What this covers

David Eagleman, a neuroscientist and CEO of Neosensory, sits down with Sean Carroll to explore how the brain functions not as fixed biological hardware but as liveware—a self-rewiring system that learns continuously from whatever data it receives. The conversation ranges across evidence that the cortex as generic computing tissue, indifferent to its assigned function, can be repurposed based on experience rather than genetic instruction. Eagleman walks through experiments in sensory substitution, including his company's vibratory wristband that converts sound into skin stimulation, allowing deaf users to perceive auditory information through touch, and work with LiDAR data that enables blind participants to sense objects behind them—something natural vision cannot do. The discussion also covers his famous free-fall experiment, in which people perceived a terrifying drop as lasting far longer than it actually did, revealing that the sense of time is constructed from memory density rather than real-time perception.

Beyond sensory augmentation, Eagleman addresses the half-baked strategy of human development, arguing that humans arrive neurologically incomplete so they can absorb cultural knowledge rather than relying purely on instinct. The brain changes across nested time scales of plasticity, from rapid synaptic adjustments to slower genetic changes, with each layer deciding whether a pattern is stable enough to encode. Eagleman emphasizes that most brain wiring comes from experience, not the genome—a point he illustrates through cases of sensory deprivation, hemispherectomy recovery, and the failure of neglected children to acquire language. He discusses the barrier of understanding the ineffability of qualia, the impossibility of explaining a sense to someone who has never experienced it, and reflects on why progress in neuroscience requires identifying governing principles rather than cataloging anatomical details. The episode engages with affordance discovery, showing how flexible the body plan is: paralyzed people learn to control robotic limbs, monkeys develop control of a third arm, and people master virtual prosthetics in minutes.

Sharpest takeaway

Eagleman argues the brain is not fixed hardware but 'liveware'—a continuously self-rewiring system that learns to drive whatever body and sensory inputs it receives, which means we can substitute, augment, and add entirely new senses by feeding correlated data through any channel.

  • The cortex is physically uniform tissue whose function is determined by which data cable plugs into it, not by genetic pre-specification
  • The brain rewires its body map and sensory allocation based on activity, enabling sensory substitution and added limbs/senses
  • Most genuine wiring of the brain comes from experience, not the genome, making humans the most plastic species

The argument · threads12 threads · 40 claims
0.85

Brain structure changes continuously at all biological scales in response to experience and input.

7 pointscentrality 5/5
  • The brain changes at all levels simultaneously—not only the strength of synaptic connections but the physical shape of neurons (which unplug, re-plug, and seek new connections), the number and distribution of receptors, biochemical cascades inside the cell, and even epigenetic changes to genome expression in the nucleus.

    “the neuron itself is like a giant tree that's unplugging and re-plugging and seeking and moving its branches all the time and looking for new connections, to the number of receptors that it distributes on the surface, to the biochemical cascades on the inside, all the way down to the genome”

  • The Human Genome Project was scientifically disappointing for the brain because the other half of what wires up your brain will never be written in a book—it is everything around you: your experiences, conversations, friends, parents, culture, and society—so nature versus nurture is a dead question since the genome sets starting instructions while every experience from birth reshapes brain structure and even feeds back to genetic expression.

    “the other half of what wires up your brain will never be written down in a book because it's everything all around us, it's all of your experiences, everything that's happened, every conversation you've had”

  • There is no weak versus strong sense of 'changing the brain'—everything you remember, from a person's name and face to an appointment time, is literally physical change synchronized over millions or hundreds of millions of synapses down to biochemical cascades and receptors and sometimes genes, with each person's representation embedded in the context of everything else they have experienced.

    “the neuron itself is like a giant tree that's unplugging and re-plugging and seeking and moving its branches all the time and looking for new connections, to the number of receptors that it distributes on the surface, to the biochemical cascades on the inside, all the way down to the genome”

  • Learning something like sheet music or riding a bike starts as a fully conscious, attention-demanding process but eventually becomes unconscious because it is burned down into the circuitry; brain imaging shows a professional's brain has almost no activity during the task while a novice's brain is highly active, because the whole purpose of brain change is speed and efficiency.

    “if you look at the activity in the brain of somebody just starting to learn a bike versus the activity of a professional bicyclist, you see that the latter case, there's almost no activity in the brain, why? Because it's been burnt into the part of the circuitry for speed and efficiency”

  • The brain changes across many simultaneous time scales, with faster-changing layers and slower, more conservative layers; each time layer evaluates whether the data is stable enough to commit a change, with conservative layers waiting to see a pattern repeat before encoding it—structurally analogous to the renormalization group in physics.

    “each time layer I'm talking about, is looking at the one above it and saying, look, do I believe the data, is that actually stable enough that I'm going to make a change?”

  • Whether information gets encoded into brain structure depends on its relevance to the organism—if something is meaningless to you, you lack the right cocktail of neurotransmitters to make any lasting change, but information important to your career or survival sticks because the right neurotransmitters signal that this matters and should be retained.

    “if I tell you some, whatever, some word in Chinese that just really doesn't mean anything to you... you don't have the right cocktail of neurotransmitters present to make any change. But if I tell you something that's really, really important to you in your career and your survival... that's really going to stick”

  • The brain has 86 billion neurons, each connected to about 10,000 neighbors, yielding roughly 0.2 quadrillion connections, with changes happening every second across all of them.

    “it's got 86 billion neurons... each one of these is connected to about 10,000 its neighbors, so you've got about 0.2 quadrillion connections”

0.81

The brain's cortex is functionally indifferent tissue that becomes specialized purely by which data input it receives.

6 pointscentrality 5/5
  • The senses we inherited from evolution are not fundamental—eyes turn photons into electrochemical spikes, ears turn air-compression waves into spikes, and the nose turns molecule mixtures into spikes—so it is all the same currency inside the brain, which raises the possibility of simply adding an entirely new sense by feeding in any new data stream.

    “your eyes are just turning photons into electrochemical spikes, your ears are turning air compression waves into electrochemical spikes, your nose is turning mixtures of molecules into spikes and so on. And it's all the same currency on the inside”

  • If you opened a hole in someone's skull and observed all the neural activity beneath it, even a professional neuroscientist could not tell whether they were looking at visual, auditory, or somatosensory cortex, because it is exactly the same stuff everywhere—just spikes and chemical signals carrying information from neuron to neuron.

    “I would have no idea as a professional neuroscientist, because it's exactly the same stuff, it's just spikes, it's these little electrical signals”

  • The cortex is the same stuff everywhere—a roughly 3-millimeter-thick wrinkly outer layer whose cellular architecture looks identical no matter where you look—so a region becomes 'visual cortex' or 'auditory cortex' purely based on which data cable plugs into it; the 2000 MIT ferret experiment rerouted eye data to what would have been auditory cortex and that tissue developed all the properties of visual cortex.

    “it's just a matter of what data cable you're plugging into it. If you're plugging in a visual data cable, then it becomes visual cortex, auditory and so on”

  • The central lesson of neuroscience on this topic is that no neural real estate ever sits unused—if a child is born blind, the visual cortex is repurposed for hearing and touch; if born deaf, the auditory cortex is taken over by other functions; and a child who undergoes a hemispherectomy before age six or seven can be fine because the remaining half rewires to take over the missing half's functions.

    “no real estate ever sits unused, no land lies fallow. Instead, it gets taken over, so if a child is born blind, the part of the brain that we would normally call the visual cortex is no longer visual”

  • Wilder Penfield discovered in the 1960s that the brain contains a map of the body, but contrary to the assumption that this map is genetically pre-specified, it is activity-dependent—if you lose an arm, the brain's body map changes to represent a body without an arm, because the brain is locked in silence and darkness and only knows the body through the information streaming up into it.

    “if you, for example, lose an arm, your map of your body and your brain will change so that it represents a body that does not have an arm. And so what this tells us is that it's activity-dependent”

  • Because no neural real estate sits unused, someone who is blind provably has better hearing and better touch, since the brain devotes more real estate to those functions instead of sharing it among vision—and the more real estate devoted to a task, the better the performance.

    “for someone who, let's say, is blind, they actually provably have better hearing and better touch because they just devote more real estate towards those other functions, and the more you have, the better”

0.67

New sensory channels can be added by feeding consistent data streams into the brain via any input method.

6 pointscentrality 4/5
  • Sending information into the brain is comparatively easy—your eyeballs already do it by reading Wikipedia, and the Neosensory wristband passes information streams non-invasively—so Eagleman has bet on inexpensive, non-invasive input methods, while getting information out of the brain remains the hard part.

    “I've put all my chips on doing inexpensive, non-invasive ways of streaming information to the brain. Getting it out is the hard part”

  • Smartwatch data such as heart rate, heart rate variability, and galvanic skin response can be delivered as spatiotemporal vibration patterns so you feel your own physiology—or, by swapping bands between spouses, feel someone else's body state in real time—an experiment Eagleman is running whose value to relationships is not yet known.

    “let's say we switch the bands, so I am feeling your physiology and you're feeling mine, and the question is, what is that experience like when you're really tapped into what someone else is feeling”

  • Because it doesn't matter how information reaches the brain as long as it gets there, a wristband (Neosensory's Buzz) with vibratory motors can capture sound in real time and convert it into patterns of vibration on the skin, allowing deaf people to come to understand the auditory world through their skin since the brain figures out what to do with any consistent input.

    “for people who are deaf, for example, we capture sound in real time and turn it into patterns of vibration on the skin with this little wrist band, and they can come to hear the world”

  • Walking with a prosthetic leg is difficult because there is no feedback, so users must watch the limb; by placing pressure and angle sensors in the prosthetic and delivering that information through a wristband, people learn to walk again because the brain is excellent at finding correlations and accepts that the feedback now arrives at the wrist rather than the leg.

    “We put pressure and angle sensors in it and then give that information through the wrist band, and people can come to walk that way again because they're getting the information, it's not coming from the leg, it's now coming through the wrist, but it doesn't matter”

  • A belt that buzzes toward magnetic north gave wearers a constant sense of direction, and crucially they retained a better sense of where north was even after taking it off—implying humans already possess a little latent magnetoreception like other animals that needs a confirming external signal to be boosted into usable perception.

    “people wore it for a few weeks and then when they took it off, they retained a better sense of where north was... what that implies is that we already have a little bit of magneto reception”

  • Feeding LiDAR data through skin vibration let blind participants sense the location of people moving around the room—including behind them—which in this respect is better than natural vision because you cannot see behind you, demonstrating that new sensory channels need not carry the constraints of the original biological sense.

    “this is better than the vision we come to the table with, because you can't see behind you... it's like being a Jedi or something and knowing when someone's approaching you from behind”

0.58

Understanding how the brain encodes information differs fundamentally from any single experience of it.

4 pointscentrality 4/5
  • Knowing everything about a phenomenon is not the same as experiencing it—explaining all the biochemical mechanisms of chocolate ice cream has no influence on your enjoyment of it—suggesting qualia is an emergent property at a higher level that is not captured by describing the underlying physical mechanisms.

    “if I explain to you everything, all the biochemical mechanisms of why chocolate ice cream... it has actually no influence on your experience or enjoyment of chocolate ice cream”

  • It is impossible to convey the internal experience of a sense to someone who lacks it—you can explain vision all day to a person born blind or green to a colorblind person and they will never get it—so a genuinely new sense like infrared vision or feeling your microbiome or the stock market cannot be understood by anyone who hasn't experienced it, because it is qualitatively unlike any existing sense.

    “if you've ever had a friend who is born blind, you can try to explain vision to her all day long, but she's never going to get it... it's totally impossible for someone who hasn't experienced it to understand what it feels like”

  • Human memory differs fundamentally from computer memory: a computer stores an exact replica of a file, but when you hear a joke you don't remember it phoneme by phoneme—you retain the gist, which depends on all your prior experiences and lets you retell it in another language, because memory is interpretive rather than a recording.

    “if I tell you a joke, you don't remember phoneme by phoneme exactly what I said, like an audio recording. Instead, you could turn around and if you speak another language, you can tell it to someone else in another language right away, because it's the gist of the joke that you've gotten”

  • People who experienced the magnetic-north sense could not always explain it to outsiders, but the participants could communicate it well to each other, because language and vocabulary are about shared (or assumed-shared) concepts—you have to experience something to build a shared language around it, even without a pre-existing special-purpose vocabulary.

    “the different participants could talk with each other and get it really well. And so this comes back to this issue that you need to experience something in order to have a shared language around it”

0.58

Brain-computer interfaces face major technical and practical barriers despite their theoretical appeal.

4 pointscentrality 4/5
  • In about 30 years we may view Neuralink as the wrong approach, because molecularly precise 3D printing could build nanorobots—100 billion fit in a capsule you swallow—that swim through the bloodstream, impregnate every neuron, and relay their signals to an external mesh-network helmet, allowing the whole brain to be read without ever opening the skull.

    “we build nano robots, you can fit 100 billion of these in a little capsule, you swallow the capsule, they swim around in your bloodstream, impregnate all the neurons in your head, and then they send out signals”

  • Neuralink is essentially a refined version of long-standing electrode-in-brain techniques (tighter, wireless, blood-vessel-avoiding), and its real payoff will be in clinical neuroscience for Parkinson's, epilepsy, and depression where reading and writing neural activity helps patients—but the mythology that everyone will get implants to text faster is implausible because no one will undergo open-head surgery for that, so consumer adoption is far off, possibly after our lifetimes.

    “no one's going to go in for an open head surgery just so you can text faster than your son can go”

  • The foremost barrier to brain-computer interfaces is that we don't understand the spiking code of the 86 billion neurons, nor even what subset would need to be measured to get a good picture; beyond the wiring diagram, we may need the changing shapes of neurons, their excitability, and the epigenome, and since we only study neurons because that is all our technology can reach, we are like the drunk looking for keys under the streetlight.

    “the very first thing would be just understanding the spiking code of the 86 million neurons. We don't actually even know what subset you would need to measure to get a pretty good picture of what's going on”

  • Full read/write access to the brain is unlikely to be a utopian goal because the skull is an inner sanctum people want kept private; we have thoughts we would never want others to read, and these conflict with our curated social lives, so actual mind-reading will face real ethical barriers for the foreseeable future while clinical tools advance.

    “You have thoughts that move through your head that you would never want other people to be able to read or know that you thought something”

0.36

Humans arrive developmentally immature but gain extraordinary learning capacity by absorbing accumulated cultural knowledge.

2 pointscentrality 3/5
  • The human strategy of arriving in the world 'half-baked' produces extraordinarily long infancies, but it pays off because a child can learn what other humans figured out before them and springboard off it, whereas a goat must relearn how to be a goat every generation since goats have no schools, libraries, or accumulated knowledge to jump to the next level.

    “what a kid can do is go and figure out what other humans have learned before him or her, and then springboard off that, as opposed to if you're a goat, you're born and you have to learn how to be a goat again, in every generation”

  • Mother Nature drops humans into the world 'half-baked' and lets the environment wire them up, an enormously successful gamble that requires the child to receive the right inputs—at minimum love, touch, and language; severely neglected children who miss these inputs have their windows of plasticity close, so even when found at age 11 or 13 their brains can no longer learn language, how to chew food, or how to interact socially.

    “their windows of plasticity close, and then they are unable to learn things like language or even how to chew food or how to interact socially”

0.36

Skills cannot be downloaded into brains wholesale because each person encodes learning relative to their unique prior experiences.

2 pointscentrality 3/5
  • A paralyzed person can control a robotic arm via an electrode array in motor cortex, and monkeys can operate a third robotic arm alongside their two natural arms; the engineering tweaks matter less than the person learning flexibly how to think to move the arm, and in cheap VR setups people learn to control a chest-mounted third arm in about three minutes—demonstrating how flexible the body plan is.

    “you can put an electrode array in their motor cortex and then hook that up and have them control a robotic arm... most of the work is just for the person themselves to learn flexibly how to think about it, to move the arm”

  • You cannot simply download a skill like flying a helicopter into someone's brain Matrix-style, because each person encodes the skill differently relative to their own prior experiences; in principle, after a very thorough 'system identification'—presenting hundreds of thousands of scenarios and observing the individual brain's responses—you might learn enough about that specific brain to add a skill on top, but that requires enormous individualized work.

    “you'd have to do a very thorough system identification first... if I could actually understand your brain and everything about the way it's working, then maybe I could download the instructions for how to fly that helicopter, but I'd have to really do a lot of work to understand your brain first”

0.36

Understanding complex systems requires identifying governing principles rather than cataloging every detail.

2 pointscentrality 3/5
  • Progress in understanding a complex system comes from identifying its governing principles rather than cataloging all its detail—just as Crick and Watson reduced the complexity of the nucleus to keeping the order of base pairs, neuroscience needs to find what principles the brain is implementing.

    “then Crick and Watson come along and say, hey, it's about keeping the order of the base pairs and then everything else is detail. That's the kind of stuff that we need for the brain”

  • AI can perform narrow tasks with superhuman performance, like distinguishing cats from dogs in photographs, but it cannot do what a three-year-old can do—walk into a room, decide where to go, navigate obstacles, feed herself, and manipulate adults—because AI took off from a cartoon version of the brain that only models units and connection strengths.

    “it can do things like distinguish photographs of cats from dogs with super human performance, but it can't do anything that a three-year-old kid can do, like walk into a room, decide where to go next, navigate a complex set of obstacles”

0.32

Conscious awareness is a small fraction of the brain's work, with most processing hidden from introspection.

1 pointcentrality 3/5
  • The conscious part of you is like a broom closet in the mansion of the brain—almost everything the brain does, you have no access to or awareness of; when you 'have an idea,' the brain has actually been working on it for days or weeks behind the scenes evaluating hypotheses, and consciousness is the last to receive the result.

    “the conscious part of you is like a broom closet in the mansion of the brain, and almost everything going on, you don't have access to”

0.27

Mind uploading faces practical and philosophical limits because offline existence without embodied drives risks losing motivation.

3 pointscentrality 2/5
  • When discussing the dangers of living in silico, Eagleman remarks that we might already be living in a simulation.

    “Well, that's right, although, as you know, we might already be there.”

  • If an uploaded mind could control its own dashboard and turn off hunger, tiredness, and other drives, it risks removing all motivation to keep living—because you cannot enjoy the good moments without the contrast of the bad moments.

    “you can't enjoy the good moments without the contrast of the bad moments”

  • Although substrate-independent uploading should in principle be possible, the body's contributions—organs like the adrenal gland sending hormones and information—can probably be summarized in roughly a 1 MB file, while the three-pound brain is the densest representation of you, so uploading reduces to capturing the important pieces of the brain and its interactions with the world.

    “all that stuff can be summarized pretty easily, like in a 1 MB file probably, and so the brain itself, those three pounds, that's the densest representation of you”

0.22

Memory is interpretive and context-dependent, not a faithful recording, and perceived time depends on memory density and reconstruction.

2 pointscentrality 2/5
  • Eagleman's free-fall experiment—dropping people 150 feet backward into a net at 70 mph while flashing information at them—found that although people perceived the terrifying event as taking much longer than it did, no one could actually see in slow motion; the apparent time dilation is a trick of memory, because fear lays down dense memories and, read back in a Bayesian manner, more recorded footage is interpreted as a longer elapsed time.

    “people thought the event took a much longer time than it did, and yet nobody was actually able to see in slow motion... it's all a trick of memory. When you're in a fearful situation, you're laying down dense memories”

  • During lockdown, time seems to crawl moment-to-moment yet feels short in retrospect because memories are location-sensitive; when confined to the same four walls it is hard to distinguish one memory from another, so estimating how much time has passed becomes difficult and the period collapses into a blur.

    “we're very location-sensitive for our memories, and so when they're the same four walls, it's really hard to distinguish one memory from another”

0.18

People trapped in their own neural worlds cannot readily imagine others' genuinely different perceptual or political experiences.

1 pointcentrality 2/5
  • Each person is trapped in the inner cosmos of their own brain, believing everything to be right and self-consistent on their own 'planet,' which is why we cannot believe others hold different political opinions and assume they must be crazy or trolling.

    “We're all trapped on our own planet, essentially... we believe everything to be right and self-consistent on our planet”