Kenneth Stanley: belief

Transcript context

…So we in our paper, we don't term it SGD. We term it conventional SGD. And we do that on purpose because we're not trying to we're not a 100% sure if it's SGD by it's like if we just need to get rid of SGD or what the exact problem is. We're just saying the current paradigm, which encapsulates a fixed architecture, a fixed objective, and SGD as a target, as a as the thing that's doing the chasing of the target. That entire paradigm, there's something wrong with it because it's susceptible to shortcut learning. And people have known this for a long time. Like, I think Melanie Mitchell talks about this a lot, which is basically just Goodhart's law for representation learning. Right? It's basically just you can if you try to solve a task, you're gonna perfectly solve that task, but what you really wanted was a good representation of that task. And as our paper kinda shows, there's many different ways to solve the task. You can solve it with a bunch of heuristics and a bunch of if statements, or you can solve it with the right abstractions, and, there's the those are the 2 ways to solve it. Right? And what you pick, which 1 you end up, like, employing and using, it really matters for what you really want, which is adaptivity, generalization, creativity, out of distribution generalization, and continual learning, especially. And we can we're gonna talk about that a lot more later. But, yeah, I guess the point is we don't just care about the training loss. That's not like we train it on the training loss, but that's not what we care about. What we care about is the all the downstream stuff that we're gonna use it for later, which is much harder to quantify, much harder to formalize than just a training loss. Yeah. And let me add 1 thing. I think 1 of the really interesting things about the observation in this paper is it pokes a hole, I think, in a very deep assumption that we have that if the results are good, then what's underneath the hood is also good. It's a kind of assumption that's unstated most of the time. But it's clear here, when you look at the underlying representations in this paper, it's not necessarily the case. And so you can't just look at benchmark performance, is most of what we look at when we talk about is this model good and know that actually things are okay under the hood. Like Akash said, does have implication. Because you may think, well, who cares? If it does well in the benchmark, then it's good. What else matters? Why should I care? But as Akash says, there's all kinds of downstream implications if your underlying representation is terrible. And I'm sure we'll get into those. But so, just this idea that, like, you can't be confident just because things look good on the surface is thought provoking, think. It it leads to a lot of questions about what's really going on. 1 thing I really wanna add quickly to Ken's statement is I think at the end of the paper, we have a quote from or, like, something, like a quote or a paraphrase of Ken Ken's dad. And I think he says something along the lines of 2 mathematicians, they can both, like, ace, like, a math exam. 1 can go on to become, like, a great mathematician that discovers a lot of things in the field, and the other 1 can go on to discover nothing. So the test is just, like, it doesn't give you a picture of what we really care about, which is downstream, like, how they influence the field and how their research progress carries out.…

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