Tim Scarfe: uncertainty

Transcript context

…1 thing that's really challenging is if we were to actually lay out what's the data richness of comparative psychology studies across species, If you put that on a whiteboard and looked at it, you would realize we have so little data, on what intellectual capacities different animals in fact have. For example, the lamprey fish, which is sort of the canonical animal that's used as a model organism for the first vertebrates because it's the, of all vertebrates alive today, it's 1 of our most distant vertebrate cousins. To my knowledge, there are absolutely no studies examining the map based navigation of the lamprey fish. So we'd have no idea if it's in fact capable of recognizing things in 3 d space. Now when we look at other vertebrates like teleost fish, it seems like they're eminently capable of doing that. We look at lizards, they're eminently capable of doing that. So we sort of infer, it seems likely the first vertebrates were able to do this. We know the brain structures from which it emerges and and reptiles and telios fish are present in the lamprey. So we we sort of back into an inference that okay, well probably the lamprey fish can do that. But this is all sort of, in some sense, guessing and trying to put the pieces together from very little information. So I think that's 1 challenging aspect to reconcile. The other 1 that's really hard is in neuroscience, there's a lot of really interesting ideas about how the brain might work that have not really been tested in the wild from an AI perspective. And then there's a lot of AI systems that work really well that have diverged substantially from at least evidence what the evidence suggests is how brains work. So how do you bridge the gap between these 2 things, I think is a really fascinating space to operate in, which is like what can we learn about the brain, if anything, from the success of transformers, as an example? What can we learn, if anything, from the success of generative models in general? What can we learn from the success of and failures of modern reinforcement learning? I mean, in some ways, reinforcement learning has been a success. In other ways, it's really fallen short of what a lot of people hoped it would be. So I think the gap between neuroscience and AI is still a challenging 1 to bridge in a lot of ways. For example, Carl Ferson has all these incredible ideas in active inference. In 100 years, will we look back on this and be like Carl Ferson was onto something? If you look at the AI systems today, there's very little usage of active inference principles working in practice. So that could mean that the ideas don't have legs, or it could mean that there's a breakthrough behind the corner where we're actually missing some of the key principles that he's devising. And these are you know, questions we don't have the answers to. I think there might possibly be some breakthroughs around the corner. I I don't know if you know, but I'm I'm Carl Friston's personal publicist. I I do all of his stuff. I I I probably interviewed him more than anyone else, but I love I love great friend. He's an amazing guy. Yeah. He's an amazing man.…

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