I’d like to see more intellectual scenes that seriously think about AGI and its implications. There are surely holes in our existing frameworks, and it can be hard for people operating within them to spot. Creating new spaces with different sets of shared assumptions seems like it could help.
Absolutely not, no, we need much better discovery mechanisms for niche ideas that only isolated people talk about, so that the correct ideas can be formed.
Thank you for writing this!
Hm. I super like the notion and would like to see it implemented well. The very first example was bad enough to make me lose interest: https://russellconjugations.com/conj/1eaace137d74861f123219595a275f82 (Text from https://www.thenewatlantis.com/publications/the-anti-theology-of-the-body)
So I tried the same thing but with more surrounding text… and it was much better!… though not actually for the subset I’d already tried above. https://russellconjugations.com/conj/3a749159e066ebc4119a3871721f24fc
A longer sentence is produced by, and is asking the reader to be, putting more things together in the same [momentary working memory context]. Has advantages and disadvantages, but is not the same.
Yes, and this also applies to your version! For difficult or subtle thoughts, short sentences have to come strictly after the long sentences. If you’re having enough such thoughts, it doesn’t make sense to restrict long sentences out of communication channels; how else are you supposed to have the thoughts?
On second/third thought, I think you’re making a good point, though also I think you’re missing a different important point. And I’m not sure what the right answers are. Thanks for your engagement… If you’d be interested in thinking through this stuff in a more exploratory way on a recorded call to be maybe published, hopefully I’ll be set up for that in a week or two, LMK.
On the “self-governing” model, it might be that the blind community would want to disallow propagating blindness, while the deaf community would not disallow it:
https://pmc.ncbi.nlm.nih.gov/articles/PMC4059844/
Judy: And how’s… I mean, I know we’re talking about the blind community now, but in a DEAF (person’s own emphasis) community, some deaf couples are actually disappointed when they have an able bodied… child.
William: I believe that’s right.
Paul: I think the majority are.
Judy: Yes. Because then …
Margaret: Do they?
Judy: Oh, yes! It’s well known down at the deaf centre. So some of them would choose to have a deaf baby! (with an incredulous voice)
Moderator: Actually, a few years ago a couple chose to have a deaf baby.
Margaret: Can’t understand that!
Judy: I’ve never heard of anybody in our blind community talk like that.
Paul: To perpetuate blindness! I don’t know anybody in the blind community who’d want to do that.
I am trying to point out to you that almost everyone in our society has the view that blinding children is evil.
Our society also has the view that people should be allowed to reproduce freely, even if they’ll pass some condition on to their child.
you are influencing them at the stage of being an embryo
I’m mainly talking about engineering that happens before the embryo stage.
That’s just not a morally coherent distinction, nor is it one the law makes
Of course it’s one the law makes. IIUC it’s not even illegal for a pregnant woman to drink alcohol.
If you want to start a campaign to legalize the blinding of children, well, we have a free speech clause, you are entitled to do that. Have you considered maybe doing it separately from the genetic engineering thing?
I can’t tell if you’re strawmanning to make a point, or what, but anyway this makes absolutely no sense.
Whether you do it by genetic engineering or surgically or through some other means is entirely beside the point. Genetic engineering isn’t special.
I’m not especially distinguishing the methods, I’m mainly distinguishing whether it’s being done to a living person. See my comment upthread https://www.lesswrong.com/posts/rxcGvPrQsqoCHndwG/the-principle-of-genomic-liberty?commentId=qnafba5dx6gwoFX4a
We get genetic engineering by showing people that it is just another technology, and we can use it to do good and not evil, applying the same notions of good and evil that we would anywhere else.
I think you’re fundamentally missing that your notions of good and evil aren’t supposed to automatically be made into law. That’s not what law is for. See a very similar discussion here: https://www.lesswrong.com/posts/JFWiM7GAKfPaaLkwT/the-vision-of-bill-thurston?commentId=Xvs2y9LWbpFcydTJi
The eugenicists in early 20th century America also believed they were increasing good and getting rid of evil. Do you endorse their policies, and/or their general stance toward public policy?
Any normal person can see that a genetic engineer should be subject to the same ethical and legal constraints there as the surgeon. Arguing otherwise will endanger your purported goal of promoting this technology.
Maybe, I’m not sure and I’d like to know. This is an empirical question that I hope to find out about.
This notion of “erasing a type of person” also seems like exactly the wrong frame for this. When we cured smallpox, did we erase the type of person called “smallpox survivor”? When we feed a hungry person, are we erasing the type of person called “hungry person”? None of this is about erasing anyone. This is about fixing, or at least not intentionally breaking, people.
That’s nice that you can feel good about your intentions, but if you fail to listen to the people themselves who you’re erasing, you’re the one who’s being evil. When it comes to their own children, it’s up to them, not you. If you ask people with smallpox “is this a special consciousness, a way of life or being, which you would be sad to see disappear from the world?”, they’re not gonna say “hell yeah!”. But if you ask blind people or autistic people, some fraction of them will say “hell yeah!”. Your attitude of just going off your own judgement… I don’t know what to say about it yet, I’m not even empathizing with it yet. (If you happen to have a link to a defense of it, e.g. by a philosopher or other writer, I’d be curious.)
Now, as I’ve suggested in several places, if the blind children whose blind parents chose to make them blind later grow and say “This was terrible, it should not have happened, the state should not allow this”, THEN I’d be likely to support regulation to that effect.
i think i’m in the wrong universe, can someone at tech support reboot the servers or something? it’s not reasonable for you to screw up something as simple as putting paying customers in the right simulation . and then you’re like “here’s some picolightcones”. if you actually cared it would be micro or at least nano
Yeah, cognitive diversity is one of those aspects that could be subject to some collapse. Anomaly et al.[1] discuss this, though ultimately suggest regulatory parsimony, which I’d take even further and enshrine as a right to genomic liberty.
I feel only sort-of worried about this, though. There’s a few reasons (note: this is a biased list where I only list reasons I’m less worried; a better treatment would make the opposite case too, think about bad outcomes, investigate determinative facts, and then make judgements, etc.):
Although I want the tech to be strong, inexpensive, and widely accessible, in practice it will of course have a long road of innovation and uptake; I think I should be surprised if, 15 years after strong germline engineering technology is first relatively inexpensive (say for $10k / baby), more than half of parents in the US are using the technology.
Hopefully, many clinics would nudge parents to make a more reflective and personal choice, e.g. by asking “what about us do we want to see in our kids”, rather than just asking “how do we make a normal person / elite person / etc.”. (I’m not remotely confident clinics would do this, but.)
For better or worse, there will be a long period of time (decades, probably, at least) in which many traits, including / especially many cognitive traits (subfactors of intelligence, wisdom, curiosity, determination, etc etc) will be only weakly or not at all genomically steerable. It’s hard to define and measure many traits; and even if you can measure them, it’s hard to collect enough genotype/phenotype data to find weak polygenic associations. So there will be a lot of cognitive variation that isn’t subject to direct genomic vectoring. (Though they could be more weakly vectored if they correlate with traits, e.g. Big Five personality traits, that are studied and vectored.)
I expect the gene pool to remain quite diverse for a long time.
I expect many (most?) people to want to pass on their genes.
(As a touchpoint, though I’m probably kinda weird in this, I have some value placed on passing on most of my DNA segments at least once—though I certainly also want to pass on “worse” segments much less frequently than “better” segments.)
Most genomic vectoring methods (selection and editing) don’t affect the actual genome very much. You and I differ at O(10^6.5) genetic loci; strong editing edits O(10^2.5) loci; selection… well I’m not sure about the math, but it would probably be something like O(10^3) changes in expectation. (This doesn’t necessarily apply to whole genome synthesis, or to selection schemes that involve many donors, which could in theory hugely amplify some DNA segments in the next generation, though doing so to an extreme degree would be inadvisable.)
I therefore expect there to remain a large amount of genetic diversity. This doesn’t mean that there has to be trait-diversity, but it should imply that it would remain fairly easy (using germline engineering) to recover any previously-extant traits. In other words, even fairly extreme trait-diversity collapse would be not so hard to partially reverse (I mean, parents who want to buck the trend could do so).
For most traits, especially cognitive traits, there’s going to be lots variation that isn’t controlled by GE. E.g. nurture effects, environmental / developmental randomness, and nonlinear and weak effects of genes that haven’t been picked up in polygenic scores. This implies that, while in many cases you can significantly push the mean value of the trait among your possible future children, you can’t greatly tighten the distribution of the trait. The exact shape of this depends on the trait (how much we understand the genetics, how much overall variation there is, what the distribution-tails of the trait imply in practice for behavior). My guess is that in practice this implies you still have lots of trait-variation.
I expect that for lots of parents, for lots of traits, they just won’t especially care to or choose to genomically vector that trait much (though maybe it would be more common to avoid extremes of the trait).
For example, do I want to make my kid a bit uncommonly extraverted or introverted? Plausibly I’d develop some preference after thinking about it more, but I have less of an immediate preference compared to some other traits.
I, and I suspect many other parents, would be somewhat suspicious of many supposed polygenic scores and trait constructs and measurements. Such parents would either not GV those traits, or GV them less strongly, or only GV them after investing more thinking (and therefore would make generally better decisions).
I suspect many parents would specifically have a preference to not make genomic choices about e.g. personality traits.
I weakly expect lots of parents would want kids with a variety of values of many given traits. I would. E.g. if there’s a tradeoff between different subfactors of intelligence, I’d want one kid with a stronger genomic foundation for one of the subfactors, another with another, etc.; I might plausibly want a kid with slightly low conscientiousness (higher creativity, maybe) and another with slightly high conscientiousness (more industrious, maybe).
I expect there’s quite a lot of diversity in what parents view as desirable traits. Different parents will differently emphasize beneficence (wellbeing of the child), altruism (how the child contributes to helping others), or other criteria; and different parents will have different opinions about what traits contribute to those outcomes. I therefore suspect that collapsing all the way to just the traits that parents think they want in/for their future children, while that would plausibly be pretty bad, would still be quite far from total homongeneity.
I suspect many parents would want to see themselves in their children, i.e. pass on some of their traits. (I would.) So diversity of parents is reflected somewhat in diversity of kids.
In theory, in the medium-term (a couple generations), there should be socioeconomic feedback towards an equilibrium, where if a useful niche-strategy is undersupplied, then it’s demanded more, and some parents eventually notice and respond by trying to fill that niche. (On the other hand, it could be that some niches collapse due to undersupply, which could be harder to reverse.)
Even if the total envelope of trait-values decreases, that doesn’t necessarily mean the desirable variation decreases. In fact, what I hope, and also somewhat expect, is that you get an increased “medium frontier” of humanity: fewer extreme spikes in some dimensions (e.g. maybe no one with disagreeableness quotient >160, or sometehing), but more total weirdness because lots of kids each have several traits on which they’re 2 SDs away from the mean:
It would be interesting to poll parents to see what sorts of considerations they might take into account, and decisions they might make. One could also ask embryo screening companies.
Anomaly, Jonathan, Christopher Gyngell, and Julian Savulescu. ‘Great Minds Think Different: Preserving Cognitive Diversity in an Age of Gene Editing’. Bioethics 34, no. 1 (January 2020): 81–89. https://doi.org/10.1111/bioe.12585.
So I guess one direction this line of thinking could go is how we can get the society-level benefits of a cognitive diversity of minds without necessarily having cognitively-uneven kids grow up in pain.
Absolutely, yeah. A sort of drop-dead basic thing, which I suppose is hard to implement for some reason, is just not putting so much pressure on kids—or more precisely, not acting as though everything ought to be easy for every kid. Better would be skill at teaching individual kids by paying attention to the individual’s shape of cognition. That’s difficult because it’s labor-intensive and requires open-mindedness. I don’t know anything about the economics of education and education reform, but yeah, it would be good to fix this… AI tutors could probably improve over the status quo in many cases, but would lack some important longer-term adaptation (like, actually learning how the kid thinks and what ze can and can’t easily do).
With or without ASI, certainly morphological autonomy is more or less a universal good.
IDK what to say… I guess I’m glad you’re not in charge? @JuliaHP I’ve updated a little bit that AGI aligned to one person would be bad in practice lol.
I am the Law, the Night Watchman State, the protector of innocents who cannot protect themselves. Your children cannot prevent you from editing their genes in a way that harms them, but the law can and should.
I do think this is in interesting and important consideration here; possibly the crux is quite simply trust in the state, but maybe that’s not a crux for me, not sure.
if we had a highly competent government that could be trusted to reasonably interpret the rules,
Yeah, if this is the sort of thing you’re imagining, we’re just making a big different background assumption here.
I don’t think we have enough evidence to determine that removing the emotion of fear is “unambiguous net harm”, but it would be prohibited under your “no removing a core aspect of humanity” exception.
Yeah, on a methodological level, you’re trying to do a naive straightforward utilitarian consequentialist thing, maybe? And I’m like, this isn’t how justice and autonomy and the law work, it’s not how politics and public policy works, it’s not how society and cosmopolitanism work. (In this particular case, my justification about human dignity maybe doesn’t immediately make sense to you, but I think that not understanding the justification is a failure on your part—the justification might ultimately be wrong, I’m not at all confident, but it’s a real justification. See for example “What’s really wrong with genetice nhancement: a second look at our posthuman future”.)
Therefore, you seem to agree that going from “chronic severe depression” to “typical happiness set point” is an unambiguous good change. (Correct me if I am wrong here.)
No, this is going too far. The exception there would be for a medium / high likelihood of really bad depression, like “I can’t bring myself to work on anything for any sustained time, even stuff that’s purely for fun, I think about killing myself all the time for years and years, I am suffering greatly every day, I take no joy in anything and have no hope”, that kind of thing. Going from “once in a while gets pretty down for a few weeks, has to take a bit of time off work and be sad in bed” is probably fine, and probably has good aspects, even if it is net-bad / net-dispreferable for most people and is somewhat below typical happiness set-point. Mild high-functioning bipolar might be viewed by some people with that condition as important to who they are, and a source of strength and creativity. Or something, I don’t know. Decreasing their rates of depressive episodes by getting rid of bipolar is not an unambiguous good by any stretch.
I think a lot of people who say they anti-value intelligence are coping (I am dumb therefore dumbness is a virtue) or being tribalistic (I hate nerdy people who wear glasses, they remind me of the outgroup). If they perceived their ingroup and themselves as being intelligent, I think they would change their tune.
That’s all well and fine, but you’re still doing that thing where you say “X is unambiguously good” and I’m like “But a bunch of people say that X is bad” and you’re like “ha, well, you see, their opinion is bullshit, betcha didn’t think of that” and I’m like, we’re talking past each other lol.
Anyway thanks for engaging, I appreciate the contention and I found it helpful even though you’re so RAWNG.
I’m genuinely unsure whether or not they would. Would be interesting to know.
One example, from “ASAN Statement on Genetic Research and Autism” https://autisticadvocacy.org/wp-content/uploads/2022/03/genetic-statement-recommendations.pdf :
ASAN opposes germline gene editing in all cases. Germline gene editing is editing a person’s genes that they pass down to their children. We do not think scientists should be able to make gene edits that can be passed down to a person’s children. The practice could prevent future generations of people with any gene-related disability from being born. This is eugenics and a form of ableism.
ASAN opposes non-heritable gene editing for autism. This is when scientists edit a person’s genes in a way that can’t be passed down to their children. ASAN is against ever using this kind of gene editing for autism. We think it would be used to treat or “cure” autism. We do not want “cures” for autism. We want to continue being autistic. We want there to be rules saying people can’t use genetic research to find a “cure” for autism. We are setting this standard for autism and the autistic community because it is what most of our community members believe.
Not directly comparable, but related. (I disagree with their reasoning and conclusions, I think.)
Interestingly, they make the same suggestion I mentioned above:
Some disability communities might want non-heritable gene editing. For example, some people with epilepsy are okay with non-heritable gene editing for the genes that cause their seizures. We think it should be up to each disability community to decide if they are okay with nonheritable gene editing for their disability. Researchers and policymakers should listen to each disability community about how that community feels about non-heritable gene editing.
Yes, blind people are the experts here. If 95% of blind people wish they weren’t blind, then (unless there is good reason to believe that a specific child will be in the 5%) gene editing for blindness should be illegal.
This is absolutely not what I’m suggesting. I’m suggesting (something in the genre of) the possibility of having that if 95% of blind people decide that gene editing for blindness should be illegal, then gene editing for blindness should be illegal. It’s their autonomy that’s at issue here.
It’s straightforward to disprove: they should be able to argue for their views in a way that stands up to scrutiny.