Latest / The Psychology Undergrad Podcast / Face Blindness & Object Recognition: How the Brain Identifies What It Sees
Transcript
- 0:00Welcome to the Psychology undergrad. Hey everyone.
- 0:03Have you ever wondered why you might feel genuinely
- 0:07like viscerally bad for a robot getting kicked?
- 0:11Even though you know for an absolute fact it
- 0:14is literally just metal and wires. Yeah, it's
- 0:17such a fascinating glitch in how our brains mathematically
- 0:19calculate reality. Right. And we're going to
- 0:23answer exactly why that glitch happens today.
- 0:25We're so glad you could join us for this study
- 0:27session. Definitely. Think of us as your two
- 0:29psychology study buddies, and you are the crucial
- 0:32third member of our group sitting right here
- 0:34with us at the library table. In all our messy
- 0:37notes. Exactly. So today we were taking a deep
- 0:39dive into chapter 11 of our textbook, which is
- 0:42visual development and object recognition. The
- 0:45whole goal for this deep dive is to figure out
- 0:48the mechanics of how humans actually learn to
- 0:51see, organize and interpret what is honestly
- 0:54a highly chaotic world. Yeah, because we just
- 0:56take vision totally for granted. We really do.
- 0:58You just, you know, open your eyes and the world
- 1:01is just there. But the textbook lays out this
- 1:04incredible journey from birth where we are essentially
- 1:08flying blind to developing a brain that that
- 1:13basically performs high -level probability algebra
- 1:15every single millisecond, just to guess what
- 1:19is casting a shadow on the wall. It's wild. But
- 1:22before we get to the algebra of shadows, we need
- 1:25to trace the physical hardware development first.
- 1:27Right, starting at the literal beginning. Yeah,
- 1:29let's start with infants. Because when you picture
- 1:31a newborn baby looking up at you, you just assume
- 1:34they're studying your face, taking in your expressions.
- 1:36Just mirroring at you. Right. But the reality
- 1:38is that newborn vision is heavily, heavily impaired.
- 1:442400 visual acuity. Yeah, meaning an infant has
- 1:46to be 20 feet away to see the same level of detail
- 1:49that an adult with normal vision can see from
- 1:51400 feet away. Wow. So the world is just heavily
- 1:54blurred and and their focal distance is incredibly
- 1:57short. They can only resolve objects that are
- 1:59about 8 to 16 inches from their face. Just a
- 2:01few inches. Yeah, anything beyond that just drops
- 2:04off into a complete fog. OK, let's unpack this
- 2:06because that 8 to 16 inch focal distance that
- 2:10is exactly the distance to a caregiver's face
- 2:12during feeding the biology is so specific it's
- 2:16perfectly engineered for that specific survival
- 2:18window so a common way of thinking about this
- 2:21which actually really helps me remember it for
- 2:23the exam is to just imagine the womb as a dark
- 2:26sensory deprivation tank right you don't need
- 2:30a high -definition 4k camera when you were just
- 2:33floating and pitch blackness for nine months
- 2:35you know exactly the hardware simply hasn't had
- 2:38the light stimulus to to finish building itself
- 2:41yet. Biologically, the reason for the blurriness
- 2:44really comes down to the photoreceptors in the
- 2:47retina. The cells that actually catch the light.
- 2:49Yes. In a newborn, these cells have these massive
- 2:52inner segments which house all the cell machinery,
- 2:55but the tiny outer segments. Wait, the outer
- 2:57segments are the parts that actually capture
- 2:59the photons? Precisely. So they're tiny. So the
- 3:01light is hitting the retina. But the cells just
- 3:04don't have a large enough net to catch. Exactly.
- 3:07The visual information is severely undersampled.
- 3:10So the data coming in is super low resolution.
- 3:12And on top of that, the neural connections carrying
- 3:15that data back to the visual cortex are incredibly
- 3:18sparse. Like a bad dial -up connection. Yeah.
- 3:20It just takes time for light exposure to stimulate
- 3:23the growth of those pathways. That actually explains
- 3:26the textbook's note about how newborns look at
- 3:29faces. because they don't look at your eyes.
- 3:31No, they don't. They tend to stare at the chin
- 3:34or like the hairline because those areas have
- 3:37stark high contrast borders. The visual system
- 3:41just isn't mature enough to process the really
- 3:43fine low contrast details of... eyelashes or
- 3:48irises. Which is the exact same reason infants
- 3:50prefer high contrast black and white checkerboards
- 3:53over those soft pastel colors that people always
- 3:55paint nurseries with. Right. The aesthetic is
- 3:57for the parents, not the baby. Exactly. And that
- 4:00visual system takes significant time to really
- 4:02come online. Like depth perception, the brain
- 4:05learning to track both eyes together to create
- 4:07virgins and stereo vision that doesn't really
- 4:10kick in until about three months. The visual
- 4:12system is mostly functional by age one, but things
- 4:15like hyperacuity. Hyperacuity is, that's the
- 4:19ability to see beyond the physical limits of
- 4:21the retina, right? By having the brain sort of
- 4:23extrapolate straight lines. Yes, that's it. And
- 4:25that advanced processing doesn't fully develop
- 4:27until a child is past the age of 10. A decade.
- 4:30Yeah, the brain literally spends a decade learning
- 4:33how to see. Man. The textbook touches on a philosophical
- 4:38parallel here that I found really compelling.
- 4:41Just this universal human theme of starting in
- 4:44the dark. Oh, yeah. The chapter mentions how
- 4:47this biological progression kind of mirrors the
- 4:50way human moral and spiritual frameworks are
- 4:52often described. Right. Like we begin with a
- 4:55blurry, super limited understanding of the world,
- 4:57relying on these high contrast black and white
- 5:00rules. Good, bad, right and wrong. Exactly. And
- 5:02then over a lifetime of building connections
- 5:05and gaining con - context, we slowly move toward
- 5:07the light and toward clarity, and we learn to
- 5:10see the subtle shades of gray. It is a really
- 5:12great metaphor. Yeah. But moving from metaphor
- 5:15back to the actual physical brain, how do we
- 5:18actually know those neural pathways are sparse
- 5:21in infants, or that they're forming correctly
- 5:23in adults. Right, because we can't just... Yeah,
- 5:25we can't just open up someone's skull to watch
- 5:27the visual cortex wire itself together. We have
- 5:29to map the brain through the skull. Which brings
- 5:32us to neuroimaging. Yes. Chapter 11 breaks down
- 5:35four main tools researchers use. PETE scans.
- 5:40EEG, fMRI, and MEG. And they each measure a different
- 5:43byproduct of brain activity. They do. So I actually
- 5:46have a study in demonic for these four to make
- 5:47sure we don't mix them up on the test. You ready?
- 5:49Let's hear it. OK. EEG is for electrical speed.
- 5:52MEG is for magnetic shields. fMRI is for fine
- 5:55structure. and PETE requires a potion. A potion.
- 5:58Okay, let's test that mnemonic. The PETE scan
- 6:01potion is a mildly radioactive tracer, right?
- 6:04Right. The person is injected with this tracer,
- 6:06and as the blood flows to the active parts of
- 6:08the brain to deliver energy, the PETE scanner
- 6:10picks up the radiation. Nice. Okay, then you
- 6:13have EEG electroencephalography. The electrical
- 6:17speed? Yeah, this is the classic shower cap looking
- 6:21thing with all the electrodes. It just sits right
- 6:23on the scalp and measures the electrical brain
- 6:25waves directly. What's fascinating here is the
- 6:27conflict between these tools. Choosing a neuroimaging
- 6:30method is basically always a compromise between
- 6:33temporal resolution and spatial resolution. Okay,
- 6:36break those down for me. Temporal resolution
- 6:38is knowing exactly when a neuron fires. EEG is
- 6:43incredible for this. It tracks electrical spikes
- 6:45down to the literal millisecond. Wow. But because
- 6:48the skull smears electrical signals, an EEG has
- 6:53terrible spatial resolution. It can't tell you
- 6:55exactly where deep in the brain the signal actually
- 6:58originated. Oh, I see. So if I need to know exactly
- 7:00which millimeter of the visual cortex is processing
- 7:03a color, EEG is kind of useless. Pretty much.
- 7:05I'd need fmr. for that. Functional magnetic resonance
- 7:08imaging. My mnemonic is fine structure. Because
- 7:11fMRI gives us highly detailed 3D structural maps,
- 7:14it doesn't measure electricity at all. It tracks
- 7:16the BOLDY signal, the blood oxygen level dependence
- 7:19symbol. BOLDY, right? Yeah. When neurons in a
- 7:22specific area fire, they consume oxygen. So a
- 7:27few seconds later, oxygen -rich blood rushes
- 7:30to that specific area to refuel them. The MRI
- 7:33magnet detects the difference between oxygenated
- 7:36and deoxygenated blood. But because it relies
- 7:39on blood flow, which takes a few seconds to rush
- 7:41in, the temporal resolution is terrible. Exactly.
- 7:45You know where it happened, but you completely
- 7:46missed the exact millisecond it occurred. Okay,
- 7:49which leaves MEG, Magnetoencephalography, the
- 7:53best of both worlds. Theoretically. Right, because
- 7:56when electrical currents flow through neurons,
- 7:58they produce these tiny magnetic fields. Mag
- 8:01measures those fields, which gives you the millisecond
- 8:03speed of an EEG with much better spatial accuracy.
- 8:06But my mnemonic was magnetic shields. And the
- 8:09textbook points out that the brain's magnetic
- 8:11field is incredibly weak. So weak. Like weak
- 8:13compared to the Earth's magnetic field, or even
- 8:15just the ambient magnetism of a passing car outside
- 8:18the lab, right? Yeah, to get a clean reading,
- 8:20an Meg machine has to be housed in the specially
- 8:22constructed magnetically shielded room made of
- 8:25layers of aluminum and high permeability metal
- 8:27alloys. Which sounds cheap. Ha! No. It is prohibitively
- 8:31expensive. Which is why fMRI remains the standard
- 8:35for mapping out our visual architecture. Got
- 8:37it. And when researchers put people in those
- 8:40fMRI machines and show them images, they found
- 8:43something wild about how the visual cortex is
- 8:45organized. Oh, the pathways. Yeah. Because once
- 8:47the raw visual data hits the primary visual cortex,
- 8:51which is V1, located at the very back of the
- 8:53head, it doesn't just stay there. No. V1 is essentially
- 8:56just the sorting center. It computes the absolute
- 8:59basics, finding edges, simple contrast, line
- 9:03orientations, and very small -scale motion. Just
- 9:06the raw ingredients. Exactly. But from there,
- 9:08the processing demands split. The information
- 9:10physically divides into two completely distinct
- 9:13anatomical highways. Okay, so we need to know
- 9:15the ventral pathway and the dorsal pathway for
- 9:18the exam. Let's start with ventral. This highway
- 9:20runs downward into the temporal lobe, and it
- 9:23is heavily focused on parvocellular information.
- 9:26Think parvo for small, slow, and highly detailed.
- 9:29The ventral pathway is nicknamed the what pathway?
- 9:32The what pathway? Yeah, its entire job is identification.
- 9:36This is where you find highly specialized processing
- 9:38centers like the FFA, the fusiform face area.
- 9:42Which allows you to recognize fine facial differences.
- 9:45Exactly. It also houses the VWFA. the visual
- 9:48word form area, which processes letters and written
- 9:50words. And here's where it gets really interesting.
- 9:54I was reading this section and had a major holdup.
- 9:58Oh, what was the holdup? Well, evolution takes
- 10:01millions of years to build dedicated brain structures,
- 10:03right? Sure. But written language and alphabets
- 10:06were invented... barely a few thousand years
- 10:08ago. Evolution didn't know we were going to invent
- 10:10books. So how can we possibly have a dedicated
- 10:13biological brain region for reading? It is genuinely
- 10:16one of the great paradoxes in visual neuroscience.
- 10:20The textbook introduces a concept called neuronal
- 10:22recycling. The prevailing theory is that the
- 10:26VWFA originally evolved under immense pressure
- 10:29to recognize specific, naturally occurring shapes
- 10:32in the environment that were vital for survival.
- 10:34Like what? Like the intersecting branches of
- 10:36a tree or the specific t -junctions and y -junctions
- 10:39found in animal tracks. Wait, so my brain is
- 10:41essentially running reading software on hardware
- 10:44meant to track animal footprints? Basically,
- 10:47yeah. When humans invented writing, we didn't
- 10:50grow a new brain area. We hijacked this pre -existing,
- 10:55highly specialized shape recognition hardware
- 10:57and just re -tasks those T and Y junctions into
- 11:00the literal letters T and Y. That is insane.
- 11:03That completely changes how I think about dyslexia,
- 11:06too. Like, we are forcing an ancient survival
- 11:08system to do something it wasn't originally built
- 11:10for at all. Totally. Okay, so ventral is the
- 11:13wet pathway. The other highway goes upward into
- 11:15the parietal lobe. The dorsal pathway. Right.
- 11:18And the dorsal pathway handles magnocellular
- 11:21information. So large, fast -moving data. This
- 11:23is the where and how pathway. Not the what. Right.
- 11:26Its job is not to tell you what an object is,
- 11:28but where it is in space and how you need to
- 11:30move your body to interact with it. The textbook
- 11:33highlights two specific regions here. MT, the
- 11:36medial temporal area, and MST, the medial superior
- 11:39temporal area. These are responsible for calculating
- 11:42coherent motion and optic flow. Yep. If your
- 11:46MT was damaged, you wouldn't see a car driving
- 11:48smoothly down the street. What would you see?
- 11:51You would see a static image of a car far away,
- 11:53and then suddenly a static image of a car closer
- 11:56to you. Like a lagging video game. Exactly, like
- 11:58a lagging game. To function in the real world,
- 12:02say, to catch a baseball thrown at your head,
- 12:04you need both pathways. Right. You need the ventral
- 12:06pathway to identify that the object is a baseball
- 12:09and not a rock, and you need the dorsal pathway
- 12:11to calculate its trajectory so you can raise
- 12:13your hand to the exact... right spatial coordinate
- 12:16to catch it. Okay, but if these pathways are
- 12:20in completely different parts of the brain, how
- 12:22do they talk to each other fast enough to catch
- 12:24a 90 mile per hour ball? They are bridged by
- 12:27a massive bundle of white matter called the vertical
- 12:29occipital fasciculus. Vertical occipital fasciculus.
- 12:33Yeah, it basically serves as an ultra -high -speed
- 12:35data cable syncing the what with the where. Nice.
- 12:39I've got a memory trick to keep these two highways
- 12:41straight. Let's hear it. Ventral starts with
- 12:43V. So think vocabulary and visage, words and
- 12:48faces. That covers the VWFA and the FFA. Okay,
- 12:51I like that. And dorsal makes me think of a dolphin's
- 12:55dorsal fin. A dorsal fin cutting through the
- 12:58water helps you track where the thread is moving
- 13:00and how you need to navigate away from it. vocabulary
- 13:02and visage versus the dorsal fin. That works
- 13:05perfectly. Could. But, you know, having these
- 13:07incredibly sophisticated pathways still doesn't
- 13:10solve the fundamental problem of human vision.
- 13:13Which is? The visual information hitting your
- 13:15retina is, by the laws of physics, completely
- 13:18ambiguous. Oh, right. Because a 3D world is being
- 13:22flattened onto a 2D retina, the brain is basically
- 13:25constantly missing information. Exactly. The
- 13:27textbook calls this the aperture problem. The
- 13:29aperture problem. Yeah. If you look at a larger
- 13:31moving object through a small hole in aperture,
- 13:36you cannot tell its true direction of motion.
- 13:38Because you can't see the edges of the object.
- 13:40Right. A line moving diagonally across the hole
- 13:43looks totally identical to a line moving straight
- 13:46up. And V1 suffers from this constantly because
- 13:49each neuron only sees a tiny, tiny fraction of
- 13:52the visual field. So the brain just has to make
- 13:55guesses. Yes. And this leads to ballastable illusions
- 13:57like that famous vase face image. Oh, I love
- 14:00that one. You look at it and you see a black
- 14:02face in the center. Then, without the image changing
- 14:05at all, your perception flips and suddenly you
- 14:08see two white profiles facing each other. The
- 14:10stimulus on your retina is exactly the same,
- 14:13but your conscious perception just toggles back
- 14:16and forth. Why does it do that? It happens because
- 14:18the visual data provides equal mathematical evidence
- 14:21for both interpretations. V1 hasn't received
- 14:24enough top -down feedback from the ventral pathway
- 14:27to amplify a winning interpretation, so the brain
- 14:29just alternates between the two most likely guesses.
- 14:32So to prevent us from constantly hallucinating
- 14:34shifting reality like that everywhere we look,
- 14:36The brain relies on gestalt principles. Yes.
- 14:39It takes shortcuts to organize all these fragmented
- 14:42sensory inputs into a cohesive, predictable whole.
- 14:44The core idea there is computational efficiency.
- 14:47If your brain analyzed every single individual
- 14:50leaf on a bush before deciding what you were
- 14:53looking at, the tiger hiding behind the bush
- 14:55would just eat you. Exactly. You'd be dead. You
- 14:58need to group things instantly. And one of the
- 15:00most vital principles for that is figure ground.
- 15:03the visual system immediately forces a distinction
- 15:06between the primary object, the figure, and the
- 15:08background. Right. Then there is closure. Like
- 15:11if someone draws a square using only dashed lines,
- 15:15your retina technically only sees a series of
- 15:17disconnected line segments. But your brain automatically
- 15:20fills in the empty gaps. Yeah. You perceive a
- 15:22complete square. You are actively hallucinating
- 15:25the missing borders because a complete shape
- 15:27is just easier to process than a dozen random
- 15:30fragments. And all of these shortcuts, like similarity,
- 15:33continuity, proximity, they all fall under the
- 15:35overarching Gestalt principle of pregnancy. Pregnance.
- 15:38Yeah, pregnancy dictates that we inherently perceive
- 15:41things in the simplest, most concise, and meaningful
- 15:44way possible. The brain basically despises chaos.
- 15:48It aggressively forces order onto ambiguous inputs.
- 15:51Exactly. The brain even assumes things about
- 15:54the sun to force that order, like the textbook
- 15:57talks about the light from above assumption in
- 16:00shape from shading. Oh, this part is so cool.
- 16:03It is. Because humans evolved on a planet with
- 16:05a single light source overhead, the sun, our
- 16:08visual system, literally hardwired an assumption
- 16:11that light always comes from above. casting shadows
- 16:14downward. Right. So if you take a photograph
- 16:17of impact craters on the moon and you just turn
- 16:20it upside down, the shadows are now at the top
- 16:22of the circles. And your visual system instantly
- 16:24flips the 3D interpretation and the craters suddenly
- 16:27look like raised hills. You cannot force yourself
- 16:30to see them as craters until you flip the image
- 16:32back. It's automatic. It is. And the chapter
- 16:35raises a really interesting philosophical question
- 16:37regarding this specific bias. Oh, about the light.
- 16:40Yeah. It points out that almost every major religion
- 16:43and moral framework associates the heavens, divinity,
- 16:46and moral goodness with the concept of above
- 16:49or light. Oh, right. Conversely, the underworld,
- 16:53badness or immorality is placed below in the
- 16:55dark. It's such an important observation. The
- 16:58textbook obviously isn't making a theological
- 17:00claim, but it's pointing out that the neurological
- 17:03architecture of pregnancy, our biological drive
- 17:06to organize ambiguous sensory data by equating
- 17:09light and upwardness with clarity and safety,
- 17:13it completely mirrors how humans historically
- 17:15construct spiritual meaning. We organize the
- 17:19chaos of the moral universe using the exact same
- 17:22biological biases we use to organize the visual
- 17:25universe. We really are. But, you know, Gestalt
- 17:30principles only explain how we group these things.
- 17:33They don't really explain the actual engine doing
- 17:35the work behind the scenes. Right, the math.
- 17:37Yeah. To understand the mechanics of how the
- 17:40brain guesses what is real, the textbook introduces
- 17:43Bayesian inference. Thomas Bay's theorem is essentially
- 17:46the mathematical engine of perception. Okay,
- 17:48I looked at the formula in the book, and it was
- 17:49like, the posterior likelihood of a scene equals
- 17:53the probability of the image given the scene.
- 17:56multiplied by the probability of the scene. Wait.
- 17:59Yeah, it gets dense. You lost me entirely here.
- 18:02Are you saying my brain is literally running
- 18:04algebra equations before I can even consciously
- 18:06recognize a shadow on a wall? I am. But let's
- 18:09ditch the literal formula for a second and just
- 18:11look at the conceptual variables. Please. Okay.
- 18:15Imagine you are sitting in a chair facing a blank
- 18:18wall in your living room. There is a light behind
- 18:20you, and an unobserved object is casting a shadow
- 18:24on the wall. OK, I'm picturing it. The object
- 18:26you cannot see is what Bayes calls a latent variable.
- 18:30OK, so the shadow is the data my eyes receive,
- 18:33and the hidden object causing the shadow is the
- 18:36latent variable. Correct. Now, before the light
- 18:38even turns on, your brain already holds a mathematical
- 18:40prior. A prior? Yes. A prior is the statistical
- 18:44probability of what exists in your specific world.
- 18:47You know that a coat rack is a very in your living
- 18:49room. Sure. You also know that a highly venomous
- 18:52Komodo dragon roaming your living room is incredibly
- 18:55unlikely. Definitely. So my prior probability
- 18:57for Korak is mathematically high, and my prior
- 18:59for Komodo dragon is near zero. Exactly. Then
- 19:03the light turns on. This step is called conditioning.
- 19:06You are receiving new sensory data, the shatter.
- 19:10OK. Let's say the shadow is tall and thin with
- 19:12weird branches sticking out at the top. Well,
- 19:14that rules out the Komodo dragon entirely. Yeah.
- 19:17But it could be a coat rack, or I guess it could
- 19:19be an intruder holding a weapon with spikes.
- 19:21Right. Which brings us to the third variable,
- 19:23likelihood. Likelihood is a generative model.
- 19:27Generative model. Meaning what? It asks if the
- 19:30latent variable was actually a coat rack. How
- 19:33likely is it to generate this specific branched
- 19:35shadow? Oh, very likely. Right. If the latent
- 19:38variable was an intruder holding a spiked weapon,
- 19:40how likely is it to generate this exact shadow?
- 19:43Much less likely. Exactly. So your brain takes
- 19:46your pri - a coat rack is common in this room
- 19:48and factors in the likelihood this shadow looks
- 19:51exactly like a coat rack. Okay. It runs the math
- 19:53in milliseconds and generates a posterior conclusion.
- 19:57The brain decides. That is a coat rack, and I
- 20:00don't need to trigger a panic attack. Oh wow.
- 20:03So it takes the prior and the likelihood to equal
- 20:05the posterior. Exactly. And the brain does this
- 20:07hierarchically. V1 runs Bayesian math to guess
- 20:11the latent cause of the retina's raw data. V2
- 20:15runs Bayesian math to guess the latent cause
- 20:17of V1's edges, all the way up the ventral pathway
- 20:21until you consciously recognize the object. So
- 20:24what does this all mean for our everyday experience?
- 20:26Because this actually brings us all the way back
- 20:28to the hook from the beginning of this study
- 20:30session. The robot. The robot. The textbook uses
- 20:34the Boston Dynamics big dog robot to illustrate
- 20:37when Bayesian map gets like Completely hijacked.
- 20:40Yeah, that four -legged headless metal machine
- 20:42they designed for military transport. It looks
- 20:45absolutely nothing like a biological dog. Right.
- 20:47And there's a famous engineering video where
- 20:49a researcher walks up and kicks the robot hard
- 20:51in the side to test its gyroscopic balance. I've
- 20:54seen that. When I watch the video and the robot
- 20:56stumbles and frantically taps its metal feet
- 20:58to catch its balance, I feel a visceral, intense
- 21:01moral objection. You feel bad for it. I feel
- 21:04intense empathy for it. Why does my brain care
- 21:06about a metal box being kicked? What's fascinating
- 21:09here is, if we connect this to the bigger picture,
- 21:12it is purely a conflict of Bayesian variables.
- 21:15How so? Your prior heavily dictates that dogs
- 21:18do not look like metal boxes with wires. Visually,
- 21:22the probability of it being alive is effectively
- 21:24zero. Right. But then the robot moves. And the
- 21:27movement changes the equation. The generative
- 21:29model of likelihood completely takes over. Your
- 21:32brain asks... If this object is just an animate
- 21:36middle, how likely is it to generate the specific
- 21:39frantic stumbling reaction to being kicked? In
- 21:42nature, that likelihood is incredibly low. Right.
- 21:45Things that react to physical trauma by dynamically
- 21:47shifting their weight to survive the blow are
- 21:50almost universally alive. So the math of the
- 21:52movement likelihood completely overrides my visual
- 21:54prior. Yes. Your brain calculates a massively
- 21:57high posterior probability that this entity possesses
- 22:01agency and life. Wow. And because the subconscious
- 22:04math concludes it is alive, it instantly triggers
- 22:07your human empathy and your moral response before
- 22:10your conscious brain can even remind you that
- 22:11it's just a robot. That is deeply unsettling,
- 22:14honestly. powerful math. It really is. Well,
- 22:16if we synthesize everything we've covered today,
- 22:19the biggest takeaway from Chapter 11 is that
- 22:21seeing is not a passive recording of reality.
- 22:24Not at all. Your eyes are not just camera lenses.
- 22:28Visual perception is an active probabilistic
- 22:31hallucination tightly constrained by our biological
- 22:34hardware and our mathematical priors. It is an
- 22:36educated guess constructed entirely from ambiguity
- 22:40and you know that leaves us with a final thought
- 22:43to really mull over before the exam. Hit me.
- 22:45If our entire visual reality is constructed by
- 22:48mathematical priors that are built upon our own
- 22:50highly specific past experiences What happens
- 22:54when we encounter people whose cultural, moral,
- 22:56or life experiences are vastly different from
- 22:59ours? Well, their priors would be completely
- 23:00different. Right. So when people with different
- 23:03backgrounds look at the exact same complex situation,
- 23:06whether it's a social issue, a political event,
- 23:08or an interpersonal conflict, do they just hold
- 23:11a different opinion about the same reality? Or
- 23:13is it possible that, in a very literal, biological,
- 23:17Bayesian sense, their brains are constructing
- 23:19and seeing a completely different reality than
- 23:21we are? Because their foundational math is different.
- 23:24The posterior reality, their brain outputs is
- 23:27fundamentally different. Exactly. Wow. That really
- 23:31makes you rethink how much grace we should extend
- 23:34to people we disagree with. We are all just trying
- 23:37to sort through the muddy waters of perception,
- 23:39doing our best to build a clear picture from
- 23:42a blank slate. The underlying math of human experience.
- 23:45I love that. Thanks for studying with us today,
- 23:47everyone. Review those mnemonics, think about
- 23:50the priors shaping your own world, and go ace
- 23:52that exam. We will catch you next time.