Lex Fridman: belief

Transcript context

…I’m just going to do it once more. I think it’s a little like, let’s say it was like 1995 and Moore’s law is making the computers faster. And for some reason there had been this verbal tick that everyone was like, “Well, someday we’re going to have supercomputers. And supercomputers are going to be able to do all these things that … Once we have supercomputers, we’ll be able to sequence the genome, we’ll be able to do other things.” And so. One, it’s true, the computers are getting faster and as they get faster, they’re going to be able to do all these great things. But there’s, there’s no discrete point at which you had a supercomputer and previous computers were no. “Supercomputer” is a term we use, but it’s a vague term to just describe computers that are faster than what we have today. There’s no point at which you pass the threshold and you’re like, “Oh, my God! We’re doing a totally new type of computation and new … And so I feel that way about AGI. There’s just a smooth exponential. And if by AGI you mean AI is getting better and better, and gradually it’s going to do more and more of what humans do until it’s going to be smarter than humans, and then it’s going to get smarter even from there, then yes, I believe in AGI. But, if AGI is some discrete or separate thing, which is the way people often talk about it, then it’s a meaningless buzzword. To me, it’s just a platonic form of a powerful AI, exactly how you define it. You define it very nicely, so on the intelligence axis, it’s just on pure intelligence, it’s smarter than a Nobel Prize winner as you describe across most relevant disciplines. So okay, that’s just intelligence. So it’s both in creativity and be able to generate new ideas, all that kind of stuff in every discipline, Nobel Prize winner in their prime. It can use every modality, so this is self-explanatory, but just operate across all the modalities of the world. It can go off for many hours, days and weeks to do tasks and do its own detailed planning and only ask you help when it’s needed. This is actually interesting. I think in the essay you said … Again, it’s a bet that it’s not going to be embodied, but it can control embodied tools. So it can control tools, robots, laboratory equipment., the resource used to train it can then be repurposed to run millions of copies of it, and each of those copies would be independent that could do their own independent work. So you can do the cloning of the intelligence systems. Yeah. Yeah. You might imagine from outside the field that there’s only one of these, right? You’ve only made one. But the truth is that the scale up is very quick. We do this today,. We make a model, and then we deploy thousands, maybe tens of thousands of instances of it. I think by the time, certainly within two to three years, whether we have these super powerful AIs or not, clusters are going to get to the size where you’ll be able to deploy millions of these. And they’ll be faster than humans. And so, if your picture is, “Oh, we’ll have one and it’ll take a while to make them,” my point there was, no. Actually you have millions of them right away.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.