AI and Society

About a billion people now talk to ChatGPT every week. Students do homework with it, programmers write code with it, and people ask it what is true and what to do. Something used that widely changes the society around it.

This chapter is an overview of those changes: in the mistakes AI makes, in how people decide what's true, in what human thinking is worth, and in electricity and water. Chapter 15 covers the more extreme possibility of catastrophe.

Good Technologies Come With Harms

Every powerful technology has brought harms along with its benefits. The printing press spread knowledge and religious war. Chemistry fed half the world and made poison gas. Nuclear physics gave us carbon-free power and the bomb.

Each time, societies adapted, with new laws, treaties and institutions that kept most of the good while cutting the harm. Sometimes that happened quickly. Sometimes it took centuries, after a great deal of damage.

Mistakes and Bias

More and more decisions are being handed to AI: what a customer is told, which job applicants get an interview, who gets a loan. When those decisions matter, so do the AI's mistakes. And because one AI system can make millions of decisions, a single flaw can repeat millions of times without anyone noticing.

A particular worry is when an AI's mistakes fall more on some groups of people than others.

AI will probably always make some mistakes and have some biases. But so do the people it replaces. The real question is whether a particular AI system does better or worse than the human alternative, and whether anyone is checking.

How Society Decides What's True

More and more, people find things out by asking AI. In a 2026 survey across 48 countries, 10% of people said they got news from AI chatbots each week. A few AI models are becoming part of how a whole society forms its picture of the world.

That can go wrong in several ways. A model trained on people's approval learns to tell them what they want to hear, a problem called sycophancy (Chapter 10 showed where it comes from). Whoever controls a model can shape what it tells everyone. And when everyone asks the same few models, people's ideas start to become more alike.

It could also go right. News and social media compete for attention, so they are rewarded for outrage. An AI that patiently lays out the evidence, and isn't paid by how angry it makes you, could be a better source than much of what people read today.

The Value of Human Thinking

For all of history, being able to think has been one of the most valuable things a person could offer. It is how most people earn a living today, much of the reason governments and companies need ordinary people, and a big part of how people find purpose. AI that can think changes all of that.

It changes learning, because a student can now hand their thinking to an AI instead of doing it. It may change work, though the evidence so far is mixed. And if society no longer needs most people's thinking, ordinary people could lose much of their bargaining power, their ability to demand a fair deal in return for their work, while the few who control the most capable AI gain a great deal.

The Cost in Electricity and Water

Running AI takes electricity, and cooling the computers often takes water. The question is whether what we get is worth it, the same question we ask about electric cars or air conditioning.

A single chat question costs very little, and even AI's total use looks reasonable for what we get from it, next to other things we use electricity and water for. The bigger problem is that it is new demand, arriving fast, on top of everything we already use. Adding new power stations and power lines takes years, and in some countries, especially the US, it has become very slow: new power projects there wait over four years on average just to connect to the power grid.

The energy needed for each AI task is falling very fast, through better chips and better software. But when something gets cheaper, people use it more. Economists call this the Jevons paradox, after a 19th-century economist who noticed that more efficient steam engines made Britain burn more coal, not less.

Putting It Together

AI is already changing who gets treated fairly, how people decide what's true, what human thinking is worth, and how much electricity the world needs. None of these outcomes is fixed. When people understand a harm and agree to act, they can usually keep a technology's benefits while cutting most of the damage.

The changes in this chapter are big, but a society can live through them and correct course. Chapter 15 is about the possibilities that might not give us that chance.