TheOutsourced Mind

There is a particular sound a person makes when a machine tells them something they were about to work out for themselves. It is a small sound, somewhere between relief and surrender, and I have started to notice it everywhere.

I am a cardiologist. I have spent my working life watching people make decisions under pressure, usually about their own bodies, usually with incomplete information and a clock running. That is a strange training for a book about machines, and I want to be honest at the outset that I am not a computer scientist and this is not a book about how the technology works.

It is a book about what happens to us while we use it.

I should say what this book is not, because the shelf it sits on is crowded with things I am not writing. This is not a book about machines taking over. It is not about whether they will replace your job, or wake up, or end the world. Those may be real questions and I have no special standing to answer them. They also have a habit of going stale, because they are predictions, and predictions are perishable.

What I am writing about does not go stale. It is this: when a tool takes over a task you used to do yourself, something in you stops being practiced. That has been true of every tool that ever did this, and it will be true of the next one. It is not a warning about artificial intelligence. It is a description of people.

The reason it matters now is that the thing being taken over has changed.

Every previous tool of this kind took over storage or retrieval. The calculator held the arithmetic. The map held the route. The search engine held the facts. Useful, all of them, and each one cost a little something. But the thinking stayed where it was. You still had to decide what to ask, what the answer meant, and whether it made sense.

An intelligent machine takes over that part. It does the reasoning step. You describe a situation and it hands you a conclusion, in confident prose, and the conclusion is usually good. That is a genuinely new thing in the world, and I do not think it is well served by either of the two available reactions: the people who think it is magic and the people who think it is nothing.

I use these tools. I want that on the record early, because a book that lectures you about a thing its author refuses to touch is not worth reading. I used them to write this. They have made me faster and, in a few places, better. They have also confidently handed me things that were wrong, in the same even tone they use for things that are right, and the only reason I caught them is that I still knew enough to check.

Here is the argument, stated plainly, so you can decide whether to keep going.

You do not need to do it yourself. You need to keep enough of the skill to catch the machine when it is confidently wrong.

That is not a compromise position, and it is not a way of having it both ways. It is the same thing we decided about the calculator fifty years ago, and we were right. Nobody does long division by hand anymore, and nobody should. But we still teach children that seven times eight is fifty-six, and the reason is not nostalgia. The reason is that a person with number sense sees that the screen says five hundred and sixty and knows, instantly, without doing any arithmetic, that something has gone wrong.

That capacity, the one that notices, is the subject of this book. Where it comes from, what maintains it, what happens when it stops being used, and why it disappears without ever telling you that it has.

There are two people I am writing for, and they think they have nothing in common.

The first has embraced all of this and is a little uneasy about how much they now hand over. They are not worried about the machine. They are worried, in a way they find hard to name, about themselves.

The second will not go near it. They do not trust what they cannot see, and they have decided that the safest thing is to have nothing to do with it. I have a great deal of sympathy for this position and I think it fails, for a reason that will take a chapter to explain.

Both of these people are trying to solve the same problem, which is how to know when to trust something. Neither of them has been given a way to do it. That is what I am trying to provide.

A word about children, since learning is in the subtitle and children are the reason a lot of people pick up a book like this. Everything in here is harder where the young are concerned, and the reason is worth getting right. Everywhere else in this book, a skill is being maintained. A person had it, used it less, and lost some of it. In a classroom, the skill is being built for the first time. You cannot lose what was never there, and you also cannot tell the difference from the outside: a student who never built the capacity and a student who has it and did not use it today produce identical work.

That is a genuinely hard problem and I do not want to pretend otherwise, or to pretend that I have watched it professionally. I have not. I have watched adults lose things for forty years. On classrooms I am a parent, which is to say I am in the same position you are.

The book runs in four parts. The first walks the history, because the panic we are having now is not the first one and the previous rounds have things to teach us. The second explains what is actually different this time. The third is what I have watched happen to human capacities when they go unused, which is the part I can speak to with some authority. The fourth is what to do about it, and it is shorter than you might expect, because the answer is small.

The answer being small is not a disappointment. It is the finding.

Two men are walking outside the city walls on a hot afternoon, barefoot in a stream, talking the way friends talk when there is nowhere to be. One is Socrates. The other is a young man named Phaedrus, and the dialogue Plato wrote about that afternoon, twenty-four hundred years ago, is named for him. Somewhere in it Socrates tells a story about the invention of writing, and what he says is that writing will not make people wiser. It will make them forgetful. They will trust marks on a page instead of what they carry inside them, and they will seem to know a great deal while knowing almost nothing.

We have Socrates in writing. The complaint survives because someone wrote it down. He lost the argument to the very thing he was warning about, and the loss was so complete that we quote him from a page.

It was said again about printing.

In the 1490s a German abbot named Johannes Trithemius wrote a treatise in praise of scribes. The press had been running in Europe for about forty years, and he wanted his monks to keep copying books by hand. Copying, he argued, was a discipline and a devotion, and a monk who stopped was giving up more than the labor. He also thought the new books would not last. The word written on parchment, he said, would last a thousand years. The printed word was on paper, and he gave it two hundred at most.

His treatise was printed. It came off a press in Mainz, the city where Gutenberg started, in 1494, and copies of that printing still exist, well past the two hundred years he allowed them.

He was wrong about the paper. About the monks I am less sure. A man who copies a book by hand knows it in a way the man who owns it does not, and something did go when the copying stopped. It went quietly, and it was not what he had been watching for.

It has been said since, about other things. About spelling. About arithmetic. About knowing your way from one place to another. About remembering what you need to know. And now about judgment: about deciding what to do.

This is the first one we had pictured before it arrived: a film in 1968 gave the machine a calm voice and the name HAL, and the picture was waiting.

I hear that fear often now, and not from people who work in technology. I hear it from patients across the desk, from family at dinner, from friends who would not otherwise raise the subject with me. It is not that the machine will make a mistake. It is that it is moving too fast, that it takes over one ordinary task and then the next, and that somewhere at the end of that line it takes over the person. There is a great deal of skepticism in it, and I do not think it is foolish.

It is not only the people I know. In 2025 the Pew Research Center asked more than five thousand Americans about it. Half said they were more concerned than excited about the growing use of these machines in daily life, up from thirty-seven percent four years earlier. Fifty-three percent expected them to make people worse at thinking creatively. Sixteen percent expected them to make people better.

The loudest form of the fear is a book. In 2025 Eliezer Yudkowsky and Nate Soares published If Anyone Builds It, Everyone Dies, and it reached the best-seller list. Its argument is that a machine smarter than we are would pursue ends of its own, and that we would not survive it. I have read it. Part of it is a story (an imagined account of how it could happen) and part of it is science that is easy to believe. Serious people take it seriously, and serious people argue that it is wrong. I am not in a position to settle that, and this book does not try.

In the film there is a moment you can point to. The machine refuses, and a man is locked out. Something is seized, and you know when.

I cannot tell you what these machines will become. Nobody can. What I can tell you is what I have watched happen to people, and it does not look like the film. Nothing is seized. The daily tasks go one at a time, and each one is handed over, because the machine does it well and handing it over is easier. Nobody decides to hand over himself. What goes, goes the way the tasks did, one convenience at a time, and there is no day you could point to when it was taken.

If there is a takeover worth watching for, that is the one, and it is what this book is about. It is also the one you can do something about, because it runs on your own choices, and a choice can be made differently.

Anyone who has worked around technology knows the feeling of a tool arriving, settling in, and being replaced before you have finished learning it. And most people have noticed that the gap between one arrival and the next is not what it used to be. The thing has a name. People in the business call it technology turnover, and they use it the way you use a word everyone at the table already knows.

A number usually comes attached. People say the interval is about eighteen months, and they say it with some confidence, and the confidence traces back to Gordon Moore. In 1965 Moore, looking at the young business of putting transistors onto a single chip, observed that the number you could fit was doubling about every year. He revised it to two years in 1975. The eighteen-month figure is not his at all. It came from a colleague at Intel named Dave House, and what House was describing was not transistors but how fast the chips ran. Moore said, later and plainly, that he had never said eighteen months.

The number is a real number and it was measuring something real. It was measuring transistors on a piece of silicon. It was not measuring how long it takes a tool to reach an ordinary person's kitchen table, or how long that person has before the next one arrives, and it was never built to.