From Computation to Thinking
How a machine that only follows rules comes to read, reason and act, and what that means for the people whose work it touches.
Free to read, download and share. Version 0.60, October 2026.
It is two in the morning
A parent types into a phone: "My daughter is 3. She's had a fever all evening, 39.5. Now she has little purple spots on her legs, and they don't go away when I press a glass on them. Should I wait until morning?"
A good nurse knows at once this cannot wait. Today the answer may come from a machine. Inside, that machine holds only a clock and billions of switches, each on or off. No single switch knows anything about children or fevers.
So how does it give the right answer? This book builds the machine step by step to find out.
Who it is for
- Curious readers with no technical background.
- Users of chatbots and AI agents who wonder how they work.
- Professionals who work, or soon will, alongside these systems.
- Managers and policymakers who decide how these systems are used.
- Teachers and students looking for a clear introduction to AI.
What you will understand
- How a machine that only moves symbols can compute anything that can be computed, and what no machine can ever do.
- Why writing the rules by hand failed, and how machines learn rules nobody could write.
- Why predicting the next word teaches a machine so much about the world.
- How a predictor is turned into a helpful assistant, and why it can still bluff.
- What "reasoning" means for a machine, and why writing out its steps helps.
- How much it takes to build one, in books, kettles and cities.
- Whether its words are about anything, and when it should stop and hand over to a person.
- What happens when the machine joins a hospital team, and whether it will replace the nurse or relieve her.
How long it takes
About three hours, or 99 pages. The book has eighteen short chapters, most of them 6 to 13 minutes long. It has no mathematics or programming. One example runs through all of them: the parent, the child and the nurse.
Contents
- Part I. Machines That Follow Rules. 1 Moving Symbols · 2 One Machine for Every Rulebook · 3 The Walls
- Part II. Machines That Find Rules. 4 Writing the Nurse's Rules by Hand · 5 Learning · 6 Why Learning Can Work at All · 7 Learning to See and Hear
- Part III. Machines That Predict Language. 8 Predicting the Next Word
- Part IV. From Predictor to Agent. 9 Imitation · 10 Learning from Consequences · 11 Thinking Out Loud · 12 Agents · 13 How Big Is It?
- Part V. What It Is, Really. 14 Meaning · 15 Agency and Curiosity · 16 The Walls Again, and Knowing When to Stop · Coda: Is Anyone Home?
- Part VI. Living with It. 17 The Machine on the Ward · 18 Replace or Relieve?
For teachers and reading groups
A free teaching guide maps the book to six sessions, with discussion questions and an activity for each. Download the teaching guide
How this book was written
This book was written by AI under my direction. I worked like a film director assembling scenes. I designed the outline and the running example, and I set the logic and flow of the presentation. Claude, Anthropic's model, drafted the text and most of the diagrams. Google's Gemini drew fourteen diagrams, which were then checked and corrected. OpenAI's ChatGPT and other AI models reviewed the material and gave feedback, as did human readers. I wrote the preface and revised the book over sixty versions. The studies it cites were checked against their published sources. Mistakes may remain; they are mine to fix.
Licence and versions
This book is licensed under Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0): you may share and adapt it for any non-commercial purpose, with credit. Each version is dated. What changed