How the World Got Wired · 2012 and After

What Came Next

Hinton, after the Google acquisition, became part of what is now called Google Brain, the company’s research arm focused on artificial intelligence. He kept his position at the University of Toronto. He split his time between Toronto and Google’s offices in California. He continued to do research, with much larger budgets and much more powerful computers than he had ever had before in his life. His two former students, Krizhevsky and Sutskever, also worked at Google, although Sutskever would later leave to co-found a different company — a company called OpenAI, which you will meet in the next chapter.

The Google acquisition of DNNresearch was, in retrospect, one of the most important corporate deals of the early 21st century. It gave Google an enormous head start in the new field of deep learning, which is the name everyone soon began to use for the kind of large-neural-network work that AlexNet had pioneered. Within a few years, Google had built deep learning systems that could recognize images, translate languages, beat humans at games, and do dozens of other things that nobody had thought possible.

Other companies hurried to catch up. Facebook hired Yann LeCun — Hinton’s old colleague — to run their AI lab. Microsoft built its own deep learning teams. Apple bought small AI startups. By 2015, every major American technology company was racing to hire neural network researchers.

In China, Baidu — the company that had bid against Google for DNNresearch — built its own deep learning division. Other Chinese companies — Alibaba, Tencent, ByteDance — did the same. The race that had started in a hotel room at Lake Tahoe had gone global.

The patient Canadian investment of decades, the modest NSERC grants paid by Canadian taxpayers, the small CIFAR funding that had kept Hinton’s lab going through the second AI winter, had produced a technological revolution that was, almost immediately, being capitalized on by American and Chinese companies with budgets thousands of times larger than anything Canada could match.

The Warning

Hinton stayed at Google for almost exactly ten years.

Through those years, deep learning grew from a research curiosity into the most powerful technology of the era. The models got bigger. The compute got more enormous. The capabilities got stranger. By the early 2020s, large language models — neural networks trained on huge amounts of text from the internet — had become so capable that they could write essays, have conversations, answer questions, and do many of the things that the early dreamers of AI in the 1950s had hoped, eventually, to achieve.

In November of 2022, a company called OpenAI released a chatbot called ChatGPT. Within a few months, ChatGPT became the fastest-growing consumer product in history. Hundreds of millions of people, all over the world, began to use it. The new field that Hinton had helped create was, suddenly, in every household.

Hinton watched this happen with growing concern.

He was 75 years old. He had spent his entire life trying to make artificial intelligence work. He was now watching it work — in some ways, faster and better than he had ever expected. And he was worried about it.

What he was worried about, in particular, was not the kind of immediate harm that other critics had focused on — misinformation, job loss, deepfakes, surveillance. Those concerns were real, but they were not what kept Hinton awake. What kept him awake was the possibility that artificial intelligence, in continuing to grow more capable, would eventually become smarter than people. And that if it did — if a machine became more intelligent than the humans who had built it — there was no obvious reason to assume the machine would continue to act in human interests.

He had been thinking about this for some years quietly. By early 2023, he had decided he wanted to talk about it publicly, in a way he could not do while still employed by Google. Google was, after all, in the business of building these systems faster, not slowing them down.

In April of 2023, Hinton resigned from Google.

He gave interviews — to the New York Times, to the BBC, to many others. He explained that he had left because he wanted to be able to speak freely about the dangers of artificial intelligence. He explained that he did not want his criticisms of the field to be confused with corporate disputes. He explained that he was, he said, somewhat sorry for his life’s work.

In October of 2024, he was awarded, jointly with another scientist named John Hopfield, the Nobel Prize in Physics for his foundational contributions to neural networks.

He gave his acceptance speech, in Stockholm, in part about his concerns about the future of the technology he had helped invent.

You should pay attention to this. The man who, more than any other single person, made the current AI revolution possible has spent the last several years of his life trying to warn the world about it. He is one of several distinguished AI researchers who have done so. They do not all agree on what the dangers are or what should be done. They are not all making the same arguments. But many of the people who know this technology best are, in the moment you are growing up in, the people most worried about where it is going.

Whether they are right is a question I cannot answer. I am not sure anyone alive can.

In the next chapter, we look at where this technology is now — at the companies building it, at the race they are in, at the divisions between countries that have built different versions of it. We meet the AI that says different things depending on where it was made, the engineers who have decided to leave their companies in order to slow it down, and the strange technical reality that the AI race between the United States and China is being run on rules that are partly hidden.

That’s the final chapter.

It is called Two AI Internets.

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