Ed Boyden on Minding your Brain
Ed Boyden — neuroengineer at MIT and pioneer of optogenetics and expansion microscopy — joins Tyler Cowen in Ep. 64 to walk through the two technologies that have made his lab famous, the hard problem of consciousness, the ethics of neurotechnology, and the working methods behind a career spent building tools other people had declared impossible to fund.
Key ideas
- We understand the brain far less than the popular science literature suggests. Alzheimer’s, Parkinson’s, and epilepsy remain uncurable; most brain conditions are defined by symptoms rather than by the underlying circuitry. The treatments that exist are blunt — bathing a complexly wired system in a chemical — because we lack the maps to do better.
- Optogenetics inserts light-sensitive proteins from algae into neurons, allowing specific brain circuits to be switched on or off with light. The technique exploits a biological accident: light-to-electricity converters that evolved in single-celled organisms also work in mammalian neurons when the gene is delivered by viral vector. In mice, this has already been used to study aggression, memory, and the precursors of Alzheimer’s symptoms.
- Expansion microscopy solves the resolution problem by physically enlarging tissue rather than improving lenses. A piece of preserved brain is infused with the polymer found in nappy gel, then hydrated; the specimen swells to one hundred times its original volume while preserving the spatial arrangement of its molecules. The result is that a standard optical microscope can resolve structures that would otherwise require expensive, slow super-resolution hardware.
- Whole-brain emulation is conceivable but blocked by the unsolved problem of consciousness. Boyden is optimistic that a sufficiently detailed wiring map would eventually allow a biophysically accurate computer simulation, but he concedes there is no consciousness meter — no way to verify whether the simulation would experience anything, or whether even the biological original does.
- His lab’s method for generating ideas — the ‘tiling tree’ — converts an open-ended problem into an exhaustive combinatorial search. Starting from a goal (control the brain without killing it), he enumerates every category of physical energy that could in principle do the work, then branches each category until the leaves are testable experiments. The same technique, formalised as morphological analysis by astronomer Fritz Zwicky in the 1930s, is how optogenetics and expansion microscopy were both conceived.
Content
What we don’t know about the brain
Cowen opens by asking what a reader of all the good popular neuroscience books would still fail to understand. Boyden’s answer is disarming: the main thing they would miss is how little neuroscience actually knows. No major brain disease can be cured. Existing drug treatments work by flooding the brain — an organ with roughly a billion synaptic connections per cubic millimetre — with a chemical that affects wanted and unwanted circuits alike. The reason is a mapping problem: we cannot yet describe, in enough detail, the wiring that generates a specific emotion or decision, let alone the wiring that goes wrong in a specific disease. Mental illnesses compound this because they are currently defined by their symptoms rather than their biological substrate, which makes it hard to build targeted treatments.
Optogenetics — writing neural activity with light
The central ambition behind optogenetics is precision: to switch individual populations of neurons on or off without disturbing their neighbours. The solution borrowed from nature. Certain single-celled organisms — algae, archaea — carry proteins that act as tiny solar panels, converting light into electrical current. Because proteins are encoded by DNA, it is possible to use the established tools of gene therapy to deliver the relevant gene into specific brain cells, which then express the protein and become light-sensitive.
The delivery is viral. Light is delivered through an optical fibre implanted near the target region. When a laser pulses, the modified cells fire; everything else is unaffected. In one study Boyden describes, activating a small cluster of cells deep in the brain caused mice to become immediately aggressive — they attacked whatever was nearby, including a rubber glove — demonstrating that the behaviour could be traced to a precise circuit, not a diffuse mood state.
The near-term therapeutic pipeline runs through the same logic: use optogenetics in animals to discover exactly which pattern of activity is therapeutic, then translate that pattern into a noninvasive delivery method — a flickering movie, a patterned sound, or focused electrical fields from outside the skull. Boyden has co-founded Cognito Therapeutics to run human trials of this approach for Alzheimer’s disease.
Expansion microscopy — inflating biology to see it
Standard light microscopes cannot resolve objects much smaller than the wavelength of visible light — roughly 200 nanometres. Biomolecules are 1–2 nanometres across, about a hundred times smaller than the diffraction limit. Super-resolution microscopes work around this by optical tricks, but they are slow and expensive, and struggle with three-dimensional samples. Boyden’s group took the opposite approach: instead of improving the optics, expand the sample until it is resolvable by ordinary optics.
The technique embeds preserved brain tissue in the same swellable polymer used in disposable nappies. Treated correctly, the polymer infiltrates the specimen and, on hydration, expands evenly in all three dimensions by a linear factor of roughly four (a volume factor of around 64, which can be compounded to 10,000-fold with additional steps). The spatial relationships among molecules are preserved because every part of the specimen swells uniformly. The result is that a standard confocal microscope can now read out the positions of proteins, RNA transcripts, and synaptic structures that were previously invisible. Hundreds of research groups now use the technique to build wiring maps in small organisms — fish, flies, worms — as a path toward eventually mapping a mouse brain and, one day, a human one.
The wiring map and brain emulation
A complete wiring map — called a connectome — would specify not only where each neuron’s connections land but which molecules sit at each synapse, determining whether a connection is fast or slow, excitatory or inhibitory. With this information, Boyden thinks it would eventually be possible to build a biophysically accurate simulation capable of reproducing a brain’s decisions and emotions in software. Small precedents exist: Eve Marder at Brandeis has already produced accurate, human-understandable simulations of the stomatogastric ganglion (a simple circuit that controls digestion in crabs), and the principle generalises.
Boyden is careful, though, on what such a simulation would actually be. Since we have no way to detect consciousness — no instrument, no agreed test — we cannot know whether a simulation would be conscious, whether the biological original is conscious, or what the relationship between the two might be. The Turing test (can you tell, in conversation, whether you are talking to a machine?) he considers insufficient, given that Siri and Alexa already pass a naive version of it. Consciousness, he suggests, might become tractable only by catching the brain mid-process: mapping the neural circuits in the moment when a sensation crosses from unconscious to conscious, and trying to understand what transition occurred. This is early-stage thinking with no supporting data, but it is the research direction he finds most promising.
The picture is complicated further by the body as a whole. Gut microbes secrete molecules that cross into the brain and may influence social behaviour; long-term personality and mood likely depend on more than the brain’s own circuitry. Whether whole-brain emulation would need to model the entire body — or an entire ecosystem of human relationships — is an open question.
Mental illness, psychedelics, and meditation
On mental illness, Boyden’s position is consistent with his general stance: we are trying to treat diseases we cannot yet define biologically. The pharmaceutical approach — adjusting neurotransmitter concentrations brain-wide — works in some cases but is inevitably coarse. Ketamine, an anaesthetic that causes hallucinations at high doses, is the most striking recent exception: it relieves depression within hours rather than the three weeks typical of conventional antidepressants, suggesting that the therapeutic mechanism is something other than what those drugs were thought to be targeting.
Boyden meditates daily using a practice called Internal Family Systems, which treats the distinct drives and anxieties of the mind as independent agents — members of a family — and cultivates compassion toward them rather than suppression. He finds it transformative, though he is candid that he cannot rule out a placebo contribution. The question he raises is whether interventions that make the mind’s parts more cooperative might also reduce productive tension and creativity; he does not answer it, but regards it as a question that will become increasingly urgent as neurotechnology advances.
Consciousness and the self
On whether the mind is reducible to the brain, Boyden is a cautious physicalist: he believes brain chemistry is in principle understandable in physical terms, but he does not yet know what to do with consciousness. His own approach is procedural: if he can map the neural dynamics that precede a conscious moment, perhaps he can understand the causal process that generates it. Work by Benjamin Libet and John-Dylan Haynes on voluntary movements is suggestive — brain activity predicting a movement can be detected up to ten seconds before the subject feels they have decided to move — but Boyden does not claim this resolves the hard problem.
He finds Marvin Minsky’s ‘Society of Mind’ framing useful: the brain may be a coalition of relatively independent agents, with what we call ‘self’ emerging from their interactions rather than residing in any one place. The Internal Family Systems meditation he practises is the experiential correlate of this view — treating the parts of the mind as distinct and negotiating with them, rather than seeking a single executive. He stops short of endorsing Parfitian scepticism about personal identity, but he regards split-brain experiments as genuine evidence that there may be no unitary conscious subject.
The ethics of neurotechnology
Boyden’s worry about brain-reading and brain-writing technology is not primarily about misuse by bad actors, though he acknowledges that risk. It is about an epistemic gap: we are developing the ability to read out brain states with increasing fidelity, but we do not yet understand what those states mean. A ‘mind vocoder’ that reconstructs speech from brain signals may misclassify a false memory as an intentional lie; stimulating the dorsolateral prefrontal cortex to treat depression also, studies show, changes trust and driving ability in ways that patients were not warned about.
His proposal is a global neuroethics conference on the model of the 1975 Asilomar conference on recombinant DNA — a gathering of scientists, investors, governments, religious leaders, and lawyers to discuss what research should and should not be done, before the headlines arrive. He acknowledges that self-regulation cannot guarantee perfection, but takes Asilomar as evidence that early, structured deliberation produces better outcomes than reactive governance.
He notes a cultural obstacle: people are reluctant to discuss their brains. At a conference where 20 per cent of attendees anonymously reported using cognitive-enhancing drugs, nobody raised their hand when asked publicly. A new language — one that allows people to talk about their brain without separating it from their self — may be a precondition for the public conversation the technology requires.
How science should be funded
Boyden’s own expansion microscopy project was rejected by nine of its first ten funders before the Open Philanthropy Project stepped in. He draws a wider lesson from the history of science: the most consequential projects often fail peer review precisely because they are novel, because they span fields that no single reviewer is qualified to judge, or because the component ideas — individually well-attested — have not yet been assembled into a coherent proposal.
His three prescriptions for reform are: first, evaluate the logic of proposals rather than their surface plausibility to domain experts, using committees with genuinely non-overlapping expertise; second, allocate funding dynamically — if an experiment yields an unexpected result on a Friday afternoon, the money should follow within days rather than years; and third, go looking for ideas hiding in old literature and unfashionable fields. Optogenetics drew on light-sensitive proteins studied in single-celled organisms; expansion microscopy drew on polymer chemistry from the 1980s; CRISPR was found in yogurt cultures; green fluorescent protein came from someone obsessed with jellyfish. In each case, the breakthrough was a connection between an ignored biological fact and an urgent engineering problem.
The Ed Boyden production function
Boyden started college at 14 at the University of North Texas, motivated by a childhood conviction that philosophical questions about the meaning of life were tractable through science. He trained across chemistry, physics, electrical engineering, and neuroscience — a six-year undergraduate degree — and regards the cross-disciplinary grounding as the source of his ability to stitch together facts from different fields. He would not recommend the route for everyone, but he designs custom compressed curricula for students who come to him with unconventional backgrounds: two current PhD students in his group never finished their undergraduate degrees.
His lab takes the same approach to personnel. He has hired a professional photographer (whose silver-chemistry knowledge turned out to be exactly what a nanotechnology project needed), neurosurgeons, and college dropouts. The principle is that unique backgrounds create unique research ideas; the lab’s explicit policy is not to have formulas.
On personal habits: he wakes at 4 a.m. to get four or five hours of uninterrupted thinking time before the day’s interruptions begin. He keeps a detailed retrospective calendar logging how long things actually take him, which he uses to calibrate future estimates. He switches between environments deliberately — open Media Lab space for connecting ideas, closed laboratory space for quiet — on the same principle that his office architecture advice to scientists is not ‘open plan’ or ‘closed’ but ‘an ecosystem of environments.’
His idea-generation method, which he calls the tiling tree, was formalised by Fritz Zwicky in the 1930s as morphological analysis. Given a goal, you enumerate every distinct category of approach, branch each category into sub-categories, and repeat until every leaf is a testable hypothesis. The tree for optogenetics started with the question ‘what kinds of physical energy could control brain cell activity without killing it?’ and worked down through light, magnetism, mechanical force, and electricity until it reached the class of light-sensitive proteins that turned out to be the solution.
Overrated / underrated
Waking at 4 a.m. is, for Boyden, clearly underrated for those who value uninterrupted thinking time, though he does not claim the schedule suits everyone; he drops it at conferences when the valuable conversations happen late. Science prizes he rates positively as a storytelling mechanism — a way of communicating the value of research to a broad audience — while noting that attribution in collaborative modern science is genuinely contested, as the CRISPR disputes illustrate. Neuroeconomics he finds exciting if still early. He declines to rate birdsong as music without noting that his first neuroscience experiment, in 1998, was on zebra finches — not famous for melodiousness.
Related
- Ed Boyden — speaker; neuroengineer at MIT, pioneer of optogenetics and expansion microscopy
- Tyler Cowen — host
- Consciousness — concept; directly discussed: the hard problem, the Turing test, and whether emulation could be conscious