2022-08-18 #1 speed of evolution and speed-up of digitalisation of code

11:13AM Jan 19, 2023

Speakers:

Forrest

Keywords:

code

atoms

evolution

bias

dielectric constants

substrate

ai

thesis

talking

entities

plurality

snake

fact

represented

fox

abstracted

endemic

singular entity

digital representation

slow

This is excellent to see. But even before looking into all of this, I can point to a couple of things that would be particularly important for us to be mindful of. One is that there is a first price bias which is endemic to this. We just need to account for that, there's nothing we can do about it. But the idea or the association that evolution is slow is going to be a bias that's just going to fundamentally be in any communicative context that we occur. It won't matter what we do or how we present it or anything else, there will always be that bias that the word evolution will imply to just about everybody listening, that there's some sort of inherent slowness, or therefore some sort of inherent disadvantage.

So just be aware of that. That's not something we're going to overcome, it's just something we're going to have to live with. But when we look at this, first let's now ignore the bias and just look at it from a point of view of what is our current situation. One of the things that I think you already are aware of, is that atoms move slowly and electric fields, electromagnetism, i.e light, basically moves at the speed of light. I mean, in real wires and stuff like that, there's dielectric constants and stuff that slow it down a little bit. But basically, information moves stupendously fast relative to atoms.

When we're looking at evolution in the sense of machine process and machine intelligence, and so on and so forth, we're not talking about rearranging atoms in some sort of DNA-type code and through the processes of chemistry. We're actually talking about code as digitally stored and digitally represented. As a result, that is a very substantial speed-up in terms of just how code can move around. The other part too is that when you're making an example about the evolution of a snake versus your local fox, or lion, or something, code as represented in atoms is hard to move from one organism to another. But when we're talking about digital entities, it's going to be very quick to move from one organism to another. In other words, this is a place where the substrate argument, the orthogonality thesis is in our favor. To the degree that they believe in the orthogonality thesis, they're going to treat the notion of code as essentially being a digital construct, which is fine. I mean, I'm okay with that.

It isn't to say that the shape of the code won't be influenced by the substrate. But the fact of the code isn't necessarily going to be defined by some sort of underlying mechanical representation. It's going to be almost certainly some sort of digital representation, so it's already abstracted. As a result, there's no particular reason why you would have any barrier from the AI equivalent of a snake sharing its code with the AI equivalent of a fox. In fact, given that we can't make any assumptions about whether future AI is going to be a singular entity spanning the globe or multiple sub-entities of that, that all cooperates with some sort of community, it may be coordinated as some sort of hierarchical frame, or maybe, moreover, some sort of plurality of any kind.

Now they're mowing the lawn and they're coming right for me. Shit.