What this covers

Ed Boyden speaks with Tyler Cowen about why neuroscience is fundamentally limited by the tools available to researchers, not by the ideas driving them. The conversation ranges across how current brain science maps and controls neural circuits—particularly through optogenetics and expansion microscopy—and why this toolmaking matters for treating diseases like Alzheimer's, depression, and epilepsy that cannot yet be cured. Boyden describes the mechanics of these technologies: expansion microscopy uses a swellable polymer to enlarge preserved tissue while preserving molecular structure, making nanometer-scale biology visible on ordinary microscopes; optogenetics borrows light-responsive proteins from bacteria and algae to let researchers activate neurons with precise bursts of light, revealing how specific brain circuits control behavior. The discussion also turns to how funding structures and peer review systematically starve pioneering work—several Nobel winners and major discoveries lost money early on—and argues for science funding organized around serendipity, backward chaining, and reviving forgotten ideas instead.

The conversation branches into several distinct territories. One explores the gaps between neurotechnology and understanding: even if scanning can read brain activity with high precision, interpreting that data and using it responsibly requires equal progress in basic science. Another examines specific findings—how Dayu Lin and David Anderson activated aggression in mice by stimulating a single brain region, how Li-Huei Tsai used optogenetics to find an Alzheimer's treatment that could be delivered noninvasively via flickering video. Boyden also discusses consciousness through the lens of unconscious brain states preceding conscious decision-making, mental illness redefined by circuit damage rather than symptoms, and ethical concerns that neurotechnology should address through self-regulation modeled on earlier bioethics conferences. Questions about memory reconsolidation, brain complexity, and even whether empathy is always beneficial round out the range, alongside skepticism about the tractability of biological systems and how tools might make the next century of science tractable in new ways.

Sharpest takeaway

Boyden argues that neuroscience is fundamentally tool-limited rather than idea-limited: progress on understanding and repairing the brain depends on building technologies (optogenetics, expansion microscopy) that map and control neural circuitry, and that scientific funding and discovery should be reorganized around serendipity, backward-chaining, and reviving forgotten ideas.

  • We know surprisingly little about the brain, and most brain diseases cannot be cured because we lack circuit-level understanding
  • Tools like optogenetics and expansion microscopy unlock new questions by controlling and mapping neural activity at the circuit and molecular level
  • Pioneering science is systematically underfunded by peer review, so funding should be dynamic, logic-based, and serendipity-seeking

The argument · threads11 threads · 45 claims
0.62

Major brain diseases remain incurable and our models of consciousness, emotion, and decision-making are incomplete.

3 pointscentrality 5/5
  • We understand surprisingly little about the brain: major diseases like Alzheimer's, Parkinson's, and epilepsy cannot be cured, treatments are partial with many side effects, and we lack detailed predictive models of how emotions form or decisions are made.

    “If you think about brain diseases, like Alzheimer’s and Parkinson’s and epilepsy, basically none of these can be cured, and the treatments — if they do exist — are very partial and have a lot of side effects.”

  • A conscious brain state is always caused by a previous state during which we were not conscious, so studying the brain circuit in action as a conscious sensation comes into being—and the unconscious 'roiling sea' of processing we have no access to—may reveal what consciousness is by showing how it arises, a view supported by Libet and Haynes finding brain changes up to 10 seconds before people feel they are deciding to move.

    “Presumably that brain state was caused by a previous brain state, and during that previous brain state we were not conscious. Otherwise, we would have called that the conscious experience.”

  • Because we don't know what consciousness is, there is no 'consciousness meter' to detect whether a being is conscious and no way to create consciousness from scratch, so we cannot even be sure another person (or a simulation) is conscious, and the Turing test is insufficient since systems like Siri and Alexa would pass without us granting them consciousness.

    “There’s no consciousness meter that says, “Oh, that being is a conscious being.” There’s no way to create consciousness from scratch, as far as we can tell.”

0.62

Precise tools like optogenetics and expansion microscopy reveal how brain circuits work at molecular resolution.

3 pointscentrality 5/5
  • Instead of zooming in to overcome light's resolution limit (~200 nm, far larger than nanometer-scale biomolecules), expansion microscopy infuses preserved tissue with a swellable polymer like that in baby diapers, softens the tissue, and adds water so the specimen expands evenly by 100x or 10,000x in volume while preserving the spatial organization of molecules, making structures visible on ordinary microscopes.

    “what if, instead of zooming in like they’ve done for 300 years, let’s blow it up. ... we infuse the specimens with the same kind of chemical that you find in baby diapers, a swellable polymer.”

  • Optogenetics works by borrowing light-to-electricity converting proteins (natural 'solar panels') from bacteria, algae, and archaea, delivering their genes into brain cells via gene therapy, so that neurons can then be activated or controlled with light, letting researchers speak the brain's natural language of electrical pulses to discover how to repair it.

    “we borrow from the natural world, effectively, tiny solar panels that exist in bacteria and plants and other small — ... They make these little molecules — proteins — that convert light to electricity. We transplant them into brain cells, and then we can control the electrical pulses of brain cells with light.”

  • Running the brain-expansion process in reverse—expanding the diaper polymer with water, laser-printing material inside it, then collapsing it down—produces a piece of nanotechnology, and Boyden's group used a silver photography-style chemistry this way to make metal nanowires.

    “take the baby diaper polymer, add water to expand it, and then you can basically laser-print stuff inside of it, and then collapse it down, and you get a piece of nanotechnology.”

0.53

Funding mechanisms and peer review must reward ambitious cross-disciplinary work despite skepticism from specialists.

5 pointscentrality 4/5
  • It is very hard to fund pioneering ideas: Brian Kobilka lost funding before winning a Nobel for the G-protein-coupled receptor (the target of one-third of all drugs) and moonlighted as an ER physician; Doug Prasher, who cloned green fluorescent protein used in ~a million studies, lost funding and drove a shuttle bus; and nine of Boyden's first ten expansion-microscopy grants were rejected before the Open Philanthropy Project rescued the project.

    “Brian Kobilka, who recently won the Nobel Prize for solving the structure of the G-protein-coupled receptor ... he lost his funding because he wasn’t making progress fast enough.”

  • Funding should be dynamic rather than allocated a year after application: when someone tries something one Friday afternoon that could cure disease or yield a major insight, funding should be reallocated up or down based on real-time metrics of scientific progress rather than waiting a year.

    “What if one could dynamically allocate funding up and down based upon the real-time metrics of science?”

  • Boyden helps people craft custom, mix-and-match curricula optimized for what they want to do—telling brain-science aspirants which physics (not gravity waves), chemistry, and math are essential and which can be skipped—so that someone, even a college dropout, can get up to speed in a year or two, embodying his theme of doing on purpose what was previously done accidentally.

    “I kind of designed a little custom curriculum for them, where it’s like you need to know some physics, but frankly, gravity waves — you don’t need that if you’re studying the brain.”

  • Daring, cross-disciplinary proposals fare badly in peer review because they sound off-the-wall or no reviewer is qualified to evaluate them; rather than simply combining reviewer opinions into in-or-out decisions, funders should step back and examine why peers think as they do—whether a critique stems from missing knowledge versus a rationale that is solidly linked to ground-truthable sciences like physics and chemistry—applying logical principles to improve proposal rankings.

    “if we take a step back and apply more logical principles of evaluation to the outcomes of peer review, can we actually improve the ranking of these proposals?”

  • Science prizes work well as storytelling that celebrates achievement for a broad audience, but as science increasingly involves large collaborations over long timescales—like the long chain of contributors to CRISPR—limiting a prize to a small number of people is a flaw, partly driven by the difficulty of telling a story with a hundred characters.

    “the number of people who contribute is getting larger and larger. ... the idea that you should limit it to a small number of people — I think it’s partly because of this limitation that it’s hard to understand a story with a hundred characters.”

0.48

Brain interventions inevitably affect multiple unrelated functions, complicating treatment without understanding underlying circuitry.

3 pointscentrality 4/5
  • The hard part of neurotechnology is the neuro part: even if we gain the ability to scan brains and read out information with unprecedented accuracy, without understanding the underlying processes and what the information means we might do the wrong thing with it, so the science must accelerate as fast as the technology.

    “Sometimes I say that the hard part of neurotechnology is the neuro part.”

  • Because there are only so many brain regions but millions of things we do, intervening in one region changes many functions: stimulating the FDA-approved dorsolateral prefrontal cortex site to treat depression with noninvasive magnetic pulses has also been shown to change unrelated things like trust and driving ability, and focus-enhancing drugs like Ritalin or Adderall may sacrifice creative mind-wandering.

    “There’s only so many brain regions, but there’s millions of things we do. Of course, intervening with one region might change many things.”

  • Mental illnesses are largely defined by symptoms rather than mechanisms, and pharmaceuticals treat them by bathing the entire complexly-wired brain in a substance, so they inevitably affect regions you want to change as well as regions you'd rather leave alone; the field instead needs to define brain conditions in terms of underlying wiring and circuitry.

    “it’s bathing the brain, this complexly wired circuit, in a substance. So of course, it’s going to affect parts of the brain that you want to change as well as parts of the brain that you’d rather leave alone.”

0.46

Decoding and reading brain states is possible but dangerous without understanding what the information means.

2 pointscentrality 4/5
  • In memory reconsolidation, recalling a memory makes it fragile, and studies show false memories can be induced—people shown something in a photograph or story will, a week later, sometimes remember it as their own experience—which complicates any neurotechnological lie detector because what someone says may be a false memory rather than an intentional lie.

    “when you recall a memory, it becomes fragile, and there are studies about how even false memories can be induced.”

  • Groups have for many years used functional MRI to read out what people are seeing—showing someone a movie, scanning their brain, and statistically guessing what they are viewing—and new advances like Columbia's 'mind vocoder' continue to refine this kind of brain-decoding.

    “You can show somebody a movie, scan their brain, and you can statistically guess what they actually are seeing.”

0.46

Self-regulation and proactive global neuroethics conferences should guide neurotechnology development responsibly.

2 pointscentrality 4/5
  • Neurotechnology carries serious ethical implications that should be addressed proactively through self-regulation and a global neuroethics conference bringing together companies, scientists, investors, governments, religious leaders, and lawyers, modeled on the 1975 Asilomar bioethics conference convened by Paul Berg that successfully self-regulated molecular biology and yielded major health benefits over the following half-century.

    “around 1975 or so, Paul Berg, who won the Nobel Prize for his work in biochemistry, convened in Asilomar, California, a bioethics conference ... They decided to actually get together and figure out how to self-regulate and regulate molecular biology.”

  • Because it is very hard to know what someone else is feeling and language doesn't fully convey internal states—a source of much conflict—Boyden's long-term personal goal is a technology that lets one person read out their mind and have another literally experience that state and vice versa, to make true empathy possible.

    “What if I could read out what happens in my mind, and somebody else could experience that literal state, and vice versa?”

0.46

Cross-disciplinary hiring of misfits between fields creates environments where genuinely new science can emerge.

2 pointscentrality 4/5
  • The MIT Media Lab thrives by hiring 'misfits' working at the intersection of disparate fields—a life science combined with another science, like Kevin Esvelt's CRISPR gene drives plus ethics, politics, and sociology—providing a home for people between disciplines where new fields can emerge, which is how Boyden, rejected by most faculty searches, got his start.

    “We’re the Media Lab. Maybe our new mission is to hire misfits. ... It’s a great place for people between one field and another where there’s sort of some space.”

  • Boyden hires by getting to know individuals deeply over a long period rather than by category, valuing variability over type—taking PhD students who never finished college, a Thiel fellow, a college dropout doing tech support, and a former professional photographer whose practical silver-based photography chemistry enabled a key metal-nanowire result—because unique backgrounds can change a field for the better.

    “I learn more from individuals and their variability than from categories of people. For example, in our group at MIT, I have two PhD students. Neither finished college, actually.”

0.44

Personal experimentation with meditation and productivity systems reveals how minds work and can be integrated.

13 pointscentrality 2/5
  • Nir Grossman, a postdoc in Boyden's group, found a way to focus the effects of electricity deep in the brain from the outside, which combined with optogenetically discovered activity patterns could create a noninvasive pipeline (movies, sounds, focused electricity) to repair the brain without the expensive gene therapy required in mice.

    “Nir Grossman, who was then a postdoc in the group — we actually found a way to focus the effects of electricity deep in the brain from the outside.”

  • A group in Utah found that when brain cells are active they turn on a gene resembling the coat protein of the HIV virus, and active neurons manufacture virus-like particles that can transfer genetic material from one brain cell to another, suggesting memory formation may involve unexpected molecular mechanisms beyond synapse connections and showing how little we understand brain function.

    “when a brain cell is active, it manufactures virus-like particles that can bring genetic material from one brain cell to another.”

  • If brains are computing devices, then the ecological diversity of species—zebra finches, mockingbirds, parrots, octopuses, dolphins—provides many different ways of computing the same thing, yielding interesting computational insights and aesthetic outcomes worth exploring across the evolutionary space of possibilities.

    “if you think of brains as computing things, then ecological diversity might provide many ways of computing the same thing but in different ways that actually yield interesting computational insights or aesthetic outcomes.”

  • Practicing Internal Family Systems meditation, where you treat the parts of your mind like members of a family and show compassion and gratitude toward the anxious or despairing parts, can be transformative because it makes those parts your allies rather than rebels, unlike other meditation forms that relaxed Boyden but did not integrate the parts of his mind.

    “the parts of your mind that might be driving you anxious or causing despair — you realize what they’re trying to do for you. You, by loving them and showing gratitude to them, can make them on your team rather than trying to rebel against you.”

  • The brain may be a bunch of independent agents that work together, as Marvin Minsky argued in The Society of Mind and as split-brain experiments suggest, and Internal Family Systems meditation explicitly treats parts of the mind as having their own drives, wants, and plans, which raises the open question of whether there is a central 'you' or just competing desires creating an illusion of a controller.

    “AI pioneers like Marvin Minsky wrote about The Society of Mind, about this idea that your brain is a bunch of independent agents that can work together.”

  • Li-Huei Tsai used optogenetic tools to discover a pattern of brain activity that improved symptoms in mice engineered to have Alzheimer's, and the team then found that a movie of blinking lights could induce the same pattern noninvasively; Boyden and Tsai cofounded Cognito Therapeutics to run human trials of such movies to treat Alzheimer's.

    “through a movie — a bunch of blinking lights and so forth — you could simulate the same pattern of activity. Li-Huei and I have now cofounded a company, Cognito Therapeutics, to effectively do human trials of movies to treat Alzheimer’s.”

  • Whereas most antidepressants take three weeks or more to work, ketamine — a hallucinogenic drug — can help depressed patients feel better within tens of minutes to hours, showing that a hallucinogen can have a rapid-acting effect on a psychiatric illness.

    “Most antidepressants take three weeks or more to help people. Ketamine will help within tens of minutes to hours.”

  • Science may seem to be slowing because we are tackling combinatorially harder problems: unlike physics (few building blocks and interactions) or chemistry (a few dozen useful elements and bonds), biology has ~30,000 genes whose interactions, architecture into cells/tissues/organs, and failures we barely understand, producing a 'cognitrone explosion' of possibilities so vast that most attempts fail—but better tools to map building blocks and interactions could make 21st-century sciences as tractable as lasers and computers made 20th-century science.

    “In biology, we have what, 30,000 genes in the human genome, and while we know their sequence, for the most part, we have no idea how these gene products interact with each other ... The problem is this cognitrone explosion of possibilities is so staggeringly huge that a lot of what we try will fail.”

  • Around age eight or nine Boyden became obsessed with whether the meaning of life could be addressed through science, which made him ambitious and focused; he started college at 14 (working on an origins-of-life project), focused on chemistry for two years, then transferred to MIT for physics and electrical engineering and quantum computing.

    “Around age eight or nine I got obsessed with this question of can we address the meaning of life through science ... I started college, and I started working on this origins of life project when I was only 14.”

  • Careers have a skill phase, where you learn to program, machine metal, or do mathematics, and an impact phase, where you go solve a problem; whether you need to enter a laboratory depends on your path—Boyden's chosen skills and impact required getting his hands dirty with real matter, whereas mathematically inclined people can do great work on their own.

    “A lot of us have sort of a skill phase of our career ... Then at some point, you want to go solve a problem. I call that the impact phase.”

  • Boyden treats his own career as an experiment to learn what he is good at and enjoys—arriving at early waking and his time-management system by experimentation—and keeps not only a forward calendar but a calendar of the past logging how long tasks actually took, which over time made him much better at estimating durations and managing future tasks.

    “I also keep a calendar of the past, which logs how long it actually took for me to write that grant, or to meet with that person ... Over time, I’ve learned a lot about how long it takes me to do something.”

  • Neither fully open offices nor fully closed spaces is the right design for science; instead an 'ecosystem of environments' is best, and Boyden gets substantial productivity from deliberately switching between the Media Lab's open spaces and a classical closed-office neuroscience lab, finding inspirational architecture and connecting dots across disciplines productive.

    “maybe neither is the right approach. You might want to think about having sort of an ecosystem of environments. ... I actually get a lot of productivity out of switching environments in a deliberate way.”

  • Biological systems did not evolve to be understood—they evolved to survive and reproduce—so we cannot assume they are tractable, yet biology gives us more structure than we deserve (DNA's double helix, the readable genetic code), even as we may now be entering a realm of complexity that is less understandable and must accept reality as it is.

    “systems didn’t evolve to be understood. They evolved to survive and reproduce and all that. One can hope for structure. Biology does give you more structure than we deserve”

0.37

Science advances through forgotten papers, morphological analysis, and failure rebooting when conditions change.

4 pointscentrality 3/5
  • Search engines like Google surface the most popular things, but advancing science requires the opposite: a search engine for the most important unpopular things—and deliberately injecting an unrelated random word into searches can surface useful cross-field connections.

    “to advance science, what we almost need is a search engine for the most important unpopular things.”

  • Major breakthroughs often come from serendipity and ignored basic curiosity—CRISPR from yogurt scientists, fluorescent proteins from a jellyfish obsessive, backpropagation worked out in the 1980s-90s but enabled later by compute and data—so funders should deliberately hunt for treasure hiding in plain sight, possibly via a SWAT team of people who connect dots across fields and dynamically assemble committees of non-overlapping experts.

    “If you look at CRISPR for genome editing — that was found by some scientists working on yogurt. If you look at fluorescent proteins — that was identified by a person who just was obsessed with jellyfish.”

  • Boyden's research method is to think backward from a major problem so the tools you build actually solve it, then to enumerate every possible solution by taking the base of possibilities (e.g., all kinds of energy that could control the brain) and recursively splitting it into a tree diagram whose leaf nodes are testable ideas—a technique Fritz Zwicky called morphological analysis in the 1930s and credited for ideas like dark matter and gravitational lensing.

    “I call this method the tiling tree method, where you try to take this base of possibilities and break it down into parts. ... the leaves of the tree — the bottom nodes — are ideas you could actually test”

  • Rather than being the 100,000th person on a popular topic, Boyden reads old, forgotten papers because treasure hides in plain sight—both optogenetics and expansion microscopy began from ignored old papers from other fields—and he practices 'failure rebooting,' reviving previously failed ideas once something has changed (a new gene, faster computers, a discovery from left field), as with attempts to control brain cells with light dating back to the 1940s-1971 that only worked once the right light-sensitive molecules were discovered.

    “I sometimes practice what I call failure rebooting. We tried something, or somebody else tried something, and it didn’t work. But ... Something happened that made the world different. ... that old failed idea might be ready for prime time.”

0.34

Understanding the full brain ecosystem including the body's microbes is necessary for accurate brain emulation.

3 pointscentrality 3/5
  • The brain is part of an ecosystem—the whole body—that computes together: individual short-timescale thoughts may be contained in the brain, but over longer timescales gut microbes can secrete molecules that enter the brain and modulate complex functions, possibly even social behavior, so whole-brain emulation may require modeling the body as an ecosystem to capture long-term moods, memory, and personality.

    “there might be microbes in the gut that secrete molecules that can actually get into the brain and modulate complex functions, like maybe even social behavior, some people think.”

  • In a cubic millimeter of brain there are about 100,000 brain cells with roughly a billion connections between them, and mapping not just the wiring but which biomolecules sit on each connection (revealing whether connections are fast, slow, strong, or weak) is needed to understand how information flows and is transformed.

    “In a cubic millimeter of your brain, you have about 100,000 brain cells with about a billion connections between them.”

  • If we had a detailed map of brain wiring plus the location and identity of key biomolecules, a biophysically accurate computer simulation of a brain might someday be possible, as Eve Marder has already achieved for very small circuits like the crab's stomatogastric ganglion.

    “for very small neural circuits, like in the crab’s stomatogastric ganglion, Eve Marder at Brandeis already has been able to do biophysically accurate, human-understandable simulations.”

0.32

Specific neural clusters can trigger complex behaviors like aggression when activated with precision.

2 pointscentrality 3/5
  • Dayu Lin, David Anderson, and colleagues put a blue-light-activatable molecule into a deep brain region in mice and aimed an optical fiber laser at it; when the laser turned on, the mice became aggressive and would attack whatever was next to them, even a rubber glove, demonstrating that complex behaviors like aggression can be triggered by activating specific neural clusters.

    “When they turned on the laser, these cells were activated, and this was done in mice. The mice would become aggressive or violent. They would attack whatever was next to them, even if it was a rubber glove.”

  • The point of precisely controlling neural activity is not spectacle but discovery: if you could precisely control neural activity you might learn the principles of how to repair a brain, which could inspire better drugs or better noninvasive brain stimulation methods.

    “if you could precisely control neural activity, maybe you could actually repair a brain or discover the principles of how to repair the brain, which could inspire better drugs or better noninvasive brain stimulation methods.”