Transcription
[Music] Last December, Jeffrey Hinton was awarded the Nobel Prize for his pioneering work in machine learning, a major turning point on the road to artificial intelligence. Brook Silva Braga introduced us to this leading figure in AI back in 2023 and recently went back to visit him.
Good morning, Brooke.
Good morning.
When we first met Hinton, the world had just been introduced to OpenAI's chat GPT, triggering a kind of AI arms race. Hundreds of billions have been spent on AI in just the last two years. Hinton entered the field decades before it was cool. And now retired from Google, he has a unique independent perspective on how he got here and where we're headed.
This is a Nobel Prize. Last year, Jeffrey Hinton, for most of his life an outcast professor, was awoken by a call in the middle of the night. He was getting the Nobel Prize.
People dream of winning these things. And when you do win it, does it feel like you thought it might?
I never dreamt about winning one for physics. So, I don't know. I dreamt about winning one for figuring out how the brain works.
Yeah. But I didn't figure out how the brain works, but I want one anyway.
That's because Hinton's attempt to model the brain instead helped change the world. In 1986, he proposed using a neural network to predict the next word in a sequence, the foundational concept that today's large language models—it's an expert at everything—have built upon.
You believe then that we would get here?
Yes, but not this soon.
Cuz that was 40 years.
Yeah. I didn't think we'd get here in only 40 years.
Yeah. But 10 years ago, I didn't believe we'd get here.
Yeah. It happened fast.
Yeah.
That speed, Hinton says, means education and medicine will soon be transformed. Climate change could be solved. But mostly, the rapid progress really worries him.
The best way to understand it emotionally is we're like somebody who has this really cute tiger cub. It's just such a cute tiger cub. Unless you can be very sure that it's not going to want to kill you when it's grown up, you should worry. I'm kind of glad I'm 77.
Hinton predicts AI will make authoritarians more oppressive and hackers more effective. He now spreads his money across three banks. The exact odds of an AI apocalypse are unknowable, he says, but hazards this guess: a 10 to 20% risk AI will take over from humans.
People haven't got it yet. People haven't understood what's coming. I don't think there's a way of stopping it taking control if it wants to. The issue is can we design it in such a way that it never wants to take control, that it's always benevolent.
Those concerns have long been shared by other AI leaders: Google CEO Sundar Pichai: "It can be very harmful if deployed wrongly." Ex-AI's Elon Musk who continues to call for regulation: "It has the potential of civilizational destruction." Sam Altman, seen here before he became OpenAI's CEO: "I think AI will probably, like, most likely sort of lead to the end of the world."
But now, as these companies race each other and compete with China, Hinton worries they're foolishly, selfishly putting all of humanity at risk. If you look at what the big companies are doing right now, they're lobbying to get less AI regulation. There's hardly any regulation as it is, but they want less, um, because they want short-term profits.
Taking a stand against the establishment has been the hallmark of Hinton's life. When American AI funding required partnering with the Defense Department, he moved to Canada. When neural networks were laughed at as unworkable, he worked on them for a few decades more.
Is that a certain thing in a person?
Yeah. You have to be contrarian.
Yeah. You have to have a deep belief that everybody else could be doing things wrong and you could figure out how to do them right.
Do you have any idea where that came from?
My, my family partly. My father was like that. That's him there.
Hinton's legendary family tree includes not just his father, the prominent entomologist, but further back, George Bool, whose algebra innovations paved the way for computing. And George Everest, the surveyor who found the height of the world's tallest peak, then had it named after him. So, it hits there. Hinton's inheritance was a curious mechanic's mind.
Were you always interested in this kind of stuff, the way things work and how to fix them?
Absolutely. I loved it.
When one of our cameras fell, damaging a lens filter, Hinton wanted to fix it. But this kind of tinkering, was this desire important to your work or is this just a hobby?
No, this is a similar thing. When I would make neural net models on the computer, I would then tinker with them for a long time to find out how they behaved. And a lot of people didn't do much of that, but I loved tinkering with them.
Okay. I remember with Ilya, we used to watch it learning and we would have bets for like 25 cents on who could predict the new score best.
Ilya Sutskever, Hinton's most famous protégé, went on to be chief scientist at OpenAI. to just set up a large neural network, which is a large digital brain. In 2023, he was part of the group that pushed out CEO Sam Altman, reportedly because they didn't trust that Altman was prioritizing safety.
I was quite proud of him for firing Sam Altman, even though it was very naive.
Naive, Hinton says, because OpenAI employees were about to get millions of dollars that would be jeopardized by Altman's departure. Altman returned; Sutskever left. Hinton criticizes his former colleagues at Google more reluctantly, but says they're falling short, too.
Were you disappointed when Google went back on its promise not to support military uses of AI?
Very disappointed. But it's part of a pattern, Hinton says, adding Meta and XAI to the list of companies racing faster than they should. For example, the fraction of their computer time they spend on safety research should be a significant fraction, like a third. Right now, it's much, much less.
Hinton, now on the AI sidelines, says government regulation is needed, but doesn't expect it soon.
I'm curious if just in your normal day-to-day life, you despair. You fear for the future and assume it won't be so good.
I don't despair, but mainly because even I find it very hard to take it seriously. Ah, it's very hard to get your head around the fact that we're at this very, very special point in history where in a relatively short time everything might totally change at a change of a scale we've never seen before. Um, it's hard to absorb that emotionally.
We asked the AI labs mentioned in the piece how much of their compute is used for safety research. None of them gave us a specific number, but all have said safety is important and they support regulation in general, but they've mostly opposed regulations that have come up before lawmakers so far.
Did he indicate which sector the breach might technically happen?
Well, well, like he says, he's already worried about banks. He thinks banks are going to be a target. He has spread his money across different banks.
Oh, wow.
All right. It's a little scary, Brooke, but great reporting.
Thank you so much.
Did you get the lens fixed on the camera?
The lens is fixed.
All right, Nobel Prize winner in action.