site banner

Culture War Roundup for the week of February 26, 2024

This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.

Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.

We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:

  • Shaming.

  • Attempting to 'build consensus' or enforce ideological conformity.

  • Making sweeping generalizations to vilify a group you dislike.

  • Recruiting for a cause.

  • Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.

In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:

  • Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.

  • Be as precise and charitable as you can. Don't paraphrase unflatteringly.

  • Don't imply that someone said something they did not say, even if you think it follows from what they said.

  • Write like everyone is reading and you want them to be included in the discussion.

On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.

6
Jump in the discussion.

No email address required.

Genetic modification seems so obviously to be progress but I am starting to expect it to face a great deal of political backlash.

Did you miss GATTACA? Beggars in Spain? Hysteria around designer babies when Dolly was cloned, or the human genome draft was published?

change the genetic code so that they create Shaquille O’Neill physical traits plus 250 IQ.

That's just not anywhere close to realistic with our current level of technology and understanding. You could try cloning Shaq or whoever you think is smart, but we're laughably far away from editing your fertilized embryos for traits in that way. Like, it wouldn't happen in your lifetime even if the FDA were nuked tonight and we just did whatever we want to embryos for the next couple decades, ethics be damned.

Beggars in Spain

I've read it. I recall liking it. It is very obscure, right?

I'm not certain how popular it is, but I've seen it come up in a few contexts in the circles I travel in. People around here will cite obscure Larry Niven books fairly frequently ('On the gripping hand'...thankfully the much more crass 'Rape my lizard!' from the same novel never caught on), so I thought there were decent odds that Beggars in Spain was also well known. Particularly given the themes of transhumanism.

Like, it wouldn't happen in your lifetime even if the FDA were nuked tonight and we just did whatever we want to embryos for the next couple decades, ethics be damned.

Are you sure? Could anyone predict forty years of ethics-department-free AI progress like that? We were laughably far from good image-gen AI 5 years ago. And genetic augmentation is a particularly ethics-beholden field. We've got CRISPR which can change genes in fairly effective ways, it's not like FTL travel where we've got no good approach.

We've got CRISPR which can change genes in fairly effective ways

How many genes can it change? If we want to change a thousand of genes (because e.g. our utility function asks for it) and have no faulty changes in unintended genes, can it deliver that?

If someone wants to make a genetically modified plant, they can just try and change and just discard if edits were faulty, with humans we want reliable ones.

It can't change thousands of genes but it does demonstrate that it's possible to change genes, that presumably there are other ways to change many genes reliably. It's like how the first powered flight makes helicopters, jump-jets, jet-packs and autogyros much more plausible.

And hey, we're talking about a world where 'the FDA were nuked tonight and we just did whatever we want to embryos for the next couple decades, ethics be damned'. We can't even imagine how rapid development would be, it'd be a different paradigm.

Are you sure? Could anyone predict forty years of ethics-department-free AI progress like that?

It's true, predicting the future is a fool's game and betting against progress doubly so. We'll never be able to resolve our bet given the near certain continued existence of the FDA.

That being said:

  1. biologists (and I count myself in this camp) are morons who can't do math, code or do anything beyond draw pointless arrow diagrams. Moreover, the incentive structure actively pushes us away from solving any of these issues and instead focusing on shorter timelines, smaller scales and splashy headlines instead of any kind of substantive understanding. PGS gets around all of this because you can let biology do the work for you, but it comes with a host of other problems. You mentioned CRISPR though, so let's go with that.

Say you want to use CRISPR to...I don't know, increase the size of your gut to accommodate the caloric needs of the giant brain your 250 IQ Chinese supersoldier is going to have. There is no 'gut size' gene that you can just augment the expression of, there are massive, interlocking transcriptional networks who need to be turned on at the right times and in the right cell types. This is likely to be far beyond our capability to understand ever, so the only reasonable path forward is building an AI oracle to understand it for us. It's either going to 1) Need monstrous datasets that we probably can't generate in a reasonable way yet 2) Molecular dynamics simulation is impossibly computationally expensive, so figure out some other abstracted simulation method 3) ??? someone else tell me how they'd envision this working, my imagination fails me.

  1. There is no 'gut size' gene; there are dozens if not hundreds of genes you would need to alter, and moreover, alter in ways that are temporally, spatially and functionally correct. CRISPR just isn't capable of doing that with the precision or reliability you'd need; it's great at knocking out 1 to a handful of genes, mediocre-to-bad at activating genes, and mediocre at silencing genes - and these latter two functions are transient, so you'd need to find a way to keep expressing the CRISPR and gRNA. Probably we're again going to rely on godlike AI designing new methods for manipulating gene expression on broad scales, or maybe some Kwisatz Haderach breeding program over generations as we slowly introduce the changes we need.

  2. Delivery to somatic tissues isn't currently possible in a meaningful way, although I'm optimistic we might actually solve this in a reasonable timescale. And I suppose you'd want to edit germline cells regardless, but I thought I'd point it out.

I'm most pessimistic about (1), and moreover a decided lack of interest in TPTB (to be clear, the academic establishment. I doubt the deep state cares overmuch) in understanding these systems in a way that we could build or intelligently edit them ourselves. Godlike AI is the wildcard, but at least so far, all AI can do is hold up a mirror and regurgitate the same garbage that we write in review papers. And Eliezer tells me we're all dead in that scenario anyways, no?

I'm not a developmental biologist and only tangentially touching on human genetics so I wouldn't say this is authoritative. But that's been my impression over the last decade or so.

  1. ??? someone else tell me how they'd envision this working, my imagination fails me.

A whole mountain range of skulls of the failed experiments when tweaking genes for "well, let's try this and see if it works"/"oh no, it didn't, well fire up batch of embryos number one thousand and let's go again!" and breeding programmes and all the horrors that cutting-edge tech and no ethical oversight will permit. Morals are for losers, baby, nature is red in tooth and claw, we ain't getting 250 IQ without breaking a hatchery's worth of eggs, and they're just clumps of cells that we permitted to go all the way through gestation anyhoo. Not like they're people or anything.

Why do we need such intensive transformations? We have large populations of very smart people, large populations of very physically capable people. We can sequence genes easily, see which ones are associated with intelligence and physical performance. 250 IQ might be a 1 in a hundred billion genius but still basically possible from blind chance. All we'd be doing is fixing the dice to get an impossibly rare natural creation. Maybe he'd be pretty hungry but we have plenty of calories.

I bet there'd be a bunch of problems with mental illness, you'd need to refine it with time and experience. I know it's more complicated than 'make a statistical model of what genes the smartest people are most likely to have and tick the boxes' but we'd have time for trial and error. Forty years is multiple gamechangers worth of time.

As for how one would edit those genes, I don't know. Technology advanced quickly in the last forty years, we didn't have anything like CRISPR forty years ago. We couldn't even sequence human genes.

all AI can do is hold up a mirror and regurgitate the same garbage that we write in review papers.

Come on, it can do more than that. I got it to write some decent Top Gear fanfiction. Google had theirs doing protein folding, which is vaguely relevant.

Even more relevant: https://nyulangone.org/news/new-artificial-intelligence-tool-makes-speedy-gene-editing-possible

Not perfect but surely proof AI is useful on this front?

With no ethics, and a big budget you could go very fast.

Females develop eggs after 20 weeks so you could make 1000 per generation, polygenically screen them all, pick the best and iterate.

In just over a year you have 3 generations and the pick of 1 in a billion (of descendants of your starting stock)

No, I don't think so. We can already clone Von Neumann, and we can't actually engineer smarter people than that in the way we engineer airplanes or computers because we don't know how either intelligence or neurology work at that level.

We can certainly do iterated embryo selection on top of Von Neumann ... but I don't think that's going to work that well, you're running into non-additive effects there.

You can just take Von Neumann's genome and eliminate very rare variants (which are usually deleterous which is why they are rare), not even him was free from it (early cancer death).

Maybe that gets him 5 IQ points, certainly not 100. Or very likely that he was selected as one of the smartest out of billions people means that there's just less juice to squeeze out of that.

You don't need to know how intelligence works, you just need to sequence enough people in genome wide association studies, build a polygenic score and then run it

Right, that doesn't get you to 250 IQ or above existing variation I think.

250 IQ is 10 standard deviations from mean IQ 100, SD=15

If we have 1000 offspring per generation via egg harvesting embryos, taking the top 1% (10) they should be 2.33 SD from the mean

With 80% heritability, response to selection per generation is 1.86 SD

Thus it will take 5.36 (6) generations of selective breeding to get 250IQ with 1000 offspring per generation

Or 6 * 20 weeks = 2.3 years

If egg/sperm are from +3 SD donors, it's 3.76 (4) generations, or 1.53 years

Please consider donating towards my volcano lair lab on Kickstarter

Right, I'm suggesting that the mathematical model stops applying. Like, imagine humans are now cars. You can make a faster car than you did last time by copying the techniques of all the best car factories (this is sorta what embryo selection does). Maybe you could even go 10-20% faster than the fastest car ever by doing that, though I think non-additive effects will prevent that. But you can't make a car that goes at 1000 mph by picking out the best techniques from existing car factories. You'd need to do more technical design work than that, or have a LOT of new mutations and natural selection on them.

Maybe you could even go 10-20% faster than the fastest car ever by doing that

well if you somehow get established breed of sperm and eggcells that are 10-20% better than the best, wouldn't be that accomplishment by itself? You can have population of high IQ humans that can procreate naturally and their regression to mean is not to 100 or 105 but 160. Good?

Yes I agree that it would be good to do that, and have made many posts to that effect. Although we don't even need embryo selection for that since we can already identify smart parents, it's useful.

It's possible to study nonlinearity... though of course you are right, it's difficult and one of problems here that due to ethics problem nobody gets dataset on million IQ - genome pairs (we need to oversample at edges of distribution too)

You can't take half the components of a Ferarri and half the components of a Ford pickup, mix them together and have a working car. The piston of one wouldn't fit in the cylinder of the other.

But you can mix the genomes of males and females of the same species.

Cars have very few components so the variation is eg swapping out this brand of muffler for another one - they must all fit together. Genetic variation is extremely small (modifying less than a billionth of the system) - and mostly independent of other variation.

Imagine a car with 10,000 tunable components, that can vary without breaking the machine (life is robust to variation)

There are 8 billion cars, almost completely stock. Some have a few parts well tuned, and are fast, some have a few parts detuned and are slow

The fastest cars have 500 components perfectly tuned and are 4 or 5 standard deviations faster

What I'm saying is we look at millions of cars tuning matched to speed. We use this to work out what the best tuning is, then we select for that, making a car that uses existing components, but the combination has never existed before naturally.

We KNOW we can rapidly selectively breed animals that vary enormously from the natural stock. Look at racehorses, dogs, milk production in cows, the giant extremely fast growing chickens we eat today that lay eggs at phenomenal rates etc etc

We KNOW we can rapidly selectively breed animals that vary enormously from the natural stock.

Yes, that involves actually growing these animals to adults and testing them, but not hoping that our linear regression (19th century math) PGI is true.

Your position here is not the mainstream one, if you asked a hundred researchers in genetics if they thought your method could reach 250 IQ they would disagree. I was trying to do the sort of broad explanatory analogy we do a lot of here because it's a general interest place, but there are technical reasons why you're wrong. If you want to convince anyone, you should use rhetoric that acknowledges you're in an unfavored position.

More comments

Females develop eggs after 20 weeks so you could make 1000 per generation, polygenically screen them all, pick the best and iterate.

Those eggs are immature. I'm not a developmental biologist, but would you expect in vitro maturation protocols to work on eggs forcibly harvested from a 20 week old fetus?

I'm also not confident that there's enough genetic diversity starting from one person to get a true 1 in a billion; won't there be a bunch of alleles where neither parent has what you want? I admit that this may be a nonissue if most of the alleles you want are relatively common, I don't have a good handle on the numbers here.

Put it a different way - Do any of us have a 1 in a billion chance of giving birth to Shaq? I would tentatively guess no, modulo some genetic conditions like acromegaly. Do you have evidence that this is true?

In just over a year you have 3 generations and the pick of 1 in a billion (of descendants of your starting stock)

But what then? You have one embryo. Somatic cloning? Things are getting pretty complicated my man.

Oocyte in vitro maturation (IVM) is a thing. I am not sure if anyone has used it on foetal eggs, but it is likely possible.

You aren't restricted to the original stock you can introduce new sperm each generation

To scale it up, you could potentially encourage eggs to divide (like identical twins) then sequence one, and if its the one you want, keep splitting

I think these things are not too hard to solve, you just need time and money, and the will (and lack of ethical restraint)