superintelligence and The Alignment Problem are the serious risk-and-ethics pair, Life 3.0 is the most accessible futures book, Domingos's The Master Algorithm is the best guide to machine learning's competing schools of thought, and Kai-Fu Lee's AI Superpowers is the one book here that takes the geopolitics as seriously as the technology.
At a glance
| Book | Year | Difficulty | Core thesis |
|---|---|---|---|
| Genius Makers — Metz | 2021 | Beginner | Deep learning's triumph was decades of unglamorous persistence, not inevitability |
| Machines Who Think — McCorduck | 1979 (rev. 2004) | Intermediate | AI's founders pursued an ancient dream of thinking machines with real scientific rigor |
| The Essential Turing — Copeland (ed.) | 2004 | Scholarly | The Turing Test reframes 'can machines think?' as a testable, operational question |
| Alan Turing: The Enigma — Hodges | 1983 (upd. 2014) | Intermediate | Turing's AI question grew out of his whole life — computability, codebreaking, persecution |
| AI: A Modern Approach — Russell & Norvig | 1995 (4th ed. 2020) | Scholarly | AI as the study of rational agents, unifying search, logic, probability, and learning |
| The Master Algorithm — Domingos | 2015 | Intermediate | Machine learning has rival paradigms; a unifying Master Algorithm remains unsolved |
| Superintelligence — Bostrom | 2014 | Scholarly | Superintelligent AI could gain decisive advantage; alignment must be solved in advance |
| The Alignment Problem — Christian | 2020 | Intermediate | Getting ML systems to do what we mean, not a proxy of it, is already a live problem |
| Life 3.0 — Tegmark | 2017 | Beginner | AI could enable humanity's best or worst future, depending on choices made now |
| AI Superpowers — Lee | 2018 | Intermediate | AI's next phase is a US–China race for data scale and deployment, not just algorithms |
Frequently asked questions
What is the best book on the history of artificial intelligence?
For most readers, Cade Metz's Genius Makers (2021) is the best starting point — a well-reported narrative following the deep-learning researchers (Geoffrey Hinton, Yann LeCun, Yoshua Bengio, Demis Hassabis, and others) from academic obscurity in the 1980s to the center of the AI boom after 2012. For the field's earlier, founding history — the 1950s Dartmouth workshop through the 1970s and 1980s — Pamela McCorduck's Machines Who Think, written with direct access to AI's founders, remains the classic reference.
What book should I read to understand Alan Turing's role in AI?
Read Turing's own 1950 paper 'Computing Machinery and Intelligence,' collected with his other foundational writing in B. Jack Copeland's The Essential Turing (Oxford, 2004). For the life and context behind it, pair that with Andrew Hodges's biography Alan Turing: The Enigma — get the 2014 updated Princeton paperback, which adds a new preface reflecting on Turing's centenary and legacy.
Is there a standard textbook on artificial intelligence?
Yes — Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig, first published in 1995 and now in its 4th edition (Pearson, 2020), is used at over 1,500 universities and is widely considered the field's standard reference text. The 4th edition substantially expanded its coverage of deep learning, multiagent systems, and AI safety and ethics compared to earlier editions.
What caused the AI 'winters,' and what ended them?
The AI winters — roughly the mid-1970s and late 1980s — were funding and confidence collapses that followed overpromised results from symbolic, rule-based AI systems that failed to scale to messy real-world problems; Minsky and Papert's 1969 book Perceptrons also dampened enthusiasm for neural networks specifically for years. The field's modern revival traces to deep learning: neural networks trained on much larger datasets with much cheaper GPU compute, dramatized by AlexNet's 2012 ImageNet win, followed by the 2017 Transformer architecture that made today's large language models possible.
What's the best book on AI risk and the alignment problem?
Nick Bostrom's Superintelligence (2014) is the foundational philosophical case for taking advanced-AI risk seriously, and remains the reference point later books respond to. Brian Christian's The Alignment Problem (2020) is the best-reported account of how the alignment challenge shows up concretely inside the machine-learning research labs building today's systems, and pairs well as a more grounded, narrative follow-up to Bostrom's more abstract argument.
Explore related events on the timeline
- The (projected) Dawn of AGI on the interactive timeline
- AlexNet and the 2012 deep-learning breakthrough
- AlphaGo — deep learning's landmark game-playing victory
- GPT-2 and the rise of large language models
Sources consulted
- Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World (Penguin Random House)
- Machines Who Think: A Personal Inquiry into the History and Prospects of Artificial Intelligence (Routledge/A K Peters)
- The Essential Turing (Oxford University Press)
- Alan Turing: The Enigma, Updated Edition (Princeton University Press)
- Artificial Intelligence: A Modern Approach, 4th Edition (Pearson)
- The Master Algorithm (Hachette Book Group / Basic Books)
- Superintelligence: Paths, Dangers, Strategies (Oxford University Press)
- The Alignment Problem: Machine Learning and Human Values (W. W. Norton)
- AI Superpowers: China, Silicon Valley, and the New World Order — Wikipedia
As an Amazon Associate, AskHistoryAI earns from qualifying purchases. Recommendations are chosen editorially — see our methodology.