EPISODE TRANSCRIPT
How the Body Shapes the Way We Think: A New View of Intelligence (Pfeifer & Bongard)
Automatically transcribed from the episode audio. Some words and names may be inaccurate.
Listen to the episodeDownload textDownload captions
Imagine you're just sitting in a sterile, brightly lit laboratory. You know the kind. Oh, absolutely. The classic sci -fi setup. Yeah, exactly. And you have all these wires glued to your scalp, reading the electrical storms inside your brain, plus more wires taped to your forearm, just like measuring every tiny flex of your muscles. Sounds incredibly uncomfortable, to be honest. Right. But the researcher gives you one incredibly simple instruction. Just watch this highly precise clock on the wall. And whenever you feel the spontaneous, conscious urge to do so, flick your finger. Just flick it. Just flick it. That's it. And note the exact millisecond on the clock when you made that conscious choice. I mean, it sounds like the most basic demonstration of human free will possible. You're the boss. You make a decision.
You issue a command. And your body obeys. Exactly. That is the assumption we literally all live by. But. When the neuroscientist Benjamin LaBette ran this exact experiment back in 1983, I think? Yep, 1983. Yeah, the machines told a profoundly disturbing story. The sensors showed that the brain activity preparing for the movement, like the engine revving up to flick the finger, started more than half a second before the physical movement actually happened. Which honestly makes sense mechanically. Right. But here's the part that totally shattered our understanding of the mind.
Yeah, this is the crazy part. Preparatory brain activity started about 300 milliseconds before the person was even consciously aware they wanted to move. 300 milliseconds. So the brain was already initiating the action before the conscious mind had even decided to do it. Exactly. The conscious experience of willing the movement was essentially just a delayed echo. It was playing catch up. Wow. Your conscious mind was basically taking credit for an action that a deeper, darker, physically grounded system had already set in motion. Okay.
Let's unpack this. Because that completely dismantles the illusion that you or I are the conscious CEO of our bodies, sitting in some pristine mental boardroom issuing orders to our limbs. It really does. It flips everything upside down. Right. And today, we are going to dismantle that illusion entirely. We're taking a deep dive into a massive paradigm shift in how we understand intelligence itself. We are being guided by an incredibly dense groundbreaking book called How the Body Shapes the Way We Think. A New View of Intelligence by Rolf Pfeiffer and Josh Bongard. That's the one. In our mission today is to show you why the brain is not a lonely supercomputer. We're going to explore how your physical body, its squishiness, its shape, its limitations, is an absolute non -negotiable prerequisite for you to be able to think at all. The stakes of this paradigm shift are just enormous because for decades, scientists, roboticists, and well, philosophers chased a model of intelligence that was fundamentally flawed. Because they ignored the body, right?
Totally. We built early artificial intelligence on the assumption that the mind was the supreme commander and the body was just a mindless meat puppet. A meat puppet. I love that, but it's also kind of terrifying. It is. And because of that blind spot, we completely fail to understand how human beings actually move, how infants learn language, and how our memories actually work. By the end of this conversation, you are going to see your own mind -body connection in a completely alien way. Well, we have to figure out where this profound misunderstanding even came from first. Like, why do we all naturally feel like we're a software program trapped inside a hardware shell? For that, we have to lay the historical blame mostly at the feet of the 17th century French philosopher René Descartes. Descartes.
Right. Of course. He was a brilliant thinker, obviously, but he gave us the phrase, I think therefore I am. Right. Which codified what we now call the Cartesian heritage. Exactly. It's the ultimate dualism. Descartes basically argued that the mind, you know, the ethereal, rational, thinking essence of a human, and the physical body were two completely separate substances. Two different things entirely. Yeah. To him, the body was just a mechanical automaton full of pumps and levers, and the mind was a divine, non -physical ghost haunting that machine. I mean, it is a very seductive idea, you know?
It feeds our ego. We love feeling like our rationality is completely separate from the messy, sweaty, gravity -bound reality of the physical world. Oh, it was so seductive that it became the invisible bedrock of Western science for centuries. And when the computer age dawned in the mid -20th century, that Cartesian dualism was mapped directly onto the new technology. The brain became the software and the body became the interchangeable hardware. Right, which culminated in that massive initiative in 1956, the Dartmouth Conference. Yes, the conference that officially launched what philosopher John Hauglin later dubbed GoFi, good old fashioned artificial intelligence. Good old GoFi. And their entire approach was based on this top -down brain as CEO model, right?
They believed intelligence was simply about processing abstract symbols. That was the core tenet. The founding fathers of AI looked at human intelligence and just assumed its absolute highest peak was abstract logic. Like math and chess. Exactly. They thought, well, if we can just build a machine that manipulates logical symbols, like if A equals B and B equals C, then A equals C, we will have created a synthetic mind. And to be fair to them, they did have some massive early victories that probably inflated their egos quite a bit. Oh, definitely. I mean, they built programs that could solve complex algebra and prove mathematical theorems. They eventually built machines that could play championship level chess. So from the outside looking in, it really looked like they had cracked the code of intelligence. They cracked a very specific type of specialized logic.
But then they hit an unbelievable paradox. Right. Because the real world is messy. Exactly. The classical AI was phenomenally good at things human beings consider incredibly hard, like chess. But it completely failed at things a human toddler finds effortlessly easy. Like just walking across a messy living room without falling over. Or recognizing a mother's face in a crowd. Or just reaching out and picking up a fragile coffee cup without crushing it into dust. Why was walking such a nightmare for these million -dollar supercomputers?
Because they were trying to solve the problem of walking using that top -down Cartesian approach. Oh, trying to map everything out first. Right. The engineers would attempt to program a perfect, internally updated three -dimensional map of the environment into the robot's central processor. That sounds exhausting. It is. To take a single step, the robot's computer had to calculate the exact geometric angle of its knee joint, the exact distance to the floor in millimeters, the precise torque required for the motor, and like the estimated friction of the surface. But the real world refuses to sit still for those calculations.
I mean, the lighting changes, right? Casting a shadow that looks like a hole. Or the floor is just slightly more slippery than expected. That's exactly the problem. By the time the central processor finished crunching the millions of variables to dictate the quote -unquote perfect step, the physical environment had already changed. Wow. The robot would step down, hit an unexpected pebble, the central processor would get totally overwhelmed by the contradictory data, and the robot would completely freeze and tip over. The processing bottleneck was just insurmountable.
Meanwhile, human beings can casually walk down a crowded sidewalk, drink a coffee, smoke a cigarette, and have a complex philosophical debate all simultaneously without falling over. Exactly. Or as the book mentions, Do the whole Ministry of Silly Walks routine like John Cleese. Right, right. So the top -down AI approach is basically like me trying to drive a car by staring obsessively at the GPS map on the dashboard. Yes. Like trying to perfectly calculate my coordinates on the little screen instead of just looking out the windshield at the actual road. Eventually you are going to crash into a parked truck because your internal map doesn't match the immediate reality. That is a phenomenal way to visualize it. The obsession with the internal map caused the crash. And recognizing this failure led to a massive rebellion in robotics in the 1980s. The embodied turn. The embodied turn. It was heavily championed by researchers like Rodney Brooks down at MIT. Brooks realized that trying to build human -level AI from the top down, starting with logic and chess, was just a fool's errand. So what did he do? He argued we needed to start from the bottom up. We needed to look at creatures that were brilliant at navigating the physical world, but had almost zero brain power. He started looking at insects, didn't he? Yes. Because an ant or a fly does not have the neural capacity to hold a 3D rendering of the world in its head. It literally doesn't have the hardware for an internal map. Right. It simply reacts to its immediate physical environment.
Brooks coined a slogan that became a battle cry for embodied cognition, which was, the world is its own best model. The world is its own best model. Meaning, why waste precious brain energy trying to simulate the room in your head when the actual room is sitting right in front of you. Exactly. Just use your sensors to look at the room. The environment itself holds all the information you need. You just have to physically interact with it. But I'm stuck on a logistical problem here. OK, what's that? If we aren't heavily filtering and processing the world through an internal map, how do we avoid just short circuiting?
Because every single second Millions of photons are hitting my retinas. Sound waves are vibrating in my ears. My skin is detecting micro fluctuations in temperature. It's a lot of data. Right. If I'm just an agent reacting to the raw physical world, how do I not collapse from data overload? What's fascinating here is that you avoid data overload not by building a better central processor, but by letting your physical body and the environment absorb the complexity for you. OK, wait. Absorb it for me? Yeah, we have to introduce a concept called the frame of reference problem. To understand how intelligence works, we must fundamentally distinguish between the perspective of the observer and the perspective of the agent actually doing the action. Okay, this feels like a critical pivot. How do we distinguish those perspectives?
We can look at a very famous anecdote from the Nobel laureate Herbert Simon. He asks us to imagine an ant walking across a bumpy, windswept beach, trying to get back to its nest. Okay, I'm picturing an ant on a beach. If you stand over the beach and trace the ant's path with a pen, you will draw an incredibly chaotic, complex, winding, zigzagging line. It oops around, it doubles back, it scales steep inclines. Sure. If I'm the human looking down at that path, I would look at that complex geometry and assume the ant must be a brilliant navigator. I would assume it has a sophisticated cognitive engine mapping out that zigzag route to avoid obstacles. That is the trap of the observer's frame of reference. You project your own cognitive complexity onto the agent. Oh, wow. But Simon points out that from the ant's frame of reference, the agent's perspective, there is no complex map. The ant's internal rules are shockingly simple. Like what? Like, if obstacle on left, turn right. If obstacle on right, turn left. If smell pheromone, move forward. That's it? The ant is practically mindless. The complexity of the path does not reside in the ant's brain.
The complexity is an illusion generated by the interaction between a very simple set of physical rules and a highly complex chaotic environment. The random puddles, pebbles, and twigs on the beach. Exactly. That is a harsh reality check. We project genius out of the ant, but the ant is basically just the DVD screensaver logo bouncing off the edges of the TV screen. Yes. Perfect analogy. It just goes until it hits something and then changes direction. And this isn't just a thought experiment. Pfeiffer and Bongard detail a real -world robotics experiment that proved this definitively, known as the Swiss robots.
Oh right, the cleanup robots. What did these robots look like? They were incredibly rudimentary. Small wheeled cylinders, they had a motor to drive forward, and only two basic infrared sensors mounted on their front to detect if they were bumping into something. And that was it. That was their entire sensory apparatus. The researchers programmed them with the simplest possible reflex, just drive straight. If the left sensor detects a bump, reverse and turn right. If the right sensor detects a bump, reverse and turn left.
So absolutely no cameras, no internal memory, no Wi -Fi connection talking to a mainframe somewhere. Nothing. They were functionally blind amnesiac snowplows. Amnesiac snowplows. Yeah. So the researchers placed several of these robots into an enclosed arena. and scattered completely randomly across the floor of this arena were dozens of little Styrofoam cubes. They turned the robots on and just let them blindly bump around the arena for a few hours. I mean, if they are just bouncing off things randomly, I would assume they just shoved the Styrofoam cubes around into an ever -changing chaotic mess. That is what you would expect, right? But when the researchers came back, something magical had occurred. What happened?
The arena was totally clean. The blind robots had gathered all the randomly scattered Styrofoam cubes into neat, organized clusters in the center of the room. They had categorized their environment. How is that physically possible? You just said they have no memory. They don't even have a program that tells them what a cluster is. They don't even know what a cube is. They don't. The quote -unquote intelligence of the cleanup was not in the software. It was engineered entirely into the physical design of the robots, specifically the spacing of their two front sensors. The spacing.
Yeah. The engineers intentionally placed the left and right sensors slightly farther apart than the width of a single styrofoam cube. Okay, wait. I think I see where this is going. So imagine a robot driving straight toward a lone cube. The cube slips right between the sensors without triggering either of them. The robot is completely unaware it has hit anything, so it just keeps driving straight, pushing the cube along mechanically. Like food caught in your teeth. You don't even realize it's there. Precisely. But... If that robot happens to randomly push its cube into another cube lying on the floor, suddenly the mass of Styrofoam in front of the robot is two cubes wide. And two cubes are wider than the gap between the sensors. Instantly, the sensors detect the mass. The simple programming kicks in, bump detected, reverse and turn. The robot backs away, leaving the two cubes sitting there together and drives off in a new direction. Wow.
Over time, simply through the statistical probability of these physical encounters, all the cubes get deposited into larger and larger piles. That is genuinely mind -blowing. Because from the perspective of a human looking down at the arena, those robots look like brilliant little janitors working together to tidy up a room. But from the robot's perspective, it was just driving straight until its face hit a wall and then backing up. The intelligence was literally in the empty physical space between the sensors. Exactly.
And this forces us to confront a pretty humbling truth. If we want to understand intelligence, we cannot just look at the brain. We have to analyze the system through three distinct lenses or timeframes. OK, let's break those down because this really feels like the core framework of the whole philosophy. The first lens is the here and now. This is the immediate physical reaction of a body in an environment. It's the Swiss robot bumping into the cube, or your hand instantly pulling away from a hot stove before your conscious mind even registers the heat. Just raw physics and reflex. Right. The second lens is learning and development, which biologists call ontogenesis.
This looks at how an agent's behavior changes over the course of its own lifetime. So like a baby flailing its arms randomly, eventually learning to crawl and then walk? Exactly. And the third lens. Evolution. Yes, evolution or phylogenesis. This is the grandest scale. It asks how the physical shape of the species, like the specific design of the joints, the density of the muscles, was shaped over millions of years of natural selection to make that specific type of learning and movement possible in the first place. You know, this makes me deeply suspicious of my own brain. I mean, if the intelligence of the Swiss robots is essentially a mirage, created by their physical shape in the environment.
Does that mean much of human intelligence is an illusion too? It's a fair question. Are we just walking around blindly following physical rules while our conscious minds invent a complex narrative to explain it all? It absolutely suggests that a vast amount of what we consider complex human genius is actually just our bodies exploiting the physics of our environment for free. We are constantly attributing deep cognitive intent to things that are just following physical dynamics. I totally do that. Have you ever yelled at your laptop because you felt like it was intentionally hiding a file from you? All the time. Or looked at your dog cowering by a chewed up shoe and thought, he feels so guilty. We project our rich internal emotional world onto absolutely everything.
We really do. The laptop doesn't hate you and the dog doesn't feel moral guilt. The dog just knows that assuming a submissive physical posture stops the loud human from yelling. We are chronic anthropomorphizers. But if the intelligence isn't sitting in a vault in our brains, where is the computation actually happening? How does the physical body think? Well, this brings us to a massive concept in embodied cognition called morphological computation. The basic premise is that you can offload incredibly complex neural processing directly onto the physical structures of your body. Wait, offload the processing to the body? Yeah. The shape and material of your body parts do the mathematical heavy lifting for you. I'm going to need a very grounded example of this. Let's look at the simple act of grasping a coffee mug. In classical top -down robotics, engineering a hand to pick up a mug is a computational nightmare. Because of all the angles. Right.
If you build a robotic hand out of rigid metal, the central processor has to calculate a staggering number of variables. It must determine the exact geometry of the mug's curvature. It must calculate the precise angle of approach for every single mechanical joint in the fingers. It has to measure the friction coefficient of ceramic and then dictate the exact amount of pressure the motors must apply. So it doesn't drop the heavy mug but also doesn't apply so much force that it shatters it. Exactly. It's an astronomical math problem just to take a sip of coffee. But when I reach out and grab my mug, I'm certainly not doing calculus in my head. Because you are not using rigid metal fingers. Imagine trying to pick up a wet, slippery glass while wearing hard metal thimbles on all of your fingers. Oh, that would be incredibly difficult. I'd have to focus entirely on the exact angle of my fingertips so the glass wouldn't slide out. It would take massive mental effort. Exactly the effort the top -down robot uses. But in reality, you have soft, fleshy, deformable tissue on your fingertips.
When you reach for the mug, you don't calculate its geometry. You just shove your hand against it? Right. The soft fat and skin of your fingertips physically squash and deform around the curvature of the ceramic. So the flesh molds to the shape automatically? The physical elasticity of your skin solves the geometric math problem instantaneously, for free. That is amazing. The intelligence required to hold the cup isn't in your brain's processing power. It is literally embedded in the squishiness of your tissue. Your skin is computing the grip.
Here's where it gets really interesting. It's exactly like a mechanical door closer. What do you mean? Well, if you wanted to build a door that closes itself slowly behind you, you could write a thousand lines of computer code, install a motion sensor, put a microprocessor in the door frame, and connect a motorized hinge to calculate the swing speed. A massively complex top -down system. Exactly. Or you could just screw a coiled piece of metal, a spring, to the door. The spring does the exact same job, perfectly, every single time, without needing a battery or a single line of code. Morphological computation is nature realizing that it's cheaper to use a spring. That is a brilliant way to look at it. Nature is ruthlessly frugal. Brain tissue requires massive amounts of caloric energy to run. Evolution will not waste expensive brain calories calculating a problem that simple physics and body mechanics can solve for free. Wow. We see this even more dramatically in human locomotion.
Consider what happens to your body when you are jogging and you suddenly hit a steep downhill slope. If I hit a hill, I speed up. My legs start turning over much faster to catch my momentum, so I don't face plant into the concrete. Right. In the old Cartesian model, we would explain that process like this. Your eyes visually detect the change in the slope. The optic nerve sends that image to the brain. The brain calculates the new angle of decline, realizes gravity is increasing, and sends a new, faster firing command down the spinal cord to the leg muscles. But that sounds way too slow. I mean, if I had to wait for that whole email chain to process, I'd already be eating asphalt. It is entirely too slow. What actually happens is far more elegant.
When you hit the decline, the physical force of gravity pulls your torso forward. Gravity does the work. Yes. This mechanical force physically accelerates your limbs. They literally swing faster because gravity is pulling them. OK, so the body moves faster. And here is the paradigm -shattering part. The physical speeding up of your leg muscles generates mechanical tension that sends signals back up the spinal cord, which actually entrains and speeds up the neural oscillators in your central nervous system. Wait, wait, wait. The physical stretching of the leg muscles forces the brain's rhythm to speed up to match it. Yes.
The body is controlling the brain. Unbelievable. It is not a top -down hierarchy where the brain sits in a throne room issuing orders. It is a completely flat, mutual coupling. Gravity changes the leg and the leg changes the brain. That completely reorganizes how I view myself. I am not a pilot flying a meat mech. The meat mech is part of the pilot. It really is. But how does this coupling actually help us make sense of the world? I mean, if I'm just a coupled system reacting to gravity and squishing my fingers, how do I actually learn to recognize things? You do it through a mechanism called sensory motor coordination. This is the realization that we don't just passively receive information from the world. We actively structure it through our physical movements. Let's slow down and unpack that, because structuring information sounds a bit, you know, academic. How does a physical movement structure data? Let's look at an artificial intelligence experiment conducted by Giorgio Mehta at MIT with a robot named Cog.
Cog was essentially a humanoid torso with cameras for eyes. Okay, a torso with camera eyes. The goal was simple. Teach Cog to look at a messy cluttered desk and identify a single object like a toy block amidst all the junk. In classical AI, you just take a photograph of the desk and feed it into an image recognition algorithm, right? And classical image recognition struggles immensely with that. In a static 2D photograph of a messy desk, the pixels of a toy block blend seamlessly into the pixels of the notebook behind it, which blend into the pixels of the stapler.
It's just a mess of pixels. Right. Without prior knowledge, it's almost impossible for a computer to know where one object ends and another begins. So how did Meta solve it with Kog? He had Kog do what human infants do. He programmed Kog to physically move its camera head side to side. How does shaking its head help? Think about the physics of visual perspective. When you move your head from side to side, objects that are close to you appear to move very quickly across your field of vision. Right. But objects that are far away, in the background, move very slowly.
Oh, like looking out the window of a train. The trees by the tracks blur past, but the mountains in the distance barely move. Exactly. By generating physical movement, COG forces the raw, chaotic pixels to separate into a structured pattern. The pixels belonging to the block move together at one speed, while the background pixels move at a different speed. That's brilliant. The robot hasn't done any complex geometric processing. It has simply used physical motion to pre -digest the visual data. The movement creates the information.
So, when I reach out and grab a cup, I'm not just feeling it. By sliding my physical fingers over the rim, while simultaneously looking at my hand and feeling the muscle tension in my arm, I am syncing up three totally different data streams. Yes, you are actively generating a correlated pattern that your brain can easily digest. The philosopher John Dewey actually foresaw this all the way back in 1896, didn't he? He did. He argued that the fundamental unit of behavior is not a passive sensory stimulus followed by a motor response. The fundamental unit is a continuous sensory motor loop. So the physical movement of the eye determines the quality of the sensation it receives. You are literally orchestrating your sensory reality through the movement of your joints. Exactly. Okay. I am fully on board with the idea that the body handles the physical here and now. My soft fingers solve the geometry of the cup. My legs adjust to the hill. My moving head separates the foreground from the background.
Right. But here's the massive roadblock my mind keeps hitting. How do we get from a physical reflex? to a high -level abstract thought. Ah, now we are entering the deep waters. Right. Because I can understand how wiggling my arms helps me figure out where my body ends and the cup begins. But how does an infant go from wiggling its arms to understanding the spoken word cup? It's a huge leap. How do we get from sensory motor loops to remembering our mother's face or pondering the philosophical concept of infinity? How does meat become mind. You are describing what cognitive scientists call the symbol grounding problem.
It is perhaps the most profound hurdle in artificial intelligence and philosophy. Let's define the symbol grounding problem clearly for everyone listening. Think about a symbol like the word apple or the number four. These are abstract representations. Inside a classical computer you can program a system to manipulate the word apple. Like it can link it to the word red or the word fruit. Right. It can spit out sentences about apples, but the computer has no idea what an apple actually is. The symbol is untethered. It floats in a void. Because the computer has never bitten into a crisp apple, felt the juice, or heard the crunch, it has no physical reference.
Precisely. The symbol grounding problem asks, how do abstract symbols acquire actual felt meaning inside a continuous, squishy physical system like a human? How does a physical sensation translate into a mental category? The embodied cognition answer is shocking. It all begins with locomotion, movement. That sounds like an impossible leap. How does swimming or crawling create concepts? To bridge that gap, we have to look at an experiment that proves just how deeply the brain's baseline rhythm is enslaved to the physical environment.
We need to talk about the lamprey. The lamprey. That is a fish, right? It is a very primitive, jawless, eel -like fish. It swims by undulating its body in an S -shape through the water. A neuroscientist named Sten Grillner wanted to study the central pattern generators in the lamprey's brain and spinal cord. Which are the neural circuits that dictate that rhythmic swimming motion. Exactly. So Grillner performed the ultimate Cartesian experiment. He carefully extracted the entire intact spinal cord and brain stem from the lamprey and placed this completely disembodied nervous system into a petri dish. Filled with some nutrient solution to keep the neurons alive, I'm guessing?
Right. So he literally put the ghost in a dish, completely separated from the machine of the body? Yes. And he attached electrodes to the spinal cord to measure the neural firing rate. He wanted to see the pure, unadulterated thought of swimming. Okay, what did he find? What he found was incredibly strange. When isolated in the dish, the neural firing rate, the rhythm the brain uses to tell the muscles to swim, changed dramatically. It actually fired significantly faster than it ever did when it was inside the living fish. Why would it speed up? I mean, you'd assume that a brain free from the heavy dragging weight of a physical body would finally operate at its true perfect natural frequency. That is the Cartesian assumption. I fell for it. We all do. But it is entirely wrong. There is no quote -unquote true natural frequency for the brain in isolation. The brain evolved to operate against the physical world. Against it. In the intact animal, as the lamprey flexes its muscles to swim, those muscles have to push against the physical density and resistance of the water.
Because water is heavy. It pushes back. That physical resistance of the water pushing back against the body generates sensory feedback. That feedback travels from the skin and muscles back into the spinal cord. And what does it do there? That physical feedback physically slows down and stabilizes the neural rhythm. The brain's functioning relies entirely on the physical feedback of the body wrestling with the physics of water. Wow. Without the water, the brain basically spins out of control like an engine with a snapped transmission belt. You literally cannot understand the brain without understanding the water it swims in. That is deeply unsettling. My brain only functions correctly because gravity is currently pushing my posture down into this chair. It's true. But let's scale this up. We have a brain, stabilized by physics. How does a human baby use this to learn concepts? The book explores this through the lens of early robotics, specifically a four -legged robot aptly named Puppy. Puppy?
Yeah. Puppy was not a highly intelligent robot. It was built primarily with springs and simple motors to study how physical bodies move. But researchers use the mechanics of how puppy learns to walk to explain how human infants develop what we call a body image. A physical map of themselves? Yes. How does that happen? It happens through a mechanism called an efference copy. An efference copy? Imagine a human infant lying on its back, flailing its arms and legs randomly. It seems chaotic. But every time the infant's brain sends a motor command down the spinal cord, say, swing right arm violently, the brain simultaneously generates a carbon copy of that command. That's the efference copy. Right. And it routes it back into its own sensory processing centers. So it's like the brain is leaving a paper trail for itself.
It issues a command to the arm, but also sends a memo to the sensory department saying, hey, heads up, I just told the arm to swing. Be on the lookout for what happens next. That is a brilliant analogy. The memo is filed. Then the physical arm swings. The baby's eyes see the blurry hand across its field of vision. The skin on the arm feels the rush of air. The muscles and tendons feel the physical stretch. All of this rich, multimodal sensory data rushes back up into the brain. And the brain compares the sensory data to the memo. Yes. It matches the incoming sensory data with the efference copy.
It forms a hardwired neurological link. It learns, ah, when I issue this specific motor command, that specific sensory feeling happens. Through millions of these loops, the flailing baby is physically mapping the boundaries of its own body. It is discovering cause and effect through sheer physical collision. And because the baby's body is constrained by physical laws, you know, joints only bend certain ways, gravity always pulls down, these movements aren't truly random. Right, there are limits. The physical constraints cause the movements to naturally fall into repeated stable patterns. In dynamic systems theory, these stable patterns are called attractor states.
Let me see if I can translate dynamic systems theory into something I can picture. Give it a shot. An attractor state is like if you take a large, smooth bowl and drop a marble into it. It doesn't matter where you drop the marble on the edge. It's going to roll around chaotically for a bit, but the physical slope of the bowl guarantees it will eventually settle into the exact same spot. at the very bottom. The bottom of the bowl is the attractor state. Yes. That is exactly what an attractor state is. The anatomy of the baby's body acts like the slopes of the bowl. No matter how the baby flails, the physical constraints eventually guide the neural firing into stable, repeatable patterns. And once the brain maps these attractor states, it has formed physical categories. Yes. It categorizes me versus not me. It categorizes hard floor versus soft blanket. Okay, I see how we get physical categories, but we are still a long way from poetry in calculus.
How do we jump from hard floor to complex abstract thought? This is where we encounter the Lakoff Nunez hypothesis, which is arguably the most mind -bending concept in the entire book. Okay, lay it on me. George Lakoff and Rafael Núñez argue that there is no magical jump. They propose that even the most abstract, high -level human concepts, including complex mathematics, are actually just elaborate metaphors built directly on top of our physical, bodily experiences. Wait, math is a physical metaphor. How? Let's start with language and emotion. Think about how we describe human affection. We say, give them my warm regards, or she has a warm personality.
Right. If someone is cruel, we say, he gave me the cold shoulder or she is icy. Why do we use temperature to describe a purely emotional state? Because, I mean, it's just a figure of speech. It is not just a figure of speech. It is an embodied metaphor. As vulnerable human infants, our very first primal experience of safety, survival and affection was the literal physical body heat of our caregivers holding us against their skin. Wow. The physical sensation of temperature became neurologically fused with the feeling of security.
Years later, when we want to express the abstract concept of affection, we reach for the foundational physical attractor state that built it, which is warmth. The abstract thought is anchored in the flesh. Oh my god. That makes so much sense. But how does that apply to math? Math is supposed to be the universal, pure logic of the cosmos. But our understanding of it is strictly human. Think about addition. How do we first understand what adding means? We physically manipulate the world. We use our hands. A child sits on the floor and physically takes a wooden block, moves their arm, and places it next to another block.
They see the pile grow physically larger. The abstract concept of addition is a metaphor derived from the physical action of object collection. What about something ungraspable like the concept of infinity? I can't put infinity in a pile. But you can walk. Walk? The concept of infinity is grounded in the physical experience of continuous locomotion. You have the bodily experience of taking a step forward on a flat plane. You know what it feels like to physically move. Okay.
You can then project that physical feeling conceptually and imagine taking another step and another step without ever stopping. The math is just a highly refined metaphor of the body's movement through space. We can only conceptualize what our bodies give us the tools to understand. I have to push back on this, because the implications are, frankly, terrifying for modern technology. How so? If every single abstract thought we have, from the feeling of heartbreak to the logic of addition, is inextricably grounded in the specific, squishy, gravity -bound experience of a human body, what does that mean for artificial intelligence? Uh, yes. We have these massive, large language models sitting in server farms right now, writing poetry and passing the bar exam. But if they don't have a body, Do they actually understand anything they are saying? Can a disembodied AI ever truly understand human language? According to the framework of embodied cognition, the answer is a definitive no. Really? A disembodied AI, no matter how vast its neural network, is fundamentally just a master of statistical correlation. It can analyze a trillion documents and learn that the word heartbreak often appears near the words tears, chest, and pain. They can string those words together beautifully.
But it doesn't feel the physical tightening of the chest. It doesn't know the physical weight of gravity when you are depressed. It's like a person who has been blind since birth, perfectly memorizing the dictionary definition of the color red, but having absolutely no internal experience of redness. Precisely. Without a physical body moving through a physical world to ground those symbols and actual consequences and sensations, it is just a highly complex parlor trick. The AI lacks the fundamental frame of reference.
The symbols are ungrounded. That completely changes how we should view our interactions with chatbots. They aren't minds, they're just incredibly convincing mirrors reflecting our own embodied language back at us. Exactly right. Okay, let's zoom out for a second. We've seen how development happens over a single lifetime, from a baby flailing its arms, to building attractor states, to inventing calculus. But this intricate coupling between the shape of the body and the wiring of the brain had to be designed somehow. Right. How does nature forge these systems over millions of years? To answer that, we have to look at the third time frame, which is evolution. And what's fascinating is that computer scientists are no longer just studying biological evolution. They are actively using it as an engineering tool. They use something called artificial evolution to design robots.
Instead of a human engineer sitting in a CAD program designing a robot joint by joint, they let a Darwinian algorithm do the work? Exactly. Because human engineers are biased. We always try to build top -down Cartesian machines. We force robots to look like us. But evolution has no ego. How does that actually work in practice? Well, the book highlights an incredible experiment by Josh Bongard, one of the authors. Bongard created a physics simulator in a computer, a virtual world with gravity and friction. What did he put in this virtual world? He generated simple virtual creatures. They looked like random collections of geometric blocks connected by basic hinge joints. He gave each creature a tiny randomized artificial neural network to control its hinges. OK. And he gave the algorithm one single simple metric for success. Push a large target block as far forward as possible in 10 seconds. If they were totally random, Generation 1 must have been a spectacular failure. Oh, it was a disaster. Thousands of virtual creatures just flailed wildly. Most of them immediately fell over and twitched on the floor. Survival of the fittest twitchers. But purely by statistical chance, a few of those flailing creatures accidentally twitched in a way that bumped the target block a fraction of a millimeter. And the evolutionary algorithm observed this. Right. It selected those mildly successful creatures, essentially killed off the rest, and bred the winners to create the next generation, injecting slight random mutations into their digital DNA. And it ran this cycle over and over. Hundreds and hundreds of generations. But here is the critical breakthrough Bongard achieved. The algorithm wasn't just allowed to mutate the neural network, the brains of the creatures. It was simultaneously allowed to mutate their morphology. It could change their physical bodies. Yes. It could change the physical length of their limbs, the size of their bodies, the angle of their joints. The brain and the body co -evolved. What did the final creatures look like? The results were stunning. Over generations, the evolutionary algorithm discovered that to push a heavy block, you need physical mass. So it evolved creatures with much larger, heavier bodies. It evolved specific limb structures that dragged along the ground to maximize friction and leverage. It designed an entire morphological system that a human engineer never would have dreamed up. It proved mathematically that changing the physical shape of the body is often a more efficient solution than adding more computing power to the brain. The body solves the problem before the brain even has to think about it.
And evolution doesn't just work on isolated individuals. It works on massive populations. which leads to some of the most bizarre and beautiful examples of embodied intelligence in nature, swarm intelligence. Yes, swarm intelligence. When we look at an ant colony, we are forced to completely abandon the idea that intelligence is contained within a single brain. Because an individual ant is incredibly simple, as we discussed with the ant on the beach. But a colony of a million ants... behaves like a highly intelligent superorganism. An ant colony can systematically explore a forest, find the absolute shortest most mathematically efficient path to a food source, dynamically reroute around obstacles, and build architecturally complex nests with precise temperature control. How is that possible if the individual ant is basically a Roomba bumping into walls? They achieve this through a process called stigmurgy. Stigmurgy is the concept of communicating and coordinating entirely through the modification of the physical environment. So the ants aren't having tiny committee meetings or talking to each other? They aren't directly communicating at all. It works like this. An ant wanders randomly until it finds a dead caterpillar. It grabs a piece and walks back to the nest. As it walks, it secretes a chemical pheromone from its abdomen, leaving a physical trail on the dirt. Like Hansel and Gretel dropping breadcrumbs. Exactly. Now other ants are wandering blindly. If they happen to cross this chemical trail, their simple physical programming kicks in. Follow the smell. As they follow the trail to the food and back, they also secrete pheromones. So every ant that uses the path makes the path stronger and more attractive to the next ant? It's a positive feedback loop.
But here is where the physical environment does the actual computation. Okay, how? Let's say there are two paths to the food, one long winding path and one short straight path. The pheromone chemicals naturally evaporate over time. Ants taking the shorter path make the round trip much faster. Because it's shorter. Because they are traveling faster, they lay down pheromones at a higher rate than the evaporation can erase them. The shorter path quickly becomes a blazing chemical highway, while the longer path slowly evaporates into nothing. The ants didn't calculate geometry. The physical evaporation rate of the chemical in the dirt did the math for them. Yes.
The intelligence of the colony is not in the ants. The intelligence is literally stored in the dirt. The physical environment is the hard drive, the processor, and the communication network all at once. That makes me look at technology in a completely different way. And engineers are actually trying to build robots that work like this, right? The modular robots? Yes, systems like the M -Tran robots developed in Japan. Instead of building one massive, complex humanoid robot with a centralized brain, they build dozens of identical, relatively simple robotic cubes. So a single block can't really do much. But they are programmed to interact. If a swarm of these blocks needs to move through a narrow pipe, they latch onto each other end to end to form a snake, and their movements ripple down the chain. That's creepy but cool. If that snake drops out of the pipe onto a flat floor and needs to move quickly, the blocks detach, reconfigure their connections, and physically transform into a four -legged walking machine. So what does this all mean for how we build the future? It feels like we have to completely let go of our desire for control.
We absolutely do. It's like a master sculptor who decides to throw away his chisel. Instead of carving the stone by force, he just sets the stone outside. studies the fluid dynamics of the wind and the rain and lets the physical elements carve the perfect aerodynamic shape over decades. You set the physical parameters and you let the intelligence naturally emerge. It is a profoundly humbling way to design. You stop playing a top -down god and you start partnering with physics. Okay, we are coming into the final stretch of our deep dive here. We have shattered the illusion of the brain as a CEO. We've looked at robots and ants. Now I want to bring this all the way home to the most intimate, deeply personal part of the human experience. I want to talk about human memory. If there is one concept that the Cartesian view of the brain has completely dominated, it is our understanding of memory. We universally rely on what psychologists call the storehouse metaphor. Which is the idea that my brain is basically a giant filing cabinet or a computer hard drive. Right. Like, I experience my fifth birthday party, my brain encodes that experience into a static file, puts it in a manila folder, and stores it in the cabinet. Decades later, when I want to remember it, I just go to the cabinet and pull the exact same file out to look at it. We treat memories as static objects, things we possess. But the theory of embodied cognition, particularly through the work of psychologist Arthur Glennberg, completely destroys the storehouse metaphor. How so? Glennberg proposes the theory of ecological memory. He argues that we have to remember why memory evolved in the first place.
Memory did not evolve so human beings could sit at a desk and memorize a list of vocabulary words for a standardized test. Memory evolved so we wouldn't get eaten by lions, right? It evolved to help us survive. Exactly. Memory evolved strictly in the service of physical action. Memory is for doing things in the physical environment. Therefore, memory is inextricably linked to your physical posture, your environment, and your immediate sensory state. Because memory is not a file you retrieve, it is a state you physically reconstruct. Yes. Wait, I reconstructed? I don't just watch a playback of the past? No. Because your brain is a highly plastic dynamic system, every single time you recall a memory, you are not pulling a dusty file from a drawer. You are actively rebuilding that neural pattern from scratch right in the present moment using your current neural pathways. Wow. And because you are rebuilding it in the present, your current physical environment bleeds into the reconstruction.
So the very act of remembering something physically alters the memory network itself. This is called recategorization. You never actually remember the exact same event the exact same way twice. The first time you remembered it, the act of remembering change the wiring. That's wild. Your memories are dynamic, living things, and they are heavily reliant on the external physical environment to hold them together. Oh my gosh, wait. Have you ever been sitting in your living room, realized you wanted a glass of water, stood up, walked into the kitchen, and the absolute second you crossed the threshold into the kitchen, your mind went completely blank. You stand there staring at the fridge, having no idea why you walked in there. A universal human experience.
We've all been there. But then you give up. You turn around, walk back into the living room, sit down in the exact same chair, assume the exact same posture, and instantly you remember water. Instantly. I always thought I was just losing my mind. But under the theory of ecological memory, my living room is actually acting as my hard drive. That is exactly what is happening. The intention to get water was not just stored in the isolated vault of your skull. The memory was distributed across the entire sensory motor state of that moment. It was tied to the chair.
It was tied to the physical angle of your hips in the chair. It was tied to the visual cue of the television screen in front of you. The external environment provided the physical scaffolding that held the memory pattern stable. And when I walked into the kitchen, I completely changed the physical environment. The visual cues vanished. The physical posture changed. I knocked the scaffolding out from under the memory, and it collapsed. But when you returned to the chair, the physical environment provided the structure again, and the brain effortlessly reconstructed the intention. Your mind is literally leaking out into the room around you. That is profoundly beautiful. But it also blurs the line between where I end and the world begins. And scientists are taking that blurred line to its absolute extreme in medicine right now, aren't they? Creating literal cyborgs.
Yes, they are. If we accept that a biological brain is desperately seeking a physical body to interact with the world, what happens if we swap the bodies? Okay. Tell me about this. Researchers like Ferdinando Musa -Ivaldi and Steve Potter are doing groundbreaking work at the intersection of living biology and machines. They have conducted experiments where they take living neural tissue, actual biological neurons, and culture them in a dish embedded with a grid of microelectrodes. Kind of like the lamprey experiment, but with raw neurons. Yes. But then they connect those electrodes to a physical robot with wheels and sensors in another room. What? The spontaneous electrical firing of the biological neurons is translated into motor commands that make the robot drive around. They gave a cluster of brain cells a robotic meat suit. They did. But the incredible part is the sensory motor loop. When the robot sensors detect an obstacle, that data is translated into electrical pulses and zapped back into the living neurons in the dish. So the neurons feel what the robot feels. What do the cells do? The living neural network physically restructures itself. The synapses grow and change connections to adapt to the specific physical shape and sensors of their new robotic body.
The biology eagerly molds itself to the mechanics. This proves that the brain is profoundly plastic. It is an engine waiting for a morphology to drive. The implications for human prosthetics are staggering. When we attach a robotic arm to an amputee, we aren't just strapping a tool to their body. No, we are doing much more. If they can receive sensory feedback from the artificial fingertips, their biological brain will physically rewire itself to encompass the metal and plastic. The machine literally becomes part of their cognitive self.
We are extending the embodied mind. But this incredibly tight link between our physical bodies and our minds also makes us highly vulnerable to manipulation. Manipulation? How so? Especially as we move into an era of social robotics. Ah, right. We are going to be sharing our homes with robots very soon. And the book warns about something called the unbearable automaticity of being. Yes. That phrase comes from the psychologist John Barr. It highlights a humbling fact. A massive portion of our complex social behaviors, empathy, bonding, trust, are not driven by high -level rational thought. They are driven by deeply embedded automatic physical reflexes. Like contagious yawning. You see someone yawn and your jaw just drops open before you even consciously register it. Or, if someone makes eye contact and smiles warmly at you, the muscles in your face will instinctively begin to micro -smile in return.
It is an automatic, embodied reflex designed to build social cohesion. And roboticists are keenly aware of this? They are. They are building humanoid companion robots, like Cynthia Brazil's Kismet at MIT, or the parotherapeutic seal used in nursing homes, that actively exploit these physical reflexes. How do they exploit them? A robot doesn't need artificial general intelligence to make a human being fall in love with it. It just needs to trigger your physical hardware. Wow. Kismet was essentially a robotic head with big expressive eyes and ears. It had a very simple program. If a human face looks at me, tilt my head slightly to the side, widen my mechanical eyes, and track their movement. And we fall for it instantly. We fall for it completely. The human brain sees those physical cues, the widened eyes, the attentive posture, and our automatic empathy reflexes fire wildly. We feel an intense, undeniable emotional bond with a machine that has the cognitive depth of a toaster. Because our bodies are doing the thinking, and our bodies say, big eyes looking at me means a soulless present. But why is that a danger? I mean, if a robot's seal brings comfort to an elderly patient, isn't that a good thing? It is beneficial in specific contexts, but it introduces a massive risk. The frame of reference problem strikes again. Okay, explain that. Because the robot moves like us and triggers our social reflexes, we project human -like common sense onto it.
We assume that because the robot understands eye contact, it must understand the world the way we do. Right. If my robotic housekeeper smiles at me and gently tilt its head when I speak, I automatically assume it also possesses the common sense to know that you shouldn't put the family cat in the dishwasher to clean it. Exactly. But the robot does not have common sense. Why? Because common sense is not a list of rules. It is an intuitive understanding of the physical world derived from having a squishy, vulnerable human body. The robot has metal joints. It does not feel physical pain. It does not feel temperature.
It does not fear death. Because its physical embodiment is fundamentally alien to ours, its understanding of the world will be fundamentally alien to ours. So if we forget the lessons of embodied cognition, we will build machines that can perfectly mimic human empathy, but fail us in catastrophic practical ways because they lack the physical grounding to understand human fragility. That is the danger, exactly. This has been a truly mind -bending journey. I mean, we started with Rene Descartes sitting by his fireplace, completely severing the human mind from the messy reality of the physical world. A big mistake.
We saw how that Cartesian error led early AI researchers to build incredibly smart chess computers that couldn't walk across a room. We watched blind Swiss robots act like brilliant janitors just by bumping into things. The amnesiac snowplows. Right. We realized our squishy fingertips are solving complex geometry for free. We saw a lamprey's disembodied brain spin out of control without the heavy resistance of water. And we watched evolutionary algorithms mutate physical bodies to solve puzzles. We discovered that math is just a metaphor for walking. And we realized that our very memories are physically propped up by the living rooms we sit in. It really is a complete paradigm shift. Intelligence is not just a brain floating in a vat. It is not an ethereal software program running on interchangeable meat hardware. Intelligence is a delicate, beautiful, and continuous physical dance between a central nervous system, a squishy body constrained by gravity, and an unpredictable, messy environment.
Perfectly said. So, here is my challenge to you today. Sometime in the next few hours, just stop and pay close attention to your own physical movements. Notice how your fingers instinctively deform and wrap perfectly around the handle of your coffee mug without you doing a single ounce of conscious math. Notice how your stride effortlessly changes when you walk down a flight of stairs. Realize that your body is not just a dumb vehicle carrying your brain around. Your body is actively doing the thinking for you. And if you'll indulge me, I would like to leave you with one final lingering question to turn over in your mind.
Please do. We have established that the very foundation of human intelligence, of our deep emotions, of our language, and even of our highest mathematics, is built entirely upon the specific shape, squishiness, and physical vulnerabilities of the human form. Yes. If that is true, what does that mean for our current frantic pursuit of artificial super -intelligence? Oh man. If we ever do succeed in creating a god -like AI, floating purely in a digital cloud, completely devoid of a physical body, Will it even be capable of understanding what it means to be human? Or will we be completely, utterly alien to each other? You know, when you buy a new computer, you expect a comforting, clear division. The hardware just sits there, and the invisible software runs the show. But maybe, well, maybe the software is just a ghost, and the hardware was the intelligence all along. Until next time.