Introduction

I think reprogenetics (human germline genomic engineering) can be done in a widely acceptable and beneficial way, and should be pursued aggressively. In particular, as a strong background motivation of mine, I think accelerating strong reprogenetics is probably the best way to enable strong human intelligence amplification; and I think strong HIA is among the best ways to decrease existential risk from AGI.

A very common objection to caring much about reprogenetics is that AGI seems very likely to come soon—say, within a decade or two. (Here I mean "actual" AGI—the kind that probably doesn't already exist—the kind that has fluid intelligence and AI advantages for recursive self-improvement, which together make it likely to take over the world shortly after being created.) The objection is fairly straightforward:

AGI will probably come within a decade or two. If that's going to happen, then even if a new cohort of brilliant humans were born today, they would still be children, or would at best have barely begun contributing ideas for how to avoid extinction. Any supposed benefit, denominated in percentage points of AGI existential risk averted, is small. Therefore, reprogenetics is too slow; and if you're going to work on human intelligence amplification, you should work on adult HIA.

Now, of course this is true to an extent. If AGI comes within 15 years, reprogenetics is almost totally useless. To the extent you believe that will happen, you believe reprogenetics is quantitatively less useful in expectation. Also, having a faster HIA method would of course be great.

However, I believe that this line of reasoning has led to a very mistaken underallocation of funding, talent, and other resources towards human intelligence amplification in general, and reprogenetics in particular. So, I would like to push back in a few ways:

• On general strategic grounds, there's altruistic alpha in working on HIA.
  • (See the section "HIA, part of your nutritionally complete portfolio".)
• You shouldn't be that confident that AGI is coming within a decade or two.
  • (See the section "Against confident short timelines".)
• Putting HIA on a legibly fast and good trajectory might indirectly decrease the motivation to create AGI.
  • (See the section "HIA may indirectly slow down AGI capabilities".)
• Even given a high probability of AGI coming soon, HIA still has a large positive impact.
  • (See the section "HIA has substantial impact even with short timelines".)
• Other methods for strong HIA are not at all obviously faster than reprogenetics.
  • (See the section "Adult HIA methods aren't fast either, absent big investment".)

(Note that this isn't a comprehensive fair-and-balanced report, but rather a collection of arguments in one direction. For example, I won't here discuss reasons that HIA could increase existential risk from AGI. Also, some but not all of these arguments rely on the assumption that alignment is very difficult.)

HIA, part of your nutritionally complete portfolio

• HIA is super neglected. Would you rather be the 3000th person working on AI safety, the 300th person working on AI regulation, or the 3rd person working on accelerating HIA?

• There's broad philanthropic alpha in working on HIA. It's tractable (to accelerate), neglected, and important. In particular, because it's somewhat taboo (though probably less than you think), there's relatively low-hanging fieldbuilding fruit to pick.

• We must attend to all plausible points of intervention.

  • The world cannot be allowed to be destroyed by AGI. We must consider intervening on this course of events at every possible point.
  • HIA is widely agreed (sometimes publicly, more often in private) to be one of the few real hopes for humanity's survival. This is a place where we must not drop the ball. Currently the ball is being dropped.

• Uncoordinated resource allocation leads to unbalanced portfolios.

  • A correct belief that Sector A is more important than Sector B would lead someone to focus on Sector A exclusively when deciding where to allocate resources. This kind of focus is reasonable, given that one can't spend forever evaluating and comparing all the different possible sectors; you have to focus on one or a few.
  • People allocate talent, funding, and other resources without checking the preexisting portfolio of allocations. (This is understandable, since that portfolio isn't written down in any 1 place or in any 10 places.) Thus, they don't automatically track that a sector is underresourced.
  • It's widely accepted that AGI is coming soon, and therefore HIA is not such a useful investment. Even if this belief is correct, this belief leads to HIA being very underresourced within the global joint philanthropic portfolio.
  • This is a collective mistake. This is as true now as it was when I first wrote it 3 years ago. The candle is still burning down.

Against confident short timelines

I don't in fact think that confident short timelines (say, ">80% within 15 years" or similar) make much sense. As yet, I have not heard a clear and convincing case that AI research has already uncovered the engines that would produce a smarter-than-human general intelligence. The main arguments for AGI coming soon boil down, I think, to the rapid rise in capabilities and to the use of AI in AI research.

Regarding the rapid rise:

We have large gaps in performance between AIs and humans (sample complexity for learning, ability to generate novel good concepts).

Also, we have an apparent explanation for the rapid rise in capability: hoovering up big piles of internet data for quasi-imitation. Current AIs are skillful / knowledgeable approximately when there's a bunch of human training data for that thing. Of course, you need more elements if you want to explain all of the rapid rise, such as unhobbling via harnesses and RL, as well as computer advantages (scaled RLVR, speed, parallelism, copying, inexpensiveness).

But the core explanation of "the bulk of these capabilities come from human demonstration" holds water, as far as I know. In particular, this explanation says that current AIs got their capabilities via some route other than accumulation of crystallized intelligence that was produced by their own fluid intelligence. Namely, they got their capabilities via "copying" (broadly construed) crystallized human intelligence originally generated by human fluid intelligence.

This seems to largely explain away the rapid rise in capabilities, without invoking current AIs having much fluid intelligence. Combined with the apparent gap, with humans still well ahead on general fluid intelligence, I don't see how anyone gets to being very confident that we already have most or all of the ideas that would be needed for AGI.

Regarding the use of AI in AI research:

I expect the most important kinds of research to not be accelerated much, because they are not bottlenecked by coding in the first place. Unless you already think we're close to making AGI for other reasons, so that what's left is mainly the kind of research that is greatly accelerated by current or near-future AI, this partial acceleration shouldn't change your timelines much.

For previous discussion, see "Do confident short timelines make sense?" and "Views on when AGI comes and on strategy to reduce existential risk". I'm open to debating this with anyone who'd like to make a serious, public case for confident short timelines.

HIA may indirectly slow down AGI capabilities

The main stated justification for pursuing AGI capabilities is that AI / AGI would bring abundance for humanity. It stands to reason that if there were a credible, workable plan to get the (supposed) benefits of AGI without the huge existential risk (and other disempowerment of humans), then there would be less motive to develop AGI and it would be harder to justify pursuing AI capabilities. (See "5.1. Abundance makes less motive to make AGI".)

I don't know how large this effect would be. Presumably not very large. Presumably many people trying to increase AI capabilities just want money, power, status, or other selfish things, and would find some other justification. I'm not sure though; it's hard to put upper or lower bounds on the importance of underlying spiritual currents in society of hope, motivation, the longer-term future, and so on.

HIA has substantial impact even with short timelines

Approximate summary of this section:

We should play to our outs; outs are scarce; HIA is an out.

• In dire situations, if you're trying to be strategic about winning, it's advisable to play to your outs. That is, avoid the temptation to focus on incremental tractable gains that don't actually increase the chances of overall success. Instead, aim at paths to overall success, including by setting yourself up to take advantage of such opportunities, even if any specific path is unlikely.

• Working towards a pause / stop / ban on AGI progress is a top priority. But what is a pause for? We pause, and then what? How do we more robustly prevent existential risk? Hopefully, we can ban Red AGI research, but probably not Blue AGI research progress. (Though social / political pressure might possibly be able to significantly slow down even Blue research.) In the longer run, how do we avoid making unaligned AGI? HIA is a way to improve our long-run chances by giving humanity more brainpower to find good answers to that question.

• Suppose you do get a pause on AGI progress, but you don't have a plan for how to more robustly stop existential risk. More specifically, suppose that progress is not being made towards the outs. Then the pause is being wasted: for the moment you're bailing water out of the boat as it leaks in, but at some point more and more leaks will be sprung. You have to also be patching the hull. Ten years from now, when you have hopefully updated toward slightly longer timelines, you'll regret not getting started on the longer-term solutions back in 2026. I know I regret not working on this a decade ago, rather than bashing my head against the AGI alignment problem. HIA will make some progress on its own by default, but this argument goes through quantitatively—you'll regret not having quantitatively accelerated HIA if you could have.

  • Another lens on this: If you have a single long-term out, then the value of pulling the long-term out forward in time by one year is equal to the value of pushing the pause to last for one year longer.

• Very short timelines are pretty intractable. If AGI is actually coming in 5 years by default, unless you think alignment is easy, there just aren't many outs. Of course we still want to try, but intractability does weigh in the prioritization calculation.

• Pulling world-saving conditions forward in time is very valuable unless you're very confident extinction will have already happened.

  • If you make an intervention such that the world would get saved in 40 years rather than 50 years, you've decreased existential risk by an amount equal to the probability of AGI coming between 40 and 50 years from now. Unless you have very confident short timelines to AGI, this still decreases total existential risk by a few percentage points.
  • Decreasing total existential risk by a percentage point is very valuable!
  • As an illustration, you could have a hazard rate of 5%. In other words, at the beginning of each year (thus, conditioning on the fact that you've survived so far), you think there's a 5% chance that AGI comes within the next year. This distribution has mean <20 years and median <14 years. In this case, the world being saved at year 40 vs. at year 50 is worth a 5% decrease in existential risk.
  • As another illustration, you could have a hazard rate of 10%. This is an extremely aggressive (confident) timeline. This distribution has median <7 years. In this case, the world being saved at year 40 vs. at year 50 is still worth a ~1% decrease in existential risk! Plug into your calculator: ${0.9}^{40}-{0.9}^{50}$, aka ((0.9)^40-(0.9)^50)
  • For a more detailed model with numbers, see "The benefit of intervening sooner".

• AGI alignment is extremely hard. Therefore solving AGI alignment is not that helpful of an out.

  • This shouldn't be surprising. Minds, especially superhumanly intelligent minds, are the most complex entities that exist. It's not surprising if getting them to do some really specific thing (make the universe end up great for humans) is extremely difficult, unlike any challenge humans have faced before. As an analogy, how hard would it be for the Ancient Greeks to get to the moon? What if they have only 15 years? Do you think that task is easier or harder than solving the mysteries of values, intelligence, creativity, self-modification, and so on, well enough to put a mind in an unnatural but stable state, within 15 or even 50 years?
  • The problem of AGI alignment has major structural and technical obstacles, crucial and insoluble deep mysteries, and appears to be cognitively unnatural to work on.
  • Therefore, "just solve alignment now, fast" is not a feasible out.

• HIA is an out we can play to.

  • Alignment is very difficult because it's a very difficult intellectual (conceptual, philosophical, technical) problem; it's bottlenecked on really insightful creative careful thinking, not on money or other resources. So, strong HIA might lead to humanity being able to solve alignment.
  • If alignment is still too hard, HIA might lead to humanity being able to handle AGI existential risk in some other way. The first-rate intellects that benefit from strong HIA would probably have a better shot than us at thinking of ways to avoid existential risk.
  • For example, they might:
    • develop next-generation HIA
    • figure out how to keep in force a stronger ban on AGI
    • make progress on material problems, more immanently removing motivation to take risks with AGI
    • improve society's decision-making in general, leading to a strong social consensus against risking AGI
  • Indeed, the whole point of making an aligned AGI in the first place is that you need intelligence, but you can't amplify your own intelligence enough. Therefore, an alternative to making aligned AGI is to figure out how to amplify human intelligence enough.

Adult HIA methods aren't fast either, absent big investment

A variant of the argument against reprogenetics goes like this:

AGI is coming very soon. Reprogenetics has a built-in delay of at least 15 or 20 years while the kids grow up, before it actually helps. Instead, we should work on adult HIA, which will be able to help much sooner.

Of course, overall, this logic is valid and compelling. If I believed we could do (strong) HIA sooner than a couple decades, I would work on that. If there were actually a Manhattan-Project-scale project for adult human intelligence amplification, I would probably drop what I was working on in reprogenetics and join that effort. I would have some hope of success—assuming that we would have available the scientists, equipment, experimental volunteers, and money that would be needed to run several experimental investigations in parallel.

However, short of a Manhattan Project, I'm somewhat skeptical of adult HIA being a faster bet than reprogenetics. There are several reasons, which I'll list here. But I want to summarize the main reasons more briefly:

It's pretty expensive to intervene on an adult brain. Further, it's not actually clear how to intervene on adult brains to make them smarter, so you'd be doing exploratory research. It's probably hard to make adult brains much smarter, so you'd probably have to do a lot of exploratory research. Each experiment probably takes at least months (the intervention wouldn't work instantly), and more likely years (to develop new tech). This means that developing the method involves lots of costly, lengthy, serial experiments. So actually adult HIA takes a long time to develop—like, think decades.

The reasons in more detail:

• Reprogenetics has good-enough data on intelligence; adult HIA does not.

  • To know what genes increase intelligence, we can look at the billions of humans already living. This doesn't come remotely close to telling us everything there is to know about genes and intelligence, but it most likely has already told us enough to make a future child be likely to possess a world-class intellect. (What's left to do is stem cell biotech.)
  • There's no comparable large dataset for adult HIA interventions. If you tried an intervention, you would wait a year, collect some data—and then you would have one (1) datapoint. With a single experimental volunteer, you'd even run into challenges trying to collect longitudinal data. You'd need some way to measure cognitive capability that is at least robust to practice, let alone to adversarial hacking (e.g. if you want to do distributed science / self-experimentation).
  • It may be that any large increase in intelligence must take a long time (years, say), even if it can be done in adults. It might, for example, take a person many months or years to develop the new ways of thinking that are unlocked by an adult intervention. Without that time to develop new ways of thinking, there may not be many clear demonstrations that you've really been put on that trajectory. This further increases the time needed for experiments (increasing the serial time needed, and the fungible cost per datapoint).
  • Each new protocol may take months or years to develop. Thus, even though there's a faster feedback loop of end-to-end testing for adult interventions compared to reprogenetics, the loop itself is slow.
  • Reprogenetics, in contrast, probably doesn't need end-to-end testing, because of the abundant natural experiment data. Reprogenetics does require innovation in things like stem cell biology. But that innovation does have fast feedback loops—significantly faster than end-to-end adult HIA, in most cases. Further, these stem cell biotech problems can mostly be tested in animals, whereas end-to-end adult intelligence amplification interventions can't really be tested in animals (except for safety), because there's no great (human-generalizable) measurement of animal intelligence.

• Reprogenetics would probably work for strong HIA; adult interventions may or may not work well for strong HIA.

  • There is a clear path for strong reprogenetics to work: in vitro gametogenesis (or another epigenomic correctness method), plus strong genomic vectoring (iterated meiotic selection, chromosome selection, and/or iterated CRISPR editing).
  • There is not a clear path for adult HIA. I do not believe anyone has a near-term-feasible way that we have strong reason to think would work. (I'm somewhat excluding whole brain emulation. It would work, though I think it is actually extremely difficult and would take longer than reprogenetics. But more importantly I think it's very dangerous.) See "Overview of strong human intelligence amplification methods".
  • Without a clear path to success, adult HIA is a less compelling bet. Making big bets like this is hard enough when there's a clear path. Without a clear path, the probability of overall success is multiplicatively diminished (compared to just having the uncertainty around execution). And, it would be hard to coordinate talent and funding.

• There are general technical obstacles to adult HIA.

  • Adult brains have lost various kinds of flexibility that child brains possess. Many neurites (long-range axons, dendrites, synapses) have been pruned; genesis of neural structures has greatly decreased; structures have solidified (e.g. via perineuronal nets).
  • The childhood developmental program is evolved for canalization—making important structures in the body form properly. This development program is no longer operating in adulthood. Therefore, tweaks to brain function in adulthood won't be met by compensatory mechanisms that keep everything functioning well, compared to how much a genetic tweak will be met with compensatory mechanisms during development.
  • The genome of a single cell isn't easy to access and manipulate, but it's fairly easy compared to accessing an adult brain at scale. The blood-brain barrier, the skull, the meninges, and immune reactions make it very inconvenient to deliver drugs, transplant cells, or implant electrodes into the brain at scale.

• Reprogenetics has strong momentum, in terms of scientific foundations, user motivation, and user on-ramp.

  • Reprogenetics is constituted by statistical genetics, reproductive science, and stem cell biology. Each of these fields makes progress on its own. E.g. there's lots of work on building biobanks, constructing polygenic scores, sussing out genetic causality, characterizing germline cells, characterizing and supporting healthy gametogenesis and embryogenesis, reprogramming cells, producing gametes in vitro, measuring genes, editing genes, and so on. For this reason, reprogenetics is already fairly close to being technically feasible.
  • The technology stack of reprogenetics for decreasing disease risks is almost identical to the stack for increasing intelligence (except for the specific genes). While many people would be interested in giving their children genomic foundations for higher expected intelligence, even more people would like to give them foundations for long, healthy, mentally healthy lives. That means there's a large personal, financial, and social incentive for everyone involved (scientists, funders, parents, regulators) to make this technology work well and at scale.
  • Most of the different hypothetical future reprogenetics technologies share most of their biotech stack and social stack (motivations, politics, regulation). On the other hand, adult HIA is somewhat more diverse (brain drugs vs. brain editing vs. stem cell transplants vs. BCIs). So investments in science and fieldbuilding projects for reprogenetics may be somewhat more synergistic with each other.
  • I don't see such a large-scale motivator for adult HIA technologies; BCIs, brain drugs, brain editing, and stem cell transplants all seem fairly niche, maybe possibly useful at some point for treating some specific illnesses.
  • For these reasons, my expectation is that legal obstacles will be tractable, as the component technologies of reprogenetics are proven safe and effective.
  • Millions of babies are already born through IVF. Reprogenetics would be a small change to the procedure from the parents's perspective (perhaps even less burdensome, via the advanced reprotech needed for reprogenetics). Thus there's already a large user base who could easily adopt reprogenetics. Most adult HIA methods (anything involving brain surgery or gene editing) would probably be somewhat hard to scale up a lot; some may be quite easy to scale up (e.g. non-editing brain drugs). A comparison point might be other non-deep brain surgeries, or organ transplantation.
  • Scale is important for HIA because, however hard it is to amplify intelligence, it's probably even harder to amplify other cognitive capacities such as wisdom, philosophical insight, sanity, creativity, persistence, curiosity, etc. If some such qualities are needed to achieve the progress that humanity most needs, and they can't easily be intervened on, then they would have to arise by chance. The plan here is to make strong HIA feasible at scale; hopefully enough people will want to avail themselves of that opportunity; and hopefully some of the resulting people will be both first-rate intellects and also in possession of those other needed qualities.

• Therefore, adult HIA is fairly likely to take a rather long time.

  • Due to the technical difficulties, iterations may be months or years. E.g. each scaling up of BCI electrode count by an order of magnitude may be a multi-year research project; several such scale-ups may be needed for true strong HIA via BCIs.
  • Due to the lack of a clear method, good abundant data, and fast training-robust intelligence tests, many iterations may be required to work out a method.
  • Due to the lack of motivation and a market, there may not be so many resources available. Therefore, exploratory experiments probably can't be massively parallelized.
  • Therefore, adult HIA would likely require many serial experiments, or in other words, a long campaign of research.

Now, all of this having been said, I could easily be wrong. I've focused largely on reprogenetics, because that's what I think will work. I'm enthusiastic about roadmapping and research on adult HIA in general, and on specific methods. Hopefully, those interested in adult HIA would take these as non-dealbreaker constraints, to chew through or work around or creatively break.

Takeaways

• HIA should be among the top priorities in a promising portfolio for avoiding existential risk from AGI.
• You probably shouldn't have very confident short AGI timelines.
• Even if you do have confident short timelines, you should still consider HIA a top priority.
• HIA is extremely neglected—it has much less talent, funding, and other resources compared to what it ought to have.
• There's lots to work on, to accelerate HIA via reprogenetics. See e.g. "Berkeley Genomics Project seeking hires (and collabs)" and "Projects that might help accelerate strong reprogenetics".