Ask for a book about artificial intelligence and you will usually be handed one of two things: a textbook with equations in it, or a hot take with a thesis and a deadline. The textbook assumes you want the math. The hot take assumes you have already picked a side. Neither is much use if you simply want to understand what is happening to your work and your news feed.
The five books below take a different route. All of them are narrative nonfiction — they follow people, companies, and arguments, and explain the technology as they go. None of them asks for a math degree. They ask for the patience you would bring to a good magazine feature.
One honest caveat before the list: books about a fast-moving field age at the speed of the field. What these five hold onto is the part that moves slower — who built this, why they built it, and what they are arguing about. That part still tells you most of what you need to know.
The five books
Co-Intelligence, by Ethan Mollick (2024). The practical one. It is about how to actually work with AI rather than watch it from a distance, and it is built on experiments — including the failures, which the author reports with an honesty most tech books skip. If you want a book that changes what you do on Monday morning, this is the one. It is also less interested in where the technology is heading than in what a person with an ordinary job can do with it today.
[Supremacy](https://en.wikipedia.org/wiki/Supremacy:_AI,_ChatGPT_and_the_Race_That_Will_Change_the_World), by Parmy Olson (2024). The rivalry story. OpenAI and DeepMind both set out to build advanced AI in a spirit of idealism, and the book follows what happened when that idealism ran into compute budgets and competition. It reads like the business story it is: two labs, enormous money, and different answers about who should hold the keys.
The Thinking Machine, by Stephen Witt (2025). NVIDIA and Jensen Huang — how one company became the arms dealer of the AI boom. This is the book for understanding why nearly every AI headline is, underneath, partly a hardware story, and why one unglamorous company ended up holding the leash on the whole field.
[Nexus](https://en.wikipedia.org/wiki/Nexus:_A_Brief_History_of_Information_Networks_from_the_Stone_Age_to_AI), by Yuval Noah Harari (2024). The big picture. Harari puts AI inside the much longer history of information networks and argues, sweepingly and contentiously, about what happens now that machines have joined them. You will not agree with all of it, and the book is not shy about taking positions. That is part of its value — treat it as an argument to test your own views against, not a summary of settled fact.
[The Alignment Problem](https://en.wikipedia.org/wiki/The_Alignment_Problem), by Brian Christian (2020). The safety question — how do we make machines that pursue what we actually want — told through the researchers trying to answer it. It is the oldest book on this list and still the clearest introduction to the problem that refuses to go away.
If you only have time for two
Pair one story with one manual. "Co-Intelligence" plus "Supremacy" is the combination that covers the most ground: one tells you how to work with these systems day to day, the other tells you who built them and what they wanted. A reader who finishes both will understand more about the current moment than someone who has skimmed twenty hot takes — including most of the commentary written about them.
That pair also ages well in opposite directions: the manual wears out first as the tools change, the story lasts longer as the incentives stay. Buy the newer edition of the manual when there is one.
The short version: read Mollick first if you want to use AI. Olson or Witt if you want the story — Olson for the labs, Witt for the hardware. Harari if you want the argument. Christian if you want to worry productively.
One more note: none of these is a how-to manual for a specific tool, and that is deliberate. Tools change every quarter. The stories, the money, and the arguments outlast them — which is exactly why these books are still worth your evening.
And read them at book speed. The field will have moved again by the time you finish any of them. The people and the arguments will not have, and those are the parts worth knowing.
No comments yet