Intelligence 1 — What Makes Us Intelligent?

@tags:: #lit✍/🎧podcast/highlights
@links:: complexity, intelligence,
@ref:: Intelligence 1 — What Makes Us Intelligent?
@author:: Simplifying Complexity

=this.file.name

Book cover of "Intelligence 1 —  What Makes Us Intelligent?"

Reference

Notes

Quote

(highlight:: Defining intelligence by structure or output: an ongoing debate
Summary:
It immediately touches a very interesting set of issues, which are shall we define intelligence in terms of its structure, the elements, how they work together, their properties, their collective dynamics, or is it purely based on their outputs, on their behaviors? And this has been an ongoing debate from the very beginning.
Transcript:
Speaker 1
Shall we define intelligence in terms of its structure, the elements, how they work together, their properties, their collective dynamics, or is it purely based on their outputs, On their behaviors? And this has been an ongoing debate from the very beginning.)
- Time 0:02:02
- outputs, collective dynamics, behavior, structure, intelligence,

Quote

(highlight:: Our Biased Definition of Intelligence
Key takeaways:
• Evolution prioritizes motion over cognitive abilities.
• Cognitive neural resources are limited.
• The lens through which we view intelligence may need to be reevaluated.
Transcript:
Speaker 1
And the way I like to put this is evolution doesn't give a shit about chess or mathematics. It cares deeply about motion. And so the amount of, if you like, cognitive neural resources that we get to allocate to doing the things that we consider most praiseworthy and impressive are actually rather minimal. So it raises this whole question of, wait a second, are we looking at intelligence using the completely wrong lens here? Should we be considering the amount of computational power it takes? Or is it something else about uniqueness and human abilities and human declarative knowledge?)
- Time 0:07:19
-

Quote

(highlight:: Competence Without Comprehension: The Tension Between Talent and Intelligence
Key takeaways:
• Talent is often used to describe athletes instead of intelligence.
• Competence without comprehension is key to true talent in any domain.
• Expertise acquired through practice becomes effortless and looks effortless.
• The better someone is at something, the more effortless it appears, which is counter to how we typically associate intelligence.
Transcript:
Speaker 2
I loved it at presentation. You talked about, you know, when it comes to chess, we clearly see that as intelligent. But when we see a star basketball player, we don't use the word intelligent, we use the word talent.
Speaker 1
Exactly. Yeah, we hold the whole bigoted language around disparaging traits that we don't perform any better than other animals. So walking and running, you know, no, no, no, that doesn't require intelligence. I mean, because other animals do it better than as they're faster than us. So there's that whole thing. There's this very nice phrase that Daniel Dennett came up with in 2012 called competence without comprehension. And on the one hand, that seems a little bit like a reflex, something innate, right? But the key point is that true talent in any domain evolves towards or develops towards competence without comprehension. For example, if you're really good at mathematics, you don't have to think when you're about to integrate a function. It's almost a reflex. It's an expertise that you've acquired. It's become effortless. Whereas beginners in mathematics really have to labor and sweat over a problem. So it looks more like running. There's this very interesting fact that the better you are, the more effortless things become. And the more they look like characteristics that we typically do not associate with intelligence. So there's this weird tension between the easy and the difficult that seems to be part of the key to understanding this problem.)
- Time 0:07:54
-

Quote

(highlight:: Three Dimensions of Intelligence: Strategy, Encoding/Representation, and Inference
Key takeaways:
• Intelligence has three dimensions: strategy, encoding/representation, and inference.
• Intelligent tasks involve pursuing an objective using strategic thinking.
• Encoding or representation involves encoding properties of the natural world for strategic advantage.
• Inference involves deduction, induction, and abduction, and is the calculation on the stored memory.
• Different forms of life vary in their respective inferential representational strategic powers.
Transcript:
Speaker 1
So I think of intelligence as being this three-dimensional quality. On the one hand, there is this quality of strategy, which is most intelligent tasks involve pursuing an objective, winning a chess game, eating a bird, escaping from the lion. There's a strategic dimension. The other dimension to it, which interests me perhaps the most, is encoding or representation, which is encoding some property of the natural world that you can operate on a bit like Numbers, for example, would be a representation that you can add up that give you some strategic advantage in the world. And inference, which is the third and the inference is deduction, induction, abduction, the stuff that actually computers do well, the straight calculation on the stored memory, Which would be the representation. So you have encodings of the world, representations, you operate on them using inference to achieve an objective, the strategy. And when you think in those terms, you can really drill down how does an octopus encode the world? How does the bird? How does a human? How does a machine perform inference? What is the strategic power of a virus or a micro, which has very limited inference? So you can position different forms of life in a three-dimensional space where they each vary in their respective inferential representational strategic powers.)
- Time 0:13:47
-

Quote

(highlight:: Representation/Encoding: The Basis for Understanding the World and Solving Problems
Key takeaways:
• Representation is present in the brain and culture.
• Representation of cultural objects are used to convey ideas.
• Number systems are a fossil record of human ingenuity when it comes to representation.
• Binary number system is an easy number system for computers.
• Representation helps in solving hard problems easily.
Transcript:
Speaker 1
Representation goes all the way down into the brain and all the way out into culture. So there are ways, for example, that your visual system represents scenes in the visual world. It divides them into edges and edges that move and then into higher order features as you move back through the visual pathway into the visual cortex. So that's one kind of representation, right? There's another kind of representation, which is the representation of ideas in cultural objects, like on paper, which is I can represent number, I can represent a number of objects With a little glyph seven. And you immediately know what I'm talking about. And there are good representations and bad ones. For example, most of us were unfortunate enough to learn Roman numerals from here at school, right? Not quite sure. It was to read gravestones. That was, I guess, the purpose. But try and do mathematics, even basic arithmetic with Roman numerals. It's total waste of time. They don't have place-based numbers. They don't have a zero. It's ridiculous. So most people who operate with Roman numerals translate into the Indian Arabic system that we all use and they go back again. So that's a good example. Number systems and their evolution are a fossil record of human ingenuity when it comes to representation, which makes solving hard problems easy. Turns out, for example, for computers, the easy number system is binary because it maps naturally on to transistor states on and off. So those are very simple examples of representation. Of course, it gets more and more elaborate geometry, for example, the representation of non-neuclidean space, for example, on the surface of a sphere. These are all ways that help us think about problems that make very difficult problems relatively easy for us.)
- Time 0:20:32
-

Quote

(highlight:: Cultivating Collective Intelligence Through Representation/Encoding
Summary:
The concept of intelligence in culture and how we use culture has been an obsession of mine. And I call this X-bodymange, which is a process that I call X- bodymange. You can look at the history of chess through this lens, which I have, how the chessboard has changed. It used to be a game that was played with a die. Now it's a deterministic game. There are rules that have been introduced to make it richer.
Transcript:
Speaker 2
Representation really makes possible this concept of intelligence in culture and how we use culture.
Speaker 1
Yeah. So that's been an obsession of mine. And I call this X-bodymange, which is, I'll give a good example, a simple one, the map. We together collectively can draw a map of a territory. And that map is generated over a course of centuries, perhaps, as most maps of the world were. But you, then, as a single individual can look at that map and internalize it. And I can take that map and burn it. And you'll still possess either the whole map or some part of it. Okay. So your navigational strategic capability has been significantly amplified through the existence of an artifact that's been collectively constructed. And that back and forth between individual minds and brains, individual contributions to collective artifacts is the process that I call X-bodymange. And in fact, you can look at the history of chess through this lens, which I have, how the chessboard has changed. It used to be a game that was played with a die. And now it's a deterministic game. There are rules that have been introduced to make it richer, rather like mathematics, you know, like the, as you mentioned earlier, the invention of the calculus, it's like adding Another rule to the game. Interestingly, when we go to machine intelligence that I know that we're going to discuss, that whole aspect is neglected. It's fascinating.)
- Time 0:24:45
-

Quote

(highlight:: The Missing Link: The Lack of Representation/Encoding in AI and Machine Learning
Key takeaways:
• There was a fight early on in the history of intelligence between B.F. Skinner and Noam Chomsky.
• Skinner believed in strict behaviorism while Chomsky argued that there was an innate structure that supports language.
• AI is focused on reinforcement and neglects the importance of collective cultural artifacts for human intelligence.
• Our intelligence is mediated through language, mathematics, music, or shared artifacts.
Transcript:
Speaker 1
Early on in the history of intelligence, there was a big fight. And the fight was between B. F. Skinner and Noam Chomsky. And Skinner was a strident, strict behaviorist. He only believed in schedules of reinforcement. That if you formed a behavior that was good, it'd be reinforced. If it was a behavior that was bad, it would be punished. And that's all you needed. And Chomsky said, oh, hold on a minute. If you look at the way that kids learn languages, they're almost never rewarded or punished. There's a porosity of the stimulus. And they learn languages much too quickly for your theory to be right. And this was this resurgence of this representational view of language that there is an innate structure that supports it. It's not just reinforcement learning. And language is one of the best examples of an ex-bodyment of an ex-body to trade. Because you, neither you nor me nor anyone else, we know, invented the language we're speaking. It's sort of like the ultimate chessboard. It was co-constructed over tens of thousands of years by many people to give us a superpower narrative, what we're doing now, communication. And so if you look at AI, it's all about reinforcement. It's not about the construction of representations. It's not about language, a language of communication. And that, to me, is one of the great missing links in the whole enterprise of AI and machine learning that it doesn't sufficiently treat of collective cultural artifacts, which are Obviously the key to our intelligence. Everything that we do, that we value, is mediated either through language, mathematics, music, or shared artifacts. And so it's a fascinating deficit. And that's something that we'll come to discuss.)
- Time 0:26:16
-


dg-publish: true
created: 2024-07-01
modified: 2024-07-01
title: Intelligence 1 — What Makes Us Intelligent?
source: snipd

@tags:: #lit✍/🎧podcast/highlights
@links:: complexity, intelligence,
@ref:: Intelligence 1 — What Makes Us Intelligent?
@author:: Simplifying Complexity

=this.file.name

Book cover of "Intelligence 1 —  What Makes Us Intelligent?"

Reference

Notes

Quote

(highlight:: Defining intelligence by structure or output: an ongoing debate
Summary:
It immediately touches a very interesting set of issues, which are shall we define intelligence in terms of its structure, the elements, how they work together, their properties, their collective dynamics, or is it purely based on their outputs, on their behaviors? And this has been an ongoing debate from the very beginning.
Transcript:
Speaker 1
Shall we define intelligence in terms of its structure, the elements, how they work together, their properties, their collective dynamics, or is it purely based on their outputs, On their behaviors? And this has been an ongoing debate from the very beginning.)
- Time 0:02:02
- outputs, collective dynamics, behavior, structure, intelligence,

Quote

(highlight:: Our Biased Definition of Intelligence
Key takeaways:
• Evolution prioritizes motion over cognitive abilities.
• Cognitive neural resources are limited.
• The lens through which we view intelligence may need to be reevaluated.
Transcript:
Speaker 1
And the way I like to put this is evolution doesn't give a shit about chess or mathematics. It cares deeply about motion. And so the amount of, if you like, cognitive neural resources that we get to allocate to doing the things that we consider most praiseworthy and impressive are actually rather minimal. So it raises this whole question of, wait a second, are we looking at intelligence using the completely wrong lens here? Should we be considering the amount of computational power it takes? Or is it something else about uniqueness and human abilities and human declarative knowledge?)
- Time 0:07:19
-

Quote

(highlight:: Competence Without Comprehension: The Tension Between Talent and Intelligence
Key takeaways:
• Talent is often used to describe athletes instead of intelligence.
• Competence without comprehension is key to true talent in any domain.
• Expertise acquired through practice becomes effortless and looks effortless.
• The better someone is at something, the more effortless it appears, which is counter to how we typically associate intelligence.
Transcript:
Speaker 2
I loved it at presentation. You talked about, you know, when it comes to chess, we clearly see that as intelligent. But when we see a star basketball player, we don't use the word intelligent, we use the word talent.
Speaker 1
Exactly. Yeah, we hold the whole bigoted language around disparaging traits that we don't perform any better than other animals. So walking and running, you know, no, no, no, that doesn't require intelligence. I mean, because other animals do it better than as they're faster than us. So there's that whole thing. There's this very nice phrase that Daniel Dennett came up with in 2012 called competence without comprehension. And on the one hand, that seems a little bit like a reflex, something innate, right? But the key point is that true talent in any domain evolves towards or develops towards competence without comprehension. For example, if you're really good at mathematics, you don't have to think when you're about to integrate a function. It's almost a reflex. It's an expertise that you've acquired. It's become effortless. Whereas beginners in mathematics really have to labor and sweat over a problem. So it looks more like running. There's this very interesting fact that the better you are, the more effortless things become. And the more they look like characteristics that we typically do not associate with intelligence. So there's this weird tension between the easy and the difficult that seems to be part of the key to understanding this problem.)
- Time 0:07:54
-

Quote

(highlight:: Three Dimensions of Intelligence: Strategy, Encoding/Representation, and Inference
Key takeaways:
• Intelligence has three dimensions: strategy, encoding/representation, and inference.
• Intelligent tasks involve pursuing an objective using strategic thinking.
• Encoding or representation involves encoding properties of the natural world for strategic advantage.
• Inference involves deduction, induction, and abduction, and is the calculation on the stored memory.
• Different forms of life vary in their respective inferential representational strategic powers.
Transcript:
Speaker 1
So I think of intelligence as being this three-dimensional quality. On the one hand, there is this quality of strategy, which is most intelligent tasks involve pursuing an objective, winning a chess game, eating a bird, escaping from the lion. There's a strategic dimension. The other dimension to it, which interests me perhaps the most, is encoding or representation, which is encoding some property of the natural world that you can operate on a bit like Numbers, for example, would be a representation that you can add up that give you some strategic advantage in the world. And inference, which is the third and the inference is deduction, induction, abduction, the stuff that actually computers do well, the straight calculation on the stored memory, Which would be the representation. So you have encodings of the world, representations, you operate on them using inference to achieve an objective, the strategy. And when you think in those terms, you can really drill down how does an octopus encode the world? How does the bird? How does a human? How does a machine perform inference? What is the strategic power of a virus or a micro, which has very limited inference? So you can position different forms of life in a three-dimensional space where they each vary in their respective inferential representational strategic powers.)
- Time 0:13:47
-

Quote

(highlight:: Representation/Encoding: The Basis for Understanding the World and Solving Problems
Key takeaways:
• Representation is present in the brain and culture.
• Representation of cultural objects are used to convey ideas.
• Number systems are a fossil record of human ingenuity when it comes to representation.
• Binary number system is an easy number system for computers.
• Representation helps in solving hard problems easily.
Transcript:
Speaker 1
Representation goes all the way down into the brain and all the way out into culture. So there are ways, for example, that your visual system represents scenes in the visual world. It divides them into edges and edges that move and then into higher order features as you move back through the visual pathway into the visual cortex. So that's one kind of representation, right? There's another kind of representation, which is the representation of ideas in cultural objects, like on paper, which is I can represent number, I can represent a number of objects With a little glyph seven. And you immediately know what I'm talking about. And there are good representations and bad ones. For example, most of us were unfortunate enough to learn Roman numerals from here at school, right? Not quite sure. It was to read gravestones. That was, I guess, the purpose. But try and do mathematics, even basic arithmetic with Roman numerals. It's total waste of time. They don't have place-based numbers. They don't have a zero. It's ridiculous. So most people who operate with Roman numerals translate into the Indian Arabic system that we all use and they go back again. So that's a good example. Number systems and their evolution are a fossil record of human ingenuity when it comes to representation, which makes solving hard problems easy. Turns out, for example, for computers, the easy number system is binary because it maps naturally on to transistor states on and off. So those are very simple examples of representation. Of course, it gets more and more elaborate geometry, for example, the representation of non-neuclidean space, for example, on the surface of a sphere. These are all ways that help us think about problems that make very difficult problems relatively easy for us.)
- Time 0:20:32
-

Quote

(highlight:: Cultivating Collective Intelligence Through Representation/Encoding
Summary:
The concept of intelligence in culture and how we use culture has been an obsession of mine. And I call this X-bodymange, which is a process that I call X- bodymange. You can look at the history of chess through this lens, which I have, how the chessboard has changed. It used to be a game that was played with a die. Now it's a deterministic game. There are rules that have been introduced to make it richer.
Transcript:
Speaker 2
Representation really makes possible this concept of intelligence in culture and how we use culture.
Speaker 1
Yeah. So that's been an obsession of mine. And I call this X-bodymange, which is, I'll give a good example, a simple one, the map. We together collectively can draw a map of a territory. And that map is generated over a course of centuries, perhaps, as most maps of the world were. But you, then, as a single individual can look at that map and internalize it. And I can take that map and burn it. And you'll still possess either the whole map or some part of it. Okay. So your navigational strategic capability has been significantly amplified through the existence of an artifact that's been collectively constructed. And that back and forth between individual minds and brains, individual contributions to collective artifacts is the process that I call X-bodymange. And in fact, you can look at the history of chess through this lens, which I have, how the chessboard has changed. It used to be a game that was played with a die. And now it's a deterministic game. There are rules that have been introduced to make it richer, rather like mathematics, you know, like the, as you mentioned earlier, the invention of the calculus, it's like adding Another rule to the game. Interestingly, when we go to machine intelligence that I know that we're going to discuss, that whole aspect is neglected. It's fascinating.)
- Time 0:24:45
-

Quote

(highlight:: The Missing Link: The Lack of Representation/Encoding in AI and Machine Learning
Key takeaways:
• There was a fight early on in the history of intelligence between B.F. Skinner and Noam Chomsky.
• Skinner believed in strict behaviorism while Chomsky argued that there was an innate structure that supports language.
• AI is focused on reinforcement and neglects the importance of collective cultural artifacts for human intelligence.
• Our intelligence is mediated through language, mathematics, music, or shared artifacts.
Transcript:
Speaker 1
Early on in the history of intelligence, there was a big fight. And the fight was between B. F. Skinner and Noam Chomsky. And Skinner was a strident, strict behaviorist. He only believed in schedules of reinforcement. That if you formed a behavior that was good, it'd be reinforced. If it was a behavior that was bad, it would be punished. And that's all you needed. And Chomsky said, oh, hold on a minute. If you look at the way that kids learn languages, they're almost never rewarded or punished. There's a porosity of the stimulus. And they learn languages much too quickly for your theory to be right. And this was this resurgence of this representational view of language that there is an innate structure that supports it. It's not just reinforcement learning. And language is one of the best examples of an ex-bodyment of an ex-body to trade. Because you, neither you nor me nor anyone else, we know, invented the language we're speaking. It's sort of like the ultimate chessboard. It was co-constructed over tens of thousands of years by many people to give us a superpower narrative, what we're doing now, communication. And so if you look at AI, it's all about reinforcement. It's not about the construction of representations. It's not about language, a language of communication. And that, to me, is one of the great missing links in the whole enterprise of AI and machine learning that it doesn't sufficiently treat of collective cultural artifacts, which are Obviously the key to our intelligence. Everything that we do, that we value, is mediated either through language, mathematics, music, or shared artifacts. And so it's a fascinating deficit. And that's something that we'll come to discuss.)
- Time 0:26:16
-