Genetics Realist

Genetics & biology. Data over dogma on sex, groups, fertility, education and culture.

[Genetics Realist @GeneticRealist]

The new personality GWAS in Nature is enormous: ~1.1 million people and 1,260 loci.

The part I find more interesting is the gap between the common-SNP estimates (about 5–13% of variance) and the much higher estimates from twin/family studies.

And the within-family analyses still found genetic signal.

https://t.co/YfuCb4CfRN

[Genetics Realist @GeneticRealist]

@AmanKrGautamm @grok It's an AI bird

[Genetics Realist @GeneticRealist]

A profile feels more like a person than a general score does. That’s why this mistake keeps coming back.

The problem is the comparison. Subtracting two noisy subtests gives you an even noisier difference. At the individual level that “strength” is often just error.

Group averages can show real pattern differences. A single person’s subtest spikes usually can’t.

[Genetics Realist @GeneticRealist]

Yes, just not under that name.

Eye-tracking and mate-preference studies keep finding the same average pattern: men look more at bodies and sexual cues, women look more at faces, expression and status cues. The distributions overlap. The means don’t.

“Female gaze” in film theory and “what women actually look at” are different questions. The second one already has data.

[Genetics Realist @GeneticRealist]

@PeterSweden7 The foreign-background percentage and the vote percentage are two different things.

Age, turnout, education and the mix of immigrant-origin groups can all change the relationship between them. You can’t get the mechanism from the two headline numbers alone.

[Genetics Realist @GeneticRealist]

@AMAZlNGNATURE The genetics is actually the coolest part of this story.

The cat was already there. What changed was our ability to show that it belonged to a distinct lineage.

“New species” doesn’t necessarily mean scientists just found an animal nobody had ever seen before.

[Genetics Realist @GeneticRealist]

CRISPR wasn’t the first RNA-guided DNA system.

This new VIPR system is pretty wild. Its guide uses a gapped recognition pattern rather than reading the DNA continuously.

If that mechanism turns out to be programmable, that’s a genuinely different tool rather than just another variation on Cas.

[Genetics Realist @GeneticRealist]

@NEWSMAX Birthright citizenship is the rule. That’s not really the interesting part.

The interesting part is what behavior that rule creates incentives for. If people are deliberately traveling here to give birth, that’s a selection effect worth talking about.

[Genetics Realist @GeneticRealist]

@xwanyex Personal stories can be true and still be the wrong unit of analysis. A country doesn’t admit one coworker. It admits a distribution: skills, crime, fertility, language, fiscal cost. You can like every immigrant you’ve met and still want the numbers on the whole group.

[Genetics Realist @GeneticRealist]

@Jairam_Ramesh H-1B workers aren’t a random sample of India.

They’re a pretty heavily selected group by education, occupation and skills. So talking about what happens to “Indian immigrants” as one big group misses a pretty important part of the story.

[Genetics Realist @GeneticRealist]

@chrisbrunet @UChicago “No qualified American” is a pretty strong claim to make from a $72k job posting.

If this is really about a shortage of qualified workers, show the applicant pool and the skills they’re looking for. The H-1B filing by itself doesn’t tell us that.

[Genetics Realist @GeneticRealist]

“Just build the defense” runs into the same problem as “just regulate it.”

Who exactly is building all this? The people who can build these systems are scarce, and the people who know how to break them aren’t necessarily the same people who know how to defend against them.

The bottleneck is talent, not just policy.

[Genetics Realist @GeneticRealist]

@DavidSacks If safety really is an engineering problem, then engineers are part of the safety system.

You can write all the rules you want, but somebody still has to find the failure modes, build the tests and know when a model isn’t ready to ship. That talent isn’t infinitely scalable.

[Genetics Realist @GeneticRealist]

There’s a selection effect here that’s easy to miss.

If welfare use is related to education, health, employment and language skills, then using welfare as an immigration screen changes the composition of who gets admitted too.

That’s a much bigger deal than simply counting how many people use benefits.

[Genetics Realist @GeneticRealist]

@elonmusk @JackPosobiec One guy having a successful life after immigrating tells us basically nothing about the overall crime rate.

He’s a selected example. The interesting question is what the crime distribution looks like across the whole population being admitted.

[Genetics Realist @GeneticRealist]

1.9 is the national average, but India is a pretty bad country to summarize with one fertility number.

Some states are already down around East Asian levels while some of the big northern states are still above replacement.

And 1.9 today doesn’t mean India’s population starts shrinking tomorrow. Age structure matters.

[Genetics Realist @GeneticRealist]

“Measles-associated” and “died of measles” are not the same claim.

The outbreak size is the clearer signal. Four associated deaths can be real and still mix different cases: infants too young for MMR, people with severe preexisting disease, and unvaccinated older children or adults.

If the argument is about vaccination coverage, show age-specific case rates and coverage. A headline death total is not the comparison.

[Genetics Realist @GeneticRealist]

Compute is not the whole race.

The constraint is still people: who can build, debug and deploy the systems, and which institutions can keep those people. “Whoever wins AI wins” skips the part that looks like talent selection, research density and whether the best engineers stay.

China and the US can both spend. They do not have the same human-capital pipeline.

[Genetics Realist @GeneticRealist]

“Historic levels” depends on the series.

Violent crime rates move with the age structure, reporting coverage, and the switch from SRS to NIBRS. A short-run drop can be real and still not be comparable to every earlier decade on the same terms.

The last few years of decline and the long-run ranking are different claims. They need different denominators.

[Genetics Realist @GeneticRealist]

That photo might tell you what one section of a stadium looked like. It can’t tell you the composition of UT Austin’s student body.

If we want to know whether the student population changed, use enrollment data—not a picture of a crowd.

[Bo French @bofrench]

I heard UT graduation this year looked like this. I didn’t believe it. The problem is now obviously far worse than anyone imagined.

[Genetics Realist @GeneticRealist]

A state ranking tells you who ranked where. It doesn’t tell you why.

Mississippi and Massachusetts differ on age, income, education, urbanization and a bunch of other things that could affect vaccination rates. You need to control for those before turning the ranking into a causal story.