Couldn't be more excited to be back with Geoff. Last time we spoke, I promised you you would get the Nobel Prize last year. You thought it was a very, very low probability. Did you tell the Nobel Committee that as well?
Actually, when they called up, I didn't believe them. I thought it was a prank. Nobody had mentioned that they'd nominated me. There isn't a Nobel Prize for computer science, and actually, I went around for several days after that doing a little Bayesian calculation. That is: what's the probability that a theoretical psychologist hiding in computer science will get the Nobel Prize in physics? Well, maybe one in two million. Now, what's the probability if this is my dream that I get the Nobel Prize in physics? Well, maybe one in two. So that means it's a million times more likely it's a dream than reality. And I remember when I was a kid, I used to dream about flying. Most kids do, I think. And I would occasionally have dreams where I remembered that last time I could fly, it turned out to be a dream. But this time it wasn't. And so I was actually worried for several days after I got it that it might be a dream. And I contented myself with one thing: if it was a dream, I would wake up, and that nightmare about Trump being president wouldn't be true.
Unfortunately, you won the Nobel Prize, and Trump was the president.
Maybe he'd make the trade, actually. We should take this idea seriously. Well, he obviously deserves one. But you've also been spanning the ideas from cognitive science to computer science. And one thing I find is quite fascinating is at some point, sort of human intelligence reached diminishing returns. We didn't get that much more intelligent than that. We could have just hill-climbed for, you know, a couple more million years and become increasingly intelligent. Do you think we're sort of intelligent enough, and it doesn't make sense to make humans more intelligent?
My guess is that we've got to a point in intelligence where we're going to create something else that can get intelligent much faster. So if you look at evolution, you start off with things just doing evolution. Then they sort of get development, and development has a much—has a 20-minute cycle time as opposed to evolution, which has a lifetime cycle time. And then after you've got development, you then get learning. And development has a 20-minute cycle time, whereas learning has a kind of one-second cycle time. So what happens is we keep developing faster and faster in the loops, and now what we've developed is AIs, which I think are the faster in a loop, and they're going to get much smarter.
Where do you think this will take you? Do you think we will see sort of 300 IQ models, and they would be vastly more useful? Or do you think it's more parallelizing and being able to think a million thoughts in parallel and pick the best ones? Do you think the intelligence in itself or the parallelization is the most useful thing?
Both, I think. So if you look at Go or chess, they've got the 300 IQ there already, right? No person will ever beat them again, at least maybe one game. But I think it's going to get much more intelligent than us, that's my guess. So I think today it was announced that, by the mathematicians, that it proved one of Erdős' theorems, and it wasn't something that it found somewhere in the ancient literature. It was something actually figured out in a quite clever way using a different branch of mathematics. So I think that's the sign that within mathematics, at least, it may well get much smarter than people because it's a closed system. So just as it did with chess and Go, it can just make conjectures, see if it can prove them, and just keep playing around like that. And by playing like that, it'll get smarter and smarter. And I think it may well outstrip mathematicians fairly soon, like in the next 10 years. And I'm surprised we haven't seen more examples like this yet.
It seems intuitive to me that if you had all of the world's knowledge in your head at the same time, through analogy, you should be able to discover quite a lot of new knowledge. What do you think is shifting, and do you think this is sort of a trend we should expect now with a lot of new knowledge being generated?
So I think if you just use language to train, it's amazing you can deal with spatial things at all. I mean, it's a weird fact. If you just train on language, you do understand quite a lot about space just from language. But obviously, you understand more about space if you can pick up objects and move them around and feel inertias and things. So I think that's one deficit that the models have had until recently. People have that. I think there's still a lot about people we don't understand, and there's probably still a lot of improvements we can make in the AIs. But I'm a materialist all the way through, so I don't think there's anything about people that we won't be able to get in AIs.
I would like to say I listened to the last talk, and I quite like people. And so, as we make these models increasingly better, what do you think now is the delta between Albert Einstein and these models? Is there something fundamental about these models that doesn't allow them to just do these thought experiments and generalize fantastically well from those? Or should we start to expect to see some of these quite clever thought experiments that then extract into a range of domains?
I think in the longer run, we will see that. Maybe not in the next few years, but if you think about the next 20 years, I think we'll be seeing things like that. I think if you look at human history, recent human history, there was the Copernican Revolution, and it took people a long time to accept that the Earth wasn't the center of the universe. Particularly, religions didn't like that at all. Then there was the Darwinian Revolution. It took people a long time to accept that we were animals. We're very special animals because we've got language and we got very big brains, but we are animals. I think we've got a new revolution coming when we're not the only beings around. I think we're creating beings, and it's going to take people a very long time to become happy to say we're creating beings. Right now, people are reacting just like they did with Copernicus and with Darwin and saying, 'No, no, no, no, no. I don't want—I mean, I don't believe that, that's crazy.' There's something really special about people. I mean, I think people are very special to other people, but I don't think there's anything about us that the AIs won't get in the end.
Benjamín mentioned some of the work on AlphaGo, where eventually they went from AlphaGo to AlphaZero, and they sort of skipped the pre-training step and went all in on purist reinforcement learning. Do you think that makes sense to do more generally, as we shift the training paradigms of these models to be more purist reinforcement learning? Or do you think inherently it's very useful to have this pre-training?
The pre-training makes things much more efficient. I think in closed worlds like mathematics or a game, that approach will work. And I think if you look at language models now, the language models now are like the early Go models were. And the early Go models—you've got a neural net to try and model the intuition of a Go player by getting it to try and copy the moves a Go player would make. So the database of all the moves experts made, you tried to get your neural net to pick the right move. And that's going to be limited by the data you've got from the experts and by the skill of the experts. As soon as they stopped doing that—they use that to initialize it, but as soon as they stopped doing that and started doing Monte Carlo rollout so they get their own training data, the things then took off. The equivalent for language models is going to be, you have some beliefs, you do a bit of reasoning with those beliefs, you arrive at a new thing you should believe, but you don't believe it. And now you've got an inconsistency. And so either the way you did the reasoning was wrong, and reinforcement learning can fix that, or you should change the premises, or you should change the conclusion. And so now you've got a system that can improve with no extra data. It can generate these internal inconsistencies, and I think that means these language models can get hugely smarter without a lot more data. And Demis thinks the same thing. So I assume Gemini is already doing stuff like that.
And as these models become better, do you think that there will be a compounding advantage? We're already seeing this to some extent, where the models get better, they use the models to improve the models, and it also has these sort of side effects, with the best talents joining the labs and so on. Do you think we'll see that sort of takeoff where being six months behind is not that valuable, and there will be one lab that compounds this all the way? Or do you think it will be equally distributed across four or five labs?
So, about a year ago, Eric Schmidt was talking about this a lot, about how being a year behind was no good because if it takes off, you'll just be left in the dust. I just don't know. I really sort of feel I don't know. That's a possibility. I hope it doesn't happen like that. My gut feeling is it probably won't happen like that. We won't get one thing that just gets much better than all the others very fast. They'll all get better. They'll all start using the AI to come up with new ideas for the architecture and stuff. So you'll get this recursive loop. I really just don't know the answer.
