Adam Brown: belief

Transcript context

…So this is actually a good point to chat about your actual day job. So there's these open debates about the kind of reasoning that these LLMs do. Does it correspond to true reasoning, or is it something more procedural? And it sometimes gets into a definition game. But this is maybe a good way to test our intuitions here. The kind of thing that Einstein was doing, where you start off with some thought experiments, you start off with some seeming conceptual inconsistencies in existing models, and you trace them through to some beautiful unified theory at the end, and you make incredibly productive use of these intuition pumps, that kind of reasoning. How far are our AIs from that? I have heard it said, and I kind of agree with this, that maybe the very last thing that these systems will be able to do, these LLMs will be able to do, is, given the laws of physics as we understood them at the turn of the last century, invent general relativity from that. So I think that's probably the terminal step. And then once it can do that, if it can do that, then there won't be much else to do as far as humans are concerned. It's pretty extraordinary, I mean, particularly coming from a physics background in which progress is pretty slow, to come to the AI field and see progress being so extraordinarily rapid day by day, week by week, year by year. Looking at it, it certainly looks like these LLMs and these AI systems in some sense are just interpolators, but the level of abstraction at which they're interpolating keeps going up and up and up and we keep sort of riding up that chain of abstractions. And then, presumably, from a sufficiently elevated point of view, the invention of general relativity from Newtonian physics is just interpolation at some sufficiently grandiose level of abstraction. That perhaps tells us something about the nature of intelligence, human intelligence, as well as about these large language models. If you ask me how many years until we can do that, that is not totally clear, but in some sense, general relativity was the greatest leap that humanity ever made. And once we can do that, perhaps in 10 years, then we will have fully encompassed human intelligence. Will it be of the same character as what Einstein did? Clearly, there are many disanalogies between human intelligence in these large language models, but I think at the right level of abstraction, it may be the same. Do you see early examples of the kind of thing it was? Obviously not at that level of difficulty, but you just start off with, hey, here's something funny, go think about it for a while. Is there something especially impressive you see when you kind of run that kind of experiment at the moment?…

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