Modern AI Was Invented by This Genius

Ashlee Vance:

Hinton has been working to make computers learn similar to humans for about 40 years. Almost everyone dismissed the idea as foolish, or at the very least, useless, until it completely transformed the field.

Geoffrey Hinton:

Google believes this is the future of the company. Amazon believes it to be the company's future. Apple believes it to be the company's future. My own department thinks it's just probably nonsense and we shouldn't be doing any more of it. I talked everybody into it except my own department.

Ashlee Vance:

You grew up in the UK and came from a pretty prominent family that included infamous economists and mathematicians, so I was interested in learning about your childhood.

Geoffrey Hinton:

Yes, there was quite a lot of pressure. I believe I understood I would need a PhD by the time I was around seven years old.

Ashlee Vance:

Did you rebel against that or did you go along with it?

Geoffery Hinton:

I dropped out every so often. For a period, I trained as a carpenter.

Ashlee Vance:

Geoff Hinton had an obsession with the idea of understanding how the mind functions relatively early in life. He began by getting into physiology and the anatomy of how the brain functioned. He first studied psychology before deciding to focus more on a computer science approach to understanding the brain and entered the field of artificial intelligence.

Geoffrey Hinton:

In what I believe is that building a brain would be the best way to understand such a complex machine. You could assume that you understand cars if you were to simply look at them. Suddenly realizing that certain components must be inside a car in order for it to function when building one is like discovering well there's this stuff that has to go under the hood, otherwise it doesn't work.

Ashlee Vance:

Frank Rosenblatt, a researcher in artificial intelligence, particularly, motivated Geoff as he was starting to think about these ideas.

Cade Metz:

A neural network, or "perceptron," was created by Rosenblatt in the late 1950s. It was a computer system that would imitate the brain.

Suzanne Gildert:

The fundamental concept is a network of neuron-like cells. These little computers are actually modeled after the way the human brain processes information. They actually learn by utilizing the incoming data from their sensors, just like we do. So that the neural network may gradually learn to make decisions.

Ashlee Vance:

Rosenblatt believed that by feeding a neural network a large amount of data, such as images of men and women, the network would eventually learn how to distinguish between the sexes, just like humans do. There was just one problem, it didn't work very well.

Cade Metz:

The capabilities of Rosenblatt's neural network, which consisted of a single layer of neurons, were limited. And in the late 1960s, one of his colleagues published a book that exposed these limitations, which in a way put the entire field of research on hold. No one wanted to work in this industry for at least ten years. They believed it would never succeed.

Geoffrey Hinton:

To me, it was just evident that this was the right path to take. Since the brain is a large neural network, it means that things like this must be possible since they function in our brains. There's just never any uncertainty over that.

Ashlee Vance:

What kept you motivated to continue working on this while everyone else gave up? Just the fact that you believed it to be the right path to take?

Geoffrey Hinton:

No, that everyone else was wrong.

Ashlee Vance:

Hinton decides he has a theory about how these neural networks might operate and he will pursue it no matter what. He spends some time moving between research centers in the US. He eventually becomes frustrated by the fact that the Defense Department provides funding for the majority of them and begins looking for alternative options.

Geoffrey Hinton:

I didn't want to accept funding from the Defense Department. I didn't really enjoy the concept of this stuff being used for purposes that I think were not good.

Ashlee Vance:

He suddenly hears that Canada might be interested in funding artificial intelligence.

Geoffrey Hinton:

And that was very attractive that I could go off to this civilized town and just get on with it. I chose to attend the University of Toronto, where in the mid 1980s we learned how to create more intricate neural networks that could address issues that the simpler ones couldn't.

Cade Metz:

A deep neural network with multiple layers was created by him and his partners, and it soon began to function in many different ways. In the late 1980s, a man by the name of Dean Pomerleau created a self-driving car that operated on public roads using a neural network. A system created by Yan LeCun in the 1990s that could read handwritten numbers was eventually used commercially, but once more they ran into a wall.

Geoffrey Hinton:

We didn't have enough data or computing power, so it didn't quite work as well as it could have. And many involved in AI and computer science have concluded that neural networks are just wishful thinking. Therefore, it was a huge disappointment.

Ashlee Vance:

Geoff was one of only a handful of individuals on the earth who continued to work on this technology through the 1990s and into the 2000s. When he would attend scholarly conferences, he would be banished to the back rooms. He was handled quite harshly.

Ashlee Vance:

Was there ever a point when you doubted your ability to succeed and believed this was never going to work?

Geoffrey Hinton:

I mean, there were many times when I thought, I'm not going to make this work.

Ashlee Vance:

But Geoff was consumed by this and couldn't stop. He just kept pursuing the idea that computers could learn until about 2006, when the world catches up to Hinton's ideas.

Geoffrey Hinton:

Computers were now a lot faster and now it's behaving like I thought it would behave in the mid 80s -- is solving everything.

Ashlee Vance:

Hinton algorithms received a tremendous boost with the introduction of ultra-fast computers and the massive amount of data generated on the internet. Computers all of a sudden had the ability to identify objects in images, then recognize speech, and translate between languages. In 2012, phrases like "neural nets" and "machine learning" began to appear on the New York Times front page.

Ashlee Vance:

You have to go through all these years, and then suddenly, in a matter of months, it simply takes off, and you feel as though the world has at last realized your vision.

Geoffrey Hinton:

It was sort of a relief that people finally came to their senses.

Ashlee Vance:

This was certainly a time of redemption for Hinton after years of struggle, and for Canada it meant something much more significant. The nation gained prominence as an AI superpower thanks to Hinton and his students. something that neither a person nor a machine could have foreseen.