A billion years of evolution in a single afternoon — George Church
What this covers
George Church discusses the rapid acceleration of biotechnology capabilities and their near-term consequences. The conversation ranges across aging reversal, de-extinction, novel biomaterials, and the biosecurity risks posed by dual-use biotechnology—particularly mirror life and AGI. Church argues that exponential cost reductions in DNA sequencing and synthesis, combined with advances in gene editing and AI-guided protein design, have created a moment where directed evolution can compress what evolution accomplishes over a million years into a single afternoon. The approach exploits libraries of billions of molecular variants tested in parallel, with results fed back into machine learning for iterative improvement. This "natural computing" bypasses the need to simulate molecular behavior from first principles.
The conversation covers several distinct claims. Church argues that master switches—single genes controlling complex traits—make even highly multigenic conditions amenable to targeted intervention, citing growth hormone's control over height. He contends that offense has an advantage in synthetic biology because the design space is too large to defend against—attackers can synthesize around countermeasures while defenders must anticipate them. On longer timescales, he suggests that therapies addressing aging mechanisms (not just isolated damage) could achieve longevity escape velocity around 2050. Yet the conversation turns sharp on governance. Church emphasizes that voluntary moratoria have failed, argues against rushing AGI development absent genuine crisis, and champions underutilized interventions like genetic counseling. He frames de-extinction not as recreating exact species but as identifying the minimum genetic changes needed for functional categories, and positions chirality differences in mirror organisms as a catastrophic biosecurity risk that a single person could weaponize.
Church argues that exponential improvements in biotechnology (cheap sequencing/synthesis, gene editing, AI protein design, massively parallel libraries) put us on the verge of major payoffs—aging reversal, de-extinction, novel biomaterials—but that the same dual-use power demands caution, especially around AGI and mirror life, and that simple interventions like genetic counseling are underexploited.
- Biology's exponential cost/capability curve now rivals or exceeds Moore's law, enabling library-based 'natural computing' that tests billions of variants in an afternoon
- Offense currently has the advantage in synthetic biology, so surveillance and reducing motivation matter more than moratoria alone
- Reductionist 'master switch' findings (e.g., growth hormone for height) let small numbers of edits control complex multigenic traits
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Evolution didn't develop capabilities like new base pairs because it generally goes with what works and, unlike technology, cannot jump to outcomes whose intermediates aren't incrementally useful—every change must be justified at each step—whereas human technology can build toward endpoints through non-useful intermediates.
“In technology you can jump to things where all the intermediates aren't incrementally useful. But evolution, as far as we know, is generally limited to… You have to justify every change”
By identifying which transcription factors a target cell expresses and applying just those to a stem cell, one can derive recipes—often 1 to 7 transcription factors—to convert pluripotent stem cells into almost any existing cell type in the body, an approach that underlies GC Therapeutics and reduces the number of genes that must be manipulated to reach a goal.
“you can see what transcription factors does it express at the time that it's the target? Then you say, "Let's just try those on the stem cell and see if they work." That recipe has worked quite well. It's the basis of GC Therapeutics”
Even highly multigenic traits can often be controlled by very few changes: height involves on the order of 10,000 genes each with tiny effect, yet a single gene—growth hormone (somatotropin)—produces extreme low or high stature on its own and is used clinically, illustrating that reductionism can isolate a 'master switch' for complex phenotypes.
“they tracked it down to something on the order of 10,000 genes... But if you take growth hormone, somatotropin, you have extreme examples where you'll get extremely low small stature and extremely high stature due to that one alone.”
Adding more intelligence or thinkers doesn't necessarily accelerate biology, because some problems can't be parallelized—just as nine women cannot produce a baby in one month—and we lack experience with thousands of Einstein-level intelligences operating simultaneously to know which gains are achievable.
“it's like the question of, "If you have nine women, can you do pregnancy in one month?" No, not at present.”
Mirror life is dangerous because, if developed past a certain point, it would be easy to weaponize, and a properly weaponized mirror organism could wipe out all competing (normal-chirality) life; the danger is amplified because biotechnology lets a single disgruntled person wield catastrophic power that historically required many people.
“this seems like the sort of thing that could wipe out all competing life if were properly weaponized.”
Unlike COVID-19, where delay cost millions of lives, the rush toward AGI is a completely artificial emergency—any crisis it causes is self-created—so there is no rush; we should go very slowly on AGI/ASI and instead double down on slightly narrower scientific AI, because we don't understand our own ethics well enough to educate a foreign intelligence and competition tends to erode safety rules.
“This is a completely artificial emergency. This is not like COVID-19, where millions of people were dying if we delayed the science. This is something where, if there ever is a crisis, it's because we created it”
Genetic counseling is an underhyped intervention that is competitive with gene therapy for preventing rare recessive diseases in future children: at about a hundred dollars per genome it offers at least a tenfold return on investment by avoiding millions in care costs and lost productivity, has been practiced successfully since 1985 by Dor Yeshorim, and is not eugenics because it enables choice rather than government coercion.
“What I would say is genetic counseling is underhyped. It is clearly competitive with gene therapy in a certain sense... This has been in practice since 1985 in Dor Yeshorim”
AlphaFold predicts protein structure to fractions of an Angstrom but does not predict function: substituting an alanine for a serine in a serine protease yields exactly the right fold yet the enzyme simply won't function, so structural prediction must be combined with evolutionary or experimental knowledge to determine activity.
“if you substitute an alanine for a serine in a serine protease, it will have exactly the right fold... But it won't function. It just won't function.”
Rather than simulating molecular function through assumption-laden chains from quantum electrodynamics to molecular mechanics, you can make huge libraries (up to ~10^17 variants in vitro), barcode them, and physically play out which work—a form of 'natural computing' that is 100% precise because you do the real thing, then feed the data back into conventional AI for the next round, compressing what evolution does over a million years into an afternoon.
“it's so exact. It's a hundred percent precision, because you're not simulating. You're not making assumptions... You're really doing the real thing. So you're doing a kind of natural computing.”
Estimating the prevalence of extraterrestrial life is better approached through exploration than simulation, and a key problem is that almost none of our space missions actually looked for life—components capable of detecting it were often absent, underused, or their positive results dismissed (as with Pioneer)—so priorities should be the water geysers of Jupiter's and Saturn's moons and water on Mars, given the solar system holds about 50 times more liquid water than Earth.
“The sad truth is that almost none of the missions that we've sent outside of Earth have actually looked for life.”
De-extinction should be understood not as recreating an exact ancestral species but as synthetic biology with societal goals: the dire wolf project did not make an exact copy but illustrated which few genes distinguish a dire wolf from a gray wolf, and the discipline's deeper interest is the 'minimum' question—the smallest set of changes needed to achieve a functional category—rather than perfect copies.
“With the dire wolf, we clearly didn't make an exact copy of a dire wolf. But it helped illustrate and educate people around the world on, "What is the difference between a gray wolf and a dire wolf?"”
In synthetic biology offense has the advantage because biotechnology enables smaller, harder-to-detect efforts, and defensive schemes like codon remapping work against naturally evolved viruses but are much harder to deploy against synthetically manufactured viruses since attackers can design among up to ~10^80 possible genetic codes.
“It's much harder. Again, the offense has the advantage. We can make a lot of different codes.”
Replicating a specific brain probably requires far more information than synthesizing a brain from scratch, because 10^14 synapses take vastly more bytes than the genome's billions of base pairs—analogous to how photographing every page of a book to copy it captures more raw data than re-deriving the book, though replication may sometimes be desirable to preserve a particular configuration.
“replicating a brain probably involves a lot more information than synthesizing it... 10^14 synapses is going to take a lot more bytes than the genome, which is billions rather than 10^14.”
When Church's lab introduced chip-based gene synthesis in a 2004 Nature paper—a thousandfold cost reduction—it was dismissed for about a decade and not even plotted on the Moore's-law curve for DNA synthesis, used only by collaborators and alumni, illustrating how transformative tools can be 'hidden in plain sight.'
“When we came out with the first chip-based genes in a 2004 Nature paper, basically people dismissed it for about a decade... It wasn't even listed on the Moore's law curve for DNA synthesis, even though it was thousand times cheaper.”
A revolution in generating nonstandard amino acids—incorporating, covalently or as ligands, all the stable elements of the periodic table, with potentially 34 new amino acids usable alongside the standard 20 in E. coli—will, once AI models are trained on them, rapidly enable a whole series of new biological materials including potentially room-temperature superconductors.
“there's a revolution going on in generating nonstandard amino acids, where the amino acids can either have as part covalent part of them, or as easily liganded, all the stable elements in the entire periodic table.”
People avoid genetic counseling partly because of a trolley-problem intuition that if you don't influence an outcome it's not your fault, but in fact not doing something is itself a decision—everything is your fault—so declining to screen and having an affected child is a chosen outcome, not a blameless default.
“it has to do with the trolley problem. If you don't influence it, it's not your fault. But actually everything is your fault. Not doing something is a decision.”
Over human history the destructive capability accessible to a single person has grown dramatically—from bare hands to one person with the right technology or connections being able to blow up a city—and biotechnology accelerates this trend by enabling smaller, harder-to-detect efforts, so society should aim to dial this individual capability back down.
“Today, one person with the right connections or right access to technology could blow up a city. That's a huge increase in capability. I think we want to start dialing that back a little bit somehow.”
Biology already operates at roughly 0.4 nanometer resolution in three dimensions, whereas the 'one nanometer' chip process due in 2027 is really about 40 nanometers center-to-center in mostly two dimensions—meaning biology is about a thousand times denser linearly and potentially up to a billion times denser by volume, and only needs more practice using the full periodic table.
“Biology is already at 0.4 nanometer resolution and it is in three dimensions... it could be a billion times higher density that biology is already at.”
Based on the exponentials in biotechnology and breakthroughs in treating general aging (not just isolated damage) with therapies now entering clinical trials, longevity escape velocity—where lifespan increases by more than a year per year—could be reached by around 2050, and most people alive today have a good chance of making it that far.
“I wouldn't be surprised if 2050 would be a point. If we can make it to that point, 25 years… Most people listening to this have a good chance of making it 25 years.”
Voluntary moratoria and good-citizen signups are insufficient for biosecurity; we need active surveillance, well-known consequences, and easy mechanisms for whistleblowers, as demonstrated by the germline-editing case where moratoria and disapproval existed yet someone did it anyway and many people knew—a clear failure of the voluntary system.
“we stop deluding ourselves into thinking that moratorium and voluntary signups to be good citizens is going to be sufficient. We need to also have surveillance and consequences, and mechanisms for whistleblowers”
Church's company Dyno Therapeutics demonstrated a hundredfold improvement in targeting neurons in the brain in a single campaign by using AI and testing millions of different viral capsids, illustrating rapid progress in tissue-targeted gene delivery.
“showed they could get a hundredfold improvement in targeting neurons in the brain, which is a big deal. That was just one little campaign that they did. That one experiment involved a lot of AI and a lot of testing of millions of different capsids.”
For many therapies you don't need to reach 100% of cells—sometimes 1% is enough because that fraction can produce a missing enzyme, and the enzyme need not be made in its normal location: you can repurpose muscle as a temporary immune component for a vaccine, or make a brain enzyme in the liver if the goal is just to get it into the blood.
“for some therapies you just need to get 1% because that 1% can produce some missing enzyme. And that 1% doesn't have to necessarily be in its normal place.”
Because aging is largely a cellular phenomenon driven by proteins and signaling factors in the blood, if you could replace every nucleus in the body the body would become young again without going all the way back to the embryo, and replacement cells could displace old ones by taking over their niches—with the brain being the hardest case but potentially addressable via artificially introduced stem cells that learn and join circuits.
“you could imagine that if you replaced every nucleus in the body, it would suddenly be young again without going all the way back to the embryo and forward again.”
Church selects lab members primarily for being nice—which he finds highly predictive of success in and after the lab—rather than for genius, and for multidisciplinarity, since people who have already learned two skills or languages demonstrate they can acquire a new skill and bridge disciplines, making multidisciplinary teams buildable where teams of pure disciplinarians are not.
“I typically tell them that I'm looking for people that are nice. I'm not necessarily looking for geniuses... nice, I think, is highly predictive of how well you will do in the lab and afterwards.”
Biotechnology has been improving at a rate about the same as, or slightly faster than, Moore's law, with the field able to stand on the shoulders of the electronics industry to accelerate, suggesting we are at or near the beginning of the big payoff rather than experiencing stagnation.
“We have something that's about the same speed, a little bit faster than Moore's Law in biology. It's more recent... We could stand on the shoulders of the electronics giants to go a little bit faster to catch up.”
One lab can spawn ~100 biotech companies because of a self-reinforcing positive feedback loop enabled by Boston's dense, walkable cluster of Harvard, MIT, BU, pharma, and biotech—which attracts top talent, lets spouses find jobs, and compounds early literature wins—combined with a lab culture that insists on connecting basic science to societal needs from the start and an exponential technology curve that asymmetrically rewards whoever already has the academic-to-industry pipeline running.
“Boston is a unique culture... Having a concentration of biotech and pharma and MIT and Harvard and BU and so forth, all in one pretty walkable distance... is one thing. That's the starting point.”
Biology is both far more complicated than anything humans design from scratch yet much more forgiving, operating at a very high programming level—for example, a small deviation in fetal development can produce a functioning two-headed animal with two personalities despite there being no specific selection for such a form.
“it's both much more complicated than almost anything we've designed from scratch, but it also is a lot more forgiving in a certain sense... a little deviation from the normal developmental pattern during fetal development and they both function fine.”
It is quite possible that mirror life already exists in our solar system or even on our planet, simply without having been weaponized.
“It's quite possible it's already here. In other words, we already have mirror life in our solar system or maybe even on our planet. It just hasn't been weaponized.”
Fast-replicating biological systems could in principle access human-engineering capabilities like radio communication or even nuclear reactors by treating manufactured artifacts as part of their replication cycle (like a bird building a nest), provided their range of usable materials is expanded—though extreme conditions like a fission reactor's temperature pose obvious incompatibilities.
“A bird can make a nest. You consider the whole nest as part of the replication cycle of the bird. So you can say the biological thing that replicates at a 30 minute doubling time could make a nuclear reactor. That would be its nest”
A safe superintelligence would likely slow or eliminate biology progress rather than accelerate it, because the first thing it would conclude is that biology is not relevant to it since it is not made out of biology.
“I think it would slow it down. I think it would eliminate it, because the first thing it would conclude is biology is not relevant to me because I'm not made out of biology.”
If AI safety issues are made the top priority and solved, we are likely to achieve almost perfect health, accelerated by a positive feedback loop in which more people gaining good healthcare contribute to prompting AI and building harmonious human-machine hybrid systems.
“If we handle the safety issues, and that has to be a top priority, then we're probably going to have almost perfect health. Why wouldn't we? It's going to go so fast.”
A possible (though not advocated) upside of cutting NIH and NSF budgets is that it could force more serious reliance on philanthropy and industry-sponsored research and make scientists listen more carefully to societal needs, but it could also shift scientific dominance to another nation-state such as China and risk producing a 'hypercapitalism' with its own pathologies.
“You could say that it forces us to think more seriously about philanthropy and industrial sponsored research. That could be a positive thing... It could be that it creates another nation-state that now is the dominant force. Like China could now become the next empire after the US.”
Future information-storage molecules might keep the ACGT bases but rebuild DNA's backbone out of peptides—smaller and more compatible with an expanded amino-acid collection—though binary digital storage works well for ~99% of electronic needs and it is unclear whether more than four nucleic-acid components are needed.
“Another possibility is if we changed the backbone of DNA. Maybe we keep the ACGT, but make it out of peptides now, a little bit smaller, a little bit more compatible.”
Rather than pushing toward a new level of intelligence, which is challenging and perhaps not urgent, simply 'actualizing' the people we currently have—getting all 8 billion to be super healthy and educated to the best demonstrated levels—would produce a completely different world.
“actualizing the people that we currently have would be quite impactful... What if we had 8 billion, super healthy... Einstein-level intelligence, education level the best we can come up with… That would be a completely different world.”
Biology frequently offers low-hanging fruit—levers it simply hands us—that people wrongly dismiss as improbable, with vaccines being an example of an 'amazing gift that didn't have to exist but does.'
“Very often there's low-hanging fruit that people just think is improbable. But it's there because biology has all these gifts where it just hands over to us levers that we can flip. Like vaccines are this amazing gift”
We made it through the Cold War without accidentally or intentionally detonating hydrogen bombs on enemies, though two atomic bombs were used.
“We made it through the Cold War without blowing up any hydrogen bombs, as far as I know, accidentally or intentionally on enemies. We did do two atomic bombs.”
Approaching longevity escape velocity will not feel like a sudden hit-or-miss cliff; rather, people are more likely to be healthier 25 years from now than they currently expect, as improvements accumulate gradually.
“it's not going to be some sudden point where you're going to be so sick 25 years from now that it's like hit or miss. It's more likely that you're going to be healthier 25 years from now than you thought you were going to be.”