How the World Got Wired · Epilogue — One Last Thing

About the Bubble (Again)

You read in this book about the dot-com bubble — the years between 1995 and 2000 when enormous amounts of money were poured into internet companies, most of which failed, but whose infrastructure ended up making the internet you grew up with possible.

You may live through something similar with AI. As I am writing this book, in 2026, an enormous amount of capital is flowing into AI companies and AI data centers. The pace of construction has no precedent. Some of those companies will, in five or ten years, be worth nothing. Some will become trillion-dollar businesses. Right now, no one — including the people putting up the money — can tell you in advance which will be which.

But you have seen this pattern, in this book, in chapter after chapter. Even when the financial bubble pops, the underlying technology and the infrastructure built during the boom do not disappear. The fiber cables laid in 1999 still carry the internet today. The data centers being built in 2025 will still be running in 2040, even if many of the companies that paid for them are not.

The need for computation that AI is creating is real. The capability of AI to do useful things is real. The infrastructure being built will outlast the financial cycle, the way it always does.

This is, by the way, why I find it hard to be a purely-doom person about all of this. The technology is genuinely useful. The capability gains are genuinely real. The benefits, distributed across billions of lives, are genuinely large. Even if the people building these systems get some things badly wrong — and they will, and some of those wrong things will hurt real people — the underlying tide of useful capability is moving in a direction that has the potential to be enormously good.

That doesn’t mean Hinton’s warnings are wrong. They might be right. The fact that something is mostly good doesn’t mean it can’t also produce serious harms, including, at the very edge of possibility, catastrophic ones. Both things can be true. Holding both is the work this book has been trying to teach you.

All 24 chapters of How the World Got Wired · Library · All courses