Sarah Walker
speaker
510 appearances
1 recordings
1 series
first heard Jun 2024
last heard Jun 2024
Sarah Walker’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
Yeah, so I think if you're thinking about this sort of thinking about living things as these giant causal structures or these objects that are deep in time or whatever language we end up using to describe it, it seems to me that consciousness is about The fact that we have a conscious experience is because we are these temporally extended objects.
So consciousness and the abstraction that we have in our minds is actually a manifestation of all the time that's rolled up in us. And it's just because we're so huge that we have this very large inner space. That we're experiencing. And it's also separated off from the rest of the world because we're this separate thread in time.
And so our consciousness is not exactly shared with anything else because nothing else occupies the same part of time that we occupy. But I can understand something about you maybe being conscious because you and I didn't separate. that far in the past in terms of our causal histories.
So in some sense, we can even share experiences with each other through language because of that sort of overlap in our structure.
It's not merely temporal separateness. It's about the depth in that time. So it's the reason that my conscious experience is not the same as yours is because we're separated in time. The fact that I have a conscious experience is because I'm an object that's super deep in time. So I'm huge in time.
And that means that there's a lot there that I am basically in some sense, a universe onto myself because my structure is so large relative to the amount of space that I occupy.
I think there's a horizon, and I don't know how to articulate this yet. It's a little bit like the horizon at the origin of life, where the space inside a particular structure becomes so large that it has some access to a space that's not... That doesn't feel as physical. It's almost like this idea of counterfactuals.
So I think like the past history of your horizon is just much larger than can be encompassed in like a small configuration of matter. So you can like pull this stuff into existence. This property is maybe a continuous property, but there's something really different about human level physical systems and human level ability. to understand reality.
I really love David Deutsch's conception of universal explainers, and that's related to the theory of universal computation. And I think there's some transition that happens there. But maybe to describe that a little bit better, what I can also say is what intelligence is in this framework. So you have these objects that are large in time,
they were selected to exist by constraining the possible space of objects to this particular, like all of the matter is funneled into this particular configuration of object over time. And so these objects arise through selection, but the more selection that you have embedded in you, the more possible selection you have on your future. And so the
selection and evolution we usually think about in the past sense, where selection happened in the past. But objects that are high density configurations of matter that have a lot of selection in them are also selecting agents in the universe. So they actually embody the physics of selection and they can select on possible futures.
And I guess what I'm saying with respect to consciousness and the experience we have is that there's something very deep about that structure and the nature of how we exist in that structure that has to do with how we're navigating that space and how we generate that space and how we continue to persist in that space.
I mean, I think they already do, but not in the way I hear popularly discussed. So there are obviously signatures of intelligence and... and a part of a ecosystem of intelligent systems. But I don't know that individually, you know, I would assign all the properties to them that people have.
It's a little like, so, you know, we talked about the history of eyes before and like how eyes scaled up into technological forms. And language has also had a really interesting history and got much more interesting, I think, once we started writing it down. And then, you know, inventing books and things.
But like, you know, every time that we started storing language in a new way, you know, we were kind of existentially traumatized by it.
So like, you know, the idea of written language was traumatic because it seemed like the dead were speaking to us even though they were deceased and books were traumatic because, you know, like suddenly there were lots of copies of this information available to everyone and it was going to somehow dilute it. And large language models are kind of interesting because they don't feel as static.
They're very dynamic. But if you think about language in the way I was describing before as language is this very large in time structure. And before it had been something that was distributed over human brains as a dynamic structure. And occasionally we store components of that structure. very large dynamic structure in books or in written language.
Now we can actually store the dynamics of that structure in a physical artifact, which is a large language model. And so I think about it almost like the evolution of genomes in some sense where, you know, there might've been like really primitive genes in the first living things and they didn't store a lot of information or they were like really messy things.
And then by the time you get to the Eukarya Excel, you have this really dynamic genetic architecture that's rewritable and has all of these different properties. And I think large language models are kind of like the genetic system for language in some sense, where it's allowing a sort of archiving that's highly dynamic.
And I think it's very paradoxical to us because obviously in human history, we haven't been used to conversing with anything that's not human. But now we can converse basically with a crystallization of human language in a computer.
Showing 381–400 of 510 · page 20 of 26
← Previous
Next →