Why Read When the Machine Can Answer?

Two propositions

When I speak about reading these days, I put two statements side by side and ask the audience to choose. The first: reading is the only way to harness human intellect and answer human query. The second: in the age of information overload and AI, one need not waste time reading copious amounts of books when the answers are all readily available.

A few years ago, the second statement would have sounded like a provocation. Today it sounds like common sense to many people, and I understand why. A machine will now summarise a five-hundred-page history in a paragraph, explain a difficult idea at whatever level you ask for, and do it in seconds, at any hour, without complaint. If reading was always a means of getting answers, and answers have become cheap, then the long hours with a book look like an inefficiency we can finally retire.

I do not think the second statement is foolish. I think it is mistaken about what reading is for. It treats a book as a container of answers, and the reader as someone trying to get the answers out as quickly as possible. On that view, of course a faster extraction tool wins. But the most important things reading has given me were never answers. They were better questions, a sense of how much I did not know, and the slow, unglamorous ability to weigh one claim against another. Those are not things a summary can hand over, because they are not facts. They are capacities, and capacities are built only by exercise.

This essay is my attempt to set out that case properly. It is not an argument against AI. I use these tools and find them genuinely useful. It is an argument about division of labour: about what the machine does well, what books still do better, and what remains, stubbornly, the work of the person in the chair.

What the machine does well

I find it useful to lay the question out in three columns: what AI does well, what books do better, and what judgment still requires of you.

The first column deserves to be taken seriously, because the case for reading is not helped by pretending the machine is weak. It is not. Ask it for the date of a treaty, the sequence of events in a campaign, or the main arguments of a book you have not read, and it will usually give you something serviceable at once. It retrieves information faster than any library catalogue I have used. It summarises with a fluency that would have taken a research assistant an afternoon. It recognises patterns across more text than any human could read in a lifetime.

Above all, it offers breadth and speed. It can range across economics, military history and molecular biology in the same conversation, and it never tires. For a busy professional, a student facing an examination or anyone who simply needs to know something quickly, these are real gifts. I would not advise anyone to refuse them.

But notice what all six of those strengths have in common. Each is about delivering an output: an answer, a summary, a retrieved fact, a pattern found. The machine is extraordinarily good at the destination. What it cannot do on your behalf is the journey, and in the life of the mind the journey is where most of the value lies. A summary of a book is to the book roughly what a photograph of a mountain is to the climb. The photograph is accurate, and it may be beautiful, but nobody ever got fitter by looking at one.

What books do better

The second column is where reading earns its keep: deep understanding, context and nuance, intellectual struggle, building mental models, reflection and depth.

Take understanding first. Knowing that something happened is not the same as understanding why it happened. A summary will tell you that Japan defeated Russia in 1905. A good history will make you live through the decisions: what Tokyo wanted and, just as important, what it chose not to want; how far its generals could push before the money ran out; why the diplomats were already preparing the peace while the guns were still firing. Only after travelling that road do you begin to see the shape of a limited war, and to recognise that shape when it appears again in another century under another name. The fact is portable. The understanding has to be grown.

Context and nuance follow from the same truth. An author with three hundred pages can afford qualifications, exceptions and counter-evidence. A summary has to discard them, because that is what summarising means. Yet in history, strategy and public life, the qualifications are often where the truth lives. The confident one-line verdict is usually the least reliable thing anyone can tell you about a complicated event.

Then there is struggle, which is the item on the list people most want to skip. A demanding book resists you. It uses terms you have to look up, makes arguments you have to hold in your head across chapters, and sometimes makes you go back twenty pages because you missed something. That resistance is not a defect of the medium. It is the medium doing its work. The mind, like a muscle, grows against load, and a tool that removes all the load also removes the growth.

Out of that struggle comes the most valuable thing reading builds: mental models. After enough books on war, economics or the rise and fall of states, you no longer meet each new event as a stranger. You have frames to place it in, analogies to test it against, and a sense of which questions to ask first. These models are not stored in any book. They are assembled, slowly, inside the reader, from hundreds of hours of attention.

Finally, reflection and depth. A book sets its own pace and leaves room around the words. You put it down, walk, argue with it in your head, pick it up again. An answer engine is designed to close a question as fast as possible. A good book is often designed to keep one open long enough for you to think.

What judgment still requires of you

The third column is the one I care about most, because it belongs to neither the machine nor the book. It belongs to you: weighing competing truths, making difficult decisions, recognising trade-offs, moral and ethical reasoning, knowing what matters, and, at the end of the list, wisdom.

Most serious questions do not have one true answer waiting to be retrieved. They have several partial truths in tension. Was a particular war a defensive necessity or a strategic blunder? Does a new technology deserve to be embraced, regulated or feared? Ask a machine and it will very sensibly give you both sides, neatly balanced. But balance is not judgment. At some point a person has to decide which considerations weigh more in this case, at this moment, with these stakes, and accept responsibility for the choice.

Decisions of that kind are always trade-offs. Every course of action buys something and pays for it with something else. The skill of seeing the price as clearly as the benefit is rare, and it is learnt above all from watching others make such choices: the commanders, statesmen and ordinary people whose decisions fill the pages of history and biography. Reading is the cheapest apprenticeship in judgment ever invented. You can sit beside a thousand decision-makers, see what they saw, and then see what they missed.

Moral reasoning cannot be outsourced either. A tool can tell you what different ethical traditions would say. It cannot make you the kind of person who notices when something is wrong or give you the steadiness to act on it. That disposition is formed by long exposure to human situations, many of which you will only ever meet on the page.

And beneath all of this lies the quiet skill of knowing what matters. In an age of infinite information, the scarce resource is not answers but attention. Someone must decide which of the thousand available facts are worth knowing and which questions are worth asking at all. Wisdom, if it can be defined, is perhaps just that: the habit of putting your attention where it belongs and knowing the limits of what you know.

Answering, understanding, deciding

If I had to put the whole argument in three lines, it would be these:

AI can help you answer questions.

Books can help you understand questions.

Judgment helps you decide which questions matter.

The three verbs describe a ladder. Answering is the lowest rung, and for the first time in history it has been largely automated. That is no small thing, and we should be grateful for it. But the higher rungs have not moved. Understanding a question means knowing where it came from, what it assumes, what an answer would and would not settle, and how it connects to other questions. That is what a serious book trains you to do. Deciding which questions matter is higher still. It is the difference between an analyst who answers whatever is put in front of them and a leader who knows what ought to be asked in the first place.

There is a real danger in the age of AI, and it is not that machines will think for us. It is that we will stop noticing the difference between having an answer and understanding one. A person who has read only summaries can repeat conclusions with great confidence and still have no idea how fragile they are. They have the words without the weight. Worse, because the answers come so easily, they may never discover how much they do not know. The machine is always confident in tone. The book, at its best, teaches humility.

This is why I have always said that reading is about harnessing the intellect and knowing the limits of what you know. Every good book I have read has left me with more questions than I started with, and that is a sign it worked. The more you read, the more you know how little you know. No answer engine will ever give you that feeling, because it is not designed to. It is designed to satisfy you. A book is often designed to unsettle you, and the unsettled mind is the one that keeps learning.

Used well, the two can work together. AI can clear away the lowest rung so that the reader has more time for the higher ones: checking a date, tracing a reference, explaining an unfamiliar term in the middle of a hard chapter. Used badly, it lets us skip the climb altogether while believing we have reached the top.

Back to the two propositions

I began with two statements and asked which one to choose. I choose the first, and the argument of this essay is why. The second statement is right that answers are now everywhere. Its mistake is to think answers were ever the point.

A reader does not go to a book only to find out what happened. He goes to understand why it happened, what else might have happened, and what it means for the questions in front of him today. Along the way something else happens, which no summary can reproduce. The reader discovers how much he did not know. Every serious book I have finished has left me with a longer list of questions than I started with. I count that as its greatest gift, not a failure.

None of this means turning away from the new tools. By all means let the machine check a date, trace a reference or explain a difficult term in the middle of a hard chapter. Then go back to the page. Use the summary to decide what to read, never as a substitute for reading it. Keep the books that changed your mind, because a private library is a record of the questions you have lived with.

The real risk of the age of AI is not that machines will think for us. It is that we will mistake having an answer for understanding one, and never discover the edges of our own knowledge. Reading guards against that. It harnesses the intellect, and it keeps the intellect honest about its limits. In a world flooded with answers, the person who knows which questions matter, and how little he yet knows, will be the one worth listening to.