What Science Is Left When Institutions Decide What Cannot Be Said

Lex Fridman opens with a claim about 2020 rather than a guest bio. Science, he says, is beautiful as a striving of the mind and fragile as an institution — susceptible to fear, greed, power, and ego. What the year required, and did not get, were leaders willing to talk honestly about uncertainty, about the studies still running, and about the cost of scaling any given tool. Masks, home testing, anonymized contact tracing, antivirals, and vaccines all belonged on the table. No voice should have been ruled out in advance.

There are plenty of ideas that are bad, wrong, or dangerous, Fridman grants. The moment a society decides it already knows which ones those are, it loses the only process that can find the ones that are not. He asked Bret Weinstein on the show, he says, less to ratify a shared list of conclusions than to show solidarity with the freedom to talk. Weinstein is an evolutionary biologist, author, and co-host of DarkHorse who describes himself as a reluctant radical. Fridman does not agree with him on everything. He is sure he is glad Weinstein exists.

What follows is a long argument that the pandemic was not only a public-health failure. It was a test of whether science, as a method, still had a public home once agencies and platforms began treating official guidance as truth and dissent as harm. Weinstein's channel had been restricted for saying things official bodies later walked back. The conversation keeps returning to that mismatch: the people enforcing the line were the ones who had just been wrong.

Asked what is beautiful about biology, Weinstein does not start with a theory. As a child he was fascinated by animals; as an adult he is drawn to what he calls a never-ending series of near-miracles across living nature. The mind, he says, timeshares among modules that do not yet connect: the evolutionary dynamics of a lineage, the deep-time view of that lineage, and the mechanisms that would have to exist inside to produce what we see from outside. Humans are a generalist platform that can swap software and occupy niches ancestors never saw. Mastery, whether parkour or skiing, is the trick of driving a new program down into the unconscious so the conscious mind can watch.

Fridman has been thinking about Boston Dynamics robots that do parkour and dance while remaining, in his view, hard-coded. What makes a human a human is the process of figuring it out, including the mistakes. He wants Spot to be allowed to fail in a constructed safe space, to see, to learn how to use a body, to discover something like being alive. Weinstein's reply is that the missing element in robotics and AI is a proper development: it is no coincidence that the most dominant species has the longest childhoods. Adult capability is the flip side of helplessness at birth. Consciousness, on this telling, is uncompiled code — slow, adaptable, and then driven down so it can run fast. After that, the conscious mind's job is to notice what the compiled program did not expect.

Fridman describes a louder problem: a self-critical megaphone that yells after every sentence, useful for production and brutal for gratitude. Weinstein maps it onto ancestral programming that no longer serves. Dwelling on an error is how you upgrade the program; dwelling past the point of utility is how you become Kafka's insect. He lives by what he calls a theory of close calls. A friend stepping off a curb who is nearly hit at 45 miles an hour did not have a non-event. The difference between that walk and death was luck, and luck is not a strategy. Count it as near-death and recode so it cannot happen again.

Weinstein wants the same accounting at civilizational scale. Accidents and disasters describe a distribution of what is likely to get us in the end. Fukushima's triple meltdown, the 2008 financial collapse, the Deepwater Horizon blowout, and COVID-19 with its probable origins in a Wuhan lab are, in his list, events that told us we were involved in an industrial process only after it had gone wrong. The wise question is which unnamed processes we will learn the name of at the next accident — and whether we can talk about the wisdom of running them before that day.

Each of those episodes, he says, could have been much worse. The fuel pools at Fukushima held. A Herculean effort pulled the financial system back from the brink. COVID, terrible as it is, is also a close call: if labs keep playing this game, especially under the story that a virus is about to leap from nature any day, they will create the disaster sooner. Intelligence can still produce these events. Wisdom would be to stop causing them. He does not see a political structure capable of deploying a protective algorithm, because the economic consequences would get it shot down.

Sustainability is hard to operationalize and still, he thinks, an objective a long-lived civilization has to meet. So is reversibility. The first internal combustion engine on a street would have made a climate warning sound preposterous; scale changed the meaning of the toy. If no process could be started that could not later be reversed, progress would slow and safety would rise. We are involved in too many very dangerous processes to keep skipping that thought.

Fridman asks for a probability that COVID-19 leaked from the Wuhan Institute of Virology. Weinstein keeps a flowchart. The likelihood it emerged from a lab is very high; if it did, the likelihood that lab was the Wuhan Institute is very high. Multiple kinds of evidence point to a lab, he says, and literally no evidence points to nature — either the evidence points nowhere or it points to the lab. Geographically, Wuhan is the place to start, and the lab most involved with viruses that look like SARS-CoV-2 is the place one would start first. He puts a lab origin well above 95%. An unmodified virus brought in and escaped is possible; it does not explain the anomalies in the genome. Wuhan as a false-flag distraction from another lab sits, for now, below 1%.

What would count as evidence for a natural origin? Not the historical fact that viruses jump — that is a presumption, and a presumption evaporates in the face of evidence. It did not evaporate here, he says, because people who privately emailed their grave doubts about a natural origin kept the presumption alive in public. The evidence that would reverse his flowchart is a human population in which the virus circulated before Wuhan, learning the tricks it already had when it spread, or an animal population with an ancestor epidemic. The second path should have left a great deal of early evolution. We do not see it. Becoming a pandemic virus, he argues, is two very hard jobs at once: infecting a person thoroughly enough to take, and jumping between people fast enough not to go extinct. Selection would leave a mark. A sudden perfect virus is a hopeful monster, and hopeful monsters do not work.

From the genome and from papers out of the labs that built the technology, he thinks SARS-CoV-2 is not entirely the product of splicing or of serial passaging. It looks like both. The furin cleavage site looks added in, and it was already known that such a site would raise infectivity in humans and widen tropism. The virus spreads well in humans, mink, and ferrets; ferrets have an ACE2 receptor much like ours and are the obvious passaging animal. Humanized mice and human airway tissue are also on the list of materials that may have been used. If it came from Wuhan, the protocols matter because they would be tools for fighting the virus. If it came from nature or from another lab, the world still needs to know which.

Fridman had been with Sam Harris in one respect: a deadly lab leak will happen eventually, so the future is the emergency even if this virus is not the worst possible one. Weinstein is freaking out about both clocks. People were warning about this research in 2015; a ban was bypassed and the work was offshored to China. That, he and his wife Heather have said from the beginning, does not look like a Chinese failure. It looks like a failure of the international scientific community. The other clock is whether SARS-CoV-2 becomes a permanent fact of life, doubling something like the yearly cost of influenza. He believes the virus could have been driven extinct months earlier and was not, because of mundane concerns among a small number of people running on autopilot — including pharmaceutical portfolios that needed competing therapies not to be noticed.

Fridman asks Weinstein to steel-man the case against even discussing a lab leak, including the charge that the conversation is racism against Chinese people. Weinstein's limit on steel-manning is that you can only steel-man a good-faith argument. Private emails, he says, show the public campaign was punishment — backing people down, or warning others not to follow. For people who think science is the best hope for navigating hard water, shutting down those who used its tools honorably is itself dishonorable. At the point Nicholas Wade published and the world, with no new evidence, admitted a lab leak was at least possible, the rationalization on display was that Chinese scientists were being protected. Weinstein does not want anyone targeted. He also does not accept a license to lie.

Once you create a license to lie in order to protect individuals, while the world has a stake in knowing what happened, that license will be used by whatever captures institutions. Heather and he have taken pains to say this looks like a collective scientific failure, not a national one. That message should travel with the evidence. It is not a reason to keep the evidence off the table. The claim that the well-being of a set of scientists outweighs the well-being of the world is, in his word, preposterous. The same scientists had argued that zoonotic spillover was so likely that the research had to be run. If they caused the event and were wrong about the natural risk they wrote into grant applications, the call is coming from inside the house.

Demonizing Weinstein, Josh Rogin, Yuri Deigin, and the DRASTIC group for following publicly available evidence teaches a worse lesson than any short-term panic: thou shalt not use scientific tools. Fridman has watched the effect at MIT. World-class biologists feel pressure not to speak and, more strangely, not to think — they avoid looking at the data. Some say the public cannot be trusted with the complexity and that the fight belongs in private. Combined with academic bullying, the result is self-censorship of the one resource those people have: their brains. Weinstein calls it a cryptic totalitarianism, and worse when it becomes rationalization that blinds people to what they can see. Their discussions with each other about why they did not say what they understood: that is what capture sounds like on the inside.

He does not claim to know the force that captured the institutions. It may have been a process rather than a cabal. On the inside, people tell themselves the public cannot have this discussion, that it has to stay private, and those stories are how an institution avoids returning to its mission. From lab leak to repurposed drugs, he says, you can see the deadliness in a very direct way. Fridman recognizes the same flocking in AI around autonomous weapons: discomfort, a scatter of sheep, the feeling that one person cannot do anything. In viruses the bill has already come due. Weinstein's older concept for the academic version is cultivated insecurity: when mortgage, insurance, and retirement hang on one employer, other people get to tell you how to think. The way out is enough security that going against the grain is not a family catastrophe.

The day before this recording, Weinstein had been on Joe Rogan's show with Pierre Kory of the Front Line COVID-19 Critical Care Alliance, in what Rogan called an emergency podcast. Weinstein is undescribably grateful Rogan took the risk. They talked about a censorship campaign against ivermectin that he finds utterly shocking, and about a drug whose story, he notes in passing, is not a simple morality play about Asia: a likely lab leak in China and an unlikely hero in Japan.

Ivermectin was found in the 1970s by Satoshi Ōmura in soil near a Japanese golf course and sent to a Merck scientist named Campbell. They shared the 2015 Nobel Prize. It treats the worm that causes river blindness, the pathogen behind elephantiasis, and scabies. It is extremely safe, on the WHO list of essential medicines, used in children, given away in millions of doses by Merck in Africa, and administered something like four billion times in four decades. One reason Africa may have had a milder COVID wave, Weinstein suggests, is that the drug is widely used there against parasites and appears to have a long-lasting effect. Antiviral properties showed up later. Early in the pandemic it was tested in tissues at a high dose that still confuses people — advocates, he says, are not pushing those lab doses in humans — and clinicians using it reported that it worked, especially if you got to patients early.

A meta-analysis had been circulating as a preprint and, in the week of this conversation, was finally peer-reviewed and published. Tess Lawrie of Britain's BIRD group is a second author. It showed, Weinstein says, that the drug is highly effective at treating people with the disease and 86% effective as a prophylactic. That number is high enough, given an R0 slightly above two, to drive SARS-CoV-2 toward extinction if anyone wished to deploy it. Lex points to a separate living meta-analysis of 60 studies at sites such as ivmmeta.com; Weinstein says the newly published paper is not that dashboard, though he likes the dashboard too. He is not a physician. He is a biologist on ivermectin in lieu of vaccination. Weekly dosing for prophylaxis is not the historical pattern, so long-term effects are a real question, and no drug is perfectly safe. Dr. Kory had told him of calculation errors that produced massive overdoses with no ill effect. The mystery is why a sufficiently promising signal was not investigated harder — and why discussing it was treated as a problem to be removed.

Why did YouTube and other platforms censor discussion of ivermectin? Weinstein splits the stated reason from the actual one. YouTube sits in a consortium with Twitter, Facebook, Reuters, AP, the Financial Times, the Washington Post, and others that appointed themselves arbiters of truth to stop misinformation. In this case they chose to treat WHO and CDC recommendations as synonymous with scientific truth. Even at their best, he says, those bodies are not scientific entities. They are public-health entities, and public health has assigned itself a right to lie.

The textbook justification is game theory around vaccination. If everyone else takes a shot and the herd is immune, the person who opts out takes no risk and still benefits, so public-health agencies oversimplify to stop a tragedy of the commons. Weinstein disagrees with that right even in the narrow case. Once it exists, it turns out to serve pharmaceutical companies that hold emergency-use authorizations requiring that there not be a safe and effective treatment, and that enjoy immunity from liability for harm from their product. You do not need to be a complex-systems expert, he says, to see the hazard of immunizing a company from its own product's harm while that product can exist on the market only if a better one fails to be noticed.

He does not need a signed contract between YouTube's board and Merck. This may be another autopilot, parroting the corrupt wisdom of WHO and CDC because that feels like doing the right thing. The irony is the record. Those agencies, he says, were wrong again and again in the pandemic, and his channel was right where they were wrong. YouTube then came after him because WHO and CDC disagreed with him. Fridman has heard the same casual line inside YouTube: slowing misinformation is easy, and a Google search will tell you which institutions are the arbiters. The outsourcing is the tell. The people with a worse track record wrote the rules for the people with a better one.

If Weinstein ran YouTube for a month, the rule would be almost embarrassingly simple: the limit of what is allowed is the limit of the law. If speech is legal, it is not YouTube's place to decide who may say it or who may hear it. Harm would follow. Net harm, he thinks, would not, because only when everything can be said is there a process in which better ideas win. The American Founders did not imagine that most speech would be high quality or harmless. They reasoned that the benefit of leaving everything sayable exceeds a substantial cost. What they could not have anticipated was the centrality of YouTube, Facebook, and Twitter. Had they seen it, they would not have written the First Amendment as a restraint on Congress alone. Those platforms may now be more powerful than the federal government of the Founders' worst nightmares.

Fridman pushes on the emergency case. Algorithms maximize drama, memes, and anger; those things are indifferent to truth, and in a crisis that is lethal. Weinstein grants the worry and then drops it into an older problem. Humans are the blankest slate nature has devised. A large fraction of behavior lives in culture and consciousness, not in hardware, and that layer produces mountains of garbage. Children make absurd errors; so do comment sections. The high birth rate of nonsense does not mean that what sticks over the long term is worthless. Totalitarianism has arisen many times and done carnage. Over time, he still thinks open, fairer societies are what freedom of ideas produces. He cannot prove it. He has not heard a better idea.

The hubris in imagining you will improve discussion by censoring, he says, is that most concepts at the fringe are nonsense — automatically. The heterodoxy at the fringe is indistinguishable from that nonsense at the beginning, and it is the key to progress. Declare the fringe 99% garbage and delete it, and you have thrown out the key. That is what YouTube was doing. In the immediate aftermath of a major reversal on the lab-leak question, it started restricting his channel. It should have looked in the mirror: we are censoring somebody who was just right, in a conflict with the very people on whose behalf we are now censoring. His brother Eric's caveat is one he accepts: ideas are not doxxing, and destroying lives is not a contribution even if a lawyer can defend it. Where to draw that line, he does not know.

Fridman would rather see better algorithms than hard-coded bans: a full spectrum of views, including the best biologists on vaccine risk next to the best accounts of how mRNA works, without manually inserted CDC spots. Weinstein is not ready to trust this consortium with curation after watching it fail at evidence. He wants search that is not personalized into a private derangement, plus tools to see what other perspectives exist and to read a steel-man of the other side — plate tectonics before it was respectable, not a cartoon. Under present incentives, he thinks, a product that optimized for long-term growth of the user would be a terrible business, because it would be decimated by people who play to the hour. Markets, in the book he is finishing with Heather, are superb at how and useless at what. Tell them what human well-being requires; do not let them plot the course. The name they give the destination is the fourth frontier: growth that feels like growth without depending on actual expansion, so that zero-sum politics does not turn into war.

Fridman took the two-shot Pfizer series. Weinstein did not; he has been taking ivermectin. One argument against discussing the drug, Fridman notes, is that it makes people less willing to get vaccinated. Weinstein wants two hazards on the table before anyone talks about individual side effects. First, a vaccine this narrowly focused on a single antigen is an evolutionary pressure that can drive variants which escape the protection. Breakthrough cases make that worse: a vaccinated person may shed viruses that are less visible to vaccine-induced immunity. Applying that pressure mid-pandemic, while people are still encountering the sick, gives the virus a vast canvas on which to practice escape. Second is antibody-dependent enhancement, the dengue pattern in which a second infection is far worse because the virus uses prior antibodies. That pattern, he says, has blocked past attempts at coronavirus vaccines. Whether it happens here is still too early to say. It is not clear, in the way people think it is, that vaccinating enough people ends the pandemic. It could make it worse.

The special feature is not mRNA as such. The mRNA shots and the adenovirus-vector DNA vaccines share a vulnerability: they aim at one subunit of spike, a very concentrated signal. A traditional vaccine typically presents a virus full of antigens; people who have had COVID have a broader response for the same reason. The new platforms also carry parts not used in people before, including lipid nanoparticles that distribute around the body with unknown consequences. Asked to steel-man universal vaccination, Weinstein does it cleanly. Nothing is perfectly safe; phase-three trials looked like high efficacy and high safety; dropping the landscape of available hosts fast enough for herd immunity to finish the job is the highest priority; overwhelmingly fewer people die if the population is vaccinated than if it is not; the most likely path is that the virus fades. He does not believe the probability of a worse pandemic is low enough to discount, and he thinks a wave of variants already looks disturbing.

Fridman detonates on the data. The U.S. adverse-event site is, as web design, so unpleasant and unclear that incentives to report are broken, and crappy data then get twisted into whatever narrative the collecting institutions need. Weinstein picks up the rant: estimates of underreporting in VAERS put reports at about 10% of the real total. The United States had recorded more than 5,000 unexpected deaths that seemed in time to be associated with vaccination — almost certainly an undercount, and he has seen estimates of 25,000 dead in the U.S. alone. Even if someone calls that an acceptable price for driving the virus extinct, the rest of the policy does not match the story. You would not inoculate people who have already had COVID. You would not vaccinate children, who tolerate the disease well, face the longest window for long-term effects, and would be put at risk to protect the old and infirm. "That's upside down," he says. To some set of people, vaccinating looks like a good in itself — the objective of the exercise, not herd immunity. High-quality testing, which should have been obvious from the beginning, is still missing, so even the mode of transmission had to be hard-won.

The conversation then jumps decades, to a paper Weinstein published as a graduate student under the title of the reserved-capacity hypothesis. Why does a creature that can replace its cells grow feeble with age? Senescence, the paper argues, is the downside of a cancer brake: most cells may divide only a limited number of times, the Hayflick limit, because telomeres — repetitive DNA at chromosome ends — shorten with each division until the cell stops. Cancer researchers wanted to shut down telomerase, the enzyme that lengthens those ends; aging researchers wanted to lengthen telomeres for repair. Weinstein's evolutionary claim was that this is a pleiotropic bargain of the kind George Williams predicted: early-life freedom from tumors, paid for with late-life decline.

The result everyone in the field "knew" seemed to kill the idea: mice have very long telomeres and not very long lives. Many labs had replicated the telomere measurement, so it was not a one-off bad protocol. Weinstein began to wonder whether breeding protocols designed to raise reproductive output had selected against reproductive senescence and almost eliminated selection for tumor suppression. That predicted a split: laboratory mice would have ultra-long telomeres, wild-derived mice would not. Carol Greider, later a Nobel laureate for the co-discovery of telomerase with Elizabeth Blackburn, tested the hypothesis and showed that it was true. With their telomeric fail-safe effectively disabled, these animals are unreliable models of normal senescence and tumor formation. Safety tests on them, the paper warns, likely overestimate cancer risk and underestimate tissue damage. Give such a mouse a toxin that does not kill it outright and you may lengthen its life, because dividing cells are more vulnerable: the toxin acts like chemotherapy on tumors the animal cannot otherwise suppress, and a toxic drug looks beneficial.

Weinstein thinks that paradox was not understood before the paper, and that it serves companies with the hard job of bringing compounds to market. Vioxx, which turned out to do conspicuous heart damage, is the example he names. Greider, after showing the lab-versus-wild split, told him she would keep the result in house rather than publish it. A young graduate student did not yet hear what that meant: if you know how a model will misbehave and others do not, you can look like a deeper biologist — or, on the industry side, you can enjoy a cover that makes drugs look safer than they are. She later published without mentioning that the hypothesis had come from him, then broke contact. He watched her Nobel lecture present the work from his hypothesis as a lucky fact that allowed tests otherwise impossible. Thrilling and erasing, in the same hour. What troubles him more than credit is that the flaw in the mice, as far as he knows, has not been addressed. Eric called Jackson Laboratory and could not even establish that anyone thought it was an issue. As a father, Weinstein says, he cannot protect his family from a hazard he believes lurks in the medicine cabinet, because knowing the model is broken does not tell him which drug will be the next one.

He does not think Greider forgot him. They corresponded; she put a student on the work; the student emailed when results came in; Weinstein even sent tissues as a favor on another telomere project. When he sent her his manuscript for comment, she panned it and became an antagonist. He did not want to become a telomere researcher. The year was a proof of concept for upgrading evolutionary tools, and the senescence model he offered is now widely accepted. The formative experience of peer review as a debacle did not sell him on journals as the way to make a difference. In some fields, he says, PIs parasitize their underlings until the underling is senior enough to parasitize in turn. His own mentor, Dick Alexander, refused to let students work on his topics so credit would never be confused. Synthesis, Weinstein thinks, may not live best in narrow papers. Richard Dawkins has almost no journal record in the usual sense and still reshaped the field in books. Peer review, in the joke he likes, is peer preview: your biggest competitors see the work first, in private, with a power dynamic attached. He would rather be reviewed in public, with names on the disagreement.

Fridman asks a hard question about the drug of martyrdom: the popularity of pointing out that the emperor has no clothes, then walking away instead of doing the slower work. Weinstein calls it one of the best questions he has ever been asked. He has an ego and thinks well of himself; he does not think it is in the driver's seat. He loves a eureka and the phone call after. He loves a walk in a forest with a toolkit that lets him see something deep in a creature. Isolate him in a biologically interesting place with people and pets he loves, and he could be all right. He understands the martyrdom drug from watching it more than from feeling it. He does like the fight against people he does not respect, and he likes winning. He has no interest in dying for it. Life is too good: a wonderful wife, amazing children, a lovely place. Willingness to face a hazard is not a wish for the thrill. Reputation, not a bullet, is how inconvenient people get killed now. The thrill, for him, is fighting when he is in the right — a way to channel the brutality built into men toward something useful.

Fridman hears optimism in that stance: we can win this. Weinstein's version is colder. He expects us not to be okay, and he is not okay with that. As far as he knows it is not too late; humanity could still reach the fourth frontier or something like it. He has looked into the abyss and counted more ways to fail than to get out of this phase of history. The thing he does not have to worry about is that he was a coward. If we fail as he expects, it will not be because he saw what was needed and did not try. He regards that need to sleep at night as something like a personality defect, and also as a defect with positive externalities.

On his brother Eric, the gratitude is uncomplicated. Coming second into a world that already had a force of nature in it was a pain and a gift. Eric made interesting mistakes, learned, and passed the wisdom down. Weinstein's hypothesis about birth-order effects is that they are adaptive: a second-born is not born into the same niches. Eric dominated the realm of fundamentals, so Bret was driven to the other end of the continuum, toward emergence, toward biology, a fascination he dates to age three. He still hunts for fundamentals inside the emergent layer — an echo of Eric's style applied far up the stack. Fusion, he suspects, will have to power the fourth frontier; Fridman suspects the path to fusion runs through AI, which would require scholars of brains and consciousness more than of physics, and that thought makes Weinstein a little sad, because it outsources the aha. They have been butting heads for decades. Each has moved. One recurring tension: Bret talks about an admixture of collusion and emergence, an autopilot ratio; Eric hears cover for people who had a choice and made the wrong one. Brother to brother, does he love the guy? Hell yes — as a brother, and because he is awesome.

Fridman values monogamy — one girl, ride or die — and notices that the idea is waning. Weinstein does not think it is waning because a superior system arrived. Humans are somewhat sexually dimorphic, which records a lot of polygyny in the deep past, and most people today still belong to civilizations that are at least nominally monogamous. Mating patterns, like other software, can be swapped. The primary evolutionary benefit of monogamy, he argues, is that it brings all adults into child-rearing. Human babies are labor-intensive; two parents are a huge asset. For a population that can still expand, that raises the number of offspring who can be raised. Sideline a mass of males in a polygynous system and total parental effort falls. At carrying capacity, he expects polygyny to break back out. What he sees around polyamory is a false sophistication that will decay into a few males at the top with many partners and everybody else worse off.

Selection might also favor war or genocide. Neither man is signing up. Once you notice that your values can trump the evolutionary robot in an extreme case, you can notice it in milder ones, including the fantasy of being the male at the top of a polygynous stack. Lots of sex and variety are intelligible. So is a partner to whom one is bonded especially closely. If love is the thing, there is a question about its quality when divided, and about the net for a society when people are frozen out for every man who has more. Weinstein thinks there is a monogamy program in us that is not automatic, and that a lifelong bond has gotten a bad rap it does not deserve. His advice to a young person is not to date until they find "the one." Find someone worthy, someone who can handle it, someone whose underlying values are complementary enough that you fall in love — you do not have to be perfectly compatible — and then opt out together. Get out of the system that is telling you what is sophisticated to think about love. It will be complex. Nothing worth it is not.

Advice beyond love is harder to operationalize. When students asked whether to go to graduate school, he had almost nothing useful to say, because the job market and the training market looked distorted and corrupt. The meta-advice: you do not know what is coming, so invest in tools rather than knowledge, and in uncommon combinations of tools. A robot mind still helps if the robots disappear. Computer programming plus something rarely paired with it creates a rarified space. Useless courses, in the sense of not directly applicable, can be the best ones; he still remembers an immunobiology class he was not supposed to be in. Eric's undergraduate advice was to ignore prerequisites. The meaning of life, asked last, gets an answer he knows is a horror: there is nothing that escapes the heat death of the universe. After that abyss, the proximate meaning is clear enough to be a moral imperative. Being a free human on this beautiful, crippled planet is unparalleled. The obligation is to deliver that opportunity to as many people as possible, which means sustainability that is not a totalitarian horror show — the whole fourth-frontier question. He expects we will not succeed. He is not satisfied with that. There is no reason he knows of that it is already too late, and no reason, if he has done what was in his power, to put cowardice first.

Fridman closes the recording with Darwin: ignorance more frequently begets confidence than does knowledge; it is those who know little, not those who know much, who so positively assert that this or that problem will never be solved by science. In the hours just finished, the most confident assertions had come from the most official voices — on origins, on what could be discussed, on which drugs counted as knowledge. The voices that said they were not sure, and that the data should be looked at, were the ones platforms treated as a problem. Weinstein was not silenced. His channel was restricted anyway, by a machine that, after a reversal on the lab-leak question, kept taking its instructions from the offices that had just been wrong. The mirror he wanted YouTube to look into is the same one Darwin was holding up: confidence is not a synonym for knowledge, and it is often the people with the least of the second who are most sure they already know which questions science is not allowed to ask.