So we all agree that all this "AI is going to kill us all, we must slow the pace!" stuff coming from Altman/Musk/Amodei is just their way of lobbying for the ban open source LLMs, right?
Bird account
@dialecticbio.bsky.social
3.1K followers 510 following 520 posts
The "The Dialectical Biologist" account. Structural biology & anti-racism. Loves: birds, proteins, Linux, Lewontin Hates: race science, racists, transphobes
![she does, and moving from ancestry groups to racial labels, she describes a near future in
which we will have a polygenic score “that is as strongly related, statistically, to academic
achievement in Black students as it is in White students.” [p. 191] (This conflation of an-
cestry and race is particularly consequential when considering Black Americans, who vary
greatly in their proportions of recent African and European ancestries, in ways that are
correlated with both geography and economic opportunity (e.g., Micheletti et al., 2020;
Baharian et al., 2016).) Harden warns the reader, however, that such comparisons be-
tween Black and White Americans would be “scientifically and ethically wrong” [p. 192]. To
explain why comparisons of genetic effects between groups are invalid, and will remain so
even when polygenic scores have been constructed for both, she refers to statistical issues
that arise from confounders and the lack of environmental homogeneity. But as we outline
in our review, similar problems, differing in degree rather than in kind, apply to the move
from within-family studies to standard GWAS.
Harden interprets our charge that she is trying to have it both ways as somehow related
to polygenic scores being “simultaneously meaningful. . . and limited, by virtue of their con-
textual dependence and imperfect portability” , adding “As the author Maggie Nelson wrote:
‘There is a lot to be learned from having it both ways’ ”. In reality, we were making the
point that she cannot overlook the effects of confounding when interpreting polygenic scores
within European-ancestry GWAS individuals–let alone within the racial grouping of “Black
Americans"–only to resurrect them when seeking protection against the implications of her
claims for between ancestry-group comparisons. The truth is that polygenic scores for ed-
ucational attainment are also confounded within ancestry groupings, in ways that greatly
complicate interpretatio](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ltfn4ycf2afx23cmmlfynfwn/bafkreicc2v7d22iugmdiydou4bylaugrnhef67tz2z35piofmm5y774cvm)
![How did they claim to distinguish nature from nurture? Consider just one item from Harden and her co-authors’ laundry list: “age of first sexual encounter.” Their study depends on the notion that the earlier this takes place, the more evidence we have that a person is innately driven toward general risk-taking by an absence of self-control. (A Christian morality pulls the strings behind the scenes of this science: where else does one derive the concept of abstinence as virtue?) Still, they make a noteworthy exception: “[W]e coded all individuals with responses below the age of 12 as missing [i.e., as providing no data on the question], as this could reflect childhood sexual abuse rather than earlier onset of sexual behavior.”
One does not need to be a psychologist—in fact, maybe it helps not to be—to spot the shortcomings of this quick fix. Among other things, it is possible to be sexually abused at any age. Harden and her colleagues obviously recognize that they need to filter out this possibility, in order to even pretend their number is a proxy for innate virtue or vice, as opposed to external happenstance, but the data available provides no rigorous way of distinguishing nature from nurture.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ltfn4ycf2afx23cmmlfynfwn/bafkreidy7rguiwymxwnasynp4jey6jcw7xiujxjaabcmzunnjz6xvrorxu)
![Fig. 1. Genome assemblies, annotations, and transcriptomes.
(A) Species representing the four sugar-consuming groups. [Bird photos (top to bottom): Steven L. Hunter (ML426078561), David J. Southall (ML431374521), Gerald R. Allen (ML633429605), and Charles Davies (ML205719751); photos accessioned into the Cornell Lab of Ornithology | Macaulay Library] (B) Cladogram and diet of the birds included in the study. On the basis of the percentage of dietary nectar and fruit intake, we delineated four groups of core sugar-consuming birds (red font) and core non-sugar controls (blue). Core non-sugar control sets were defined as groups phylogenetically close to sugar-consuming clades and comprising a similar number of species but not taking any nectar in the diet. These were used as direct controls in the downstream analyses. (C) Comparison of genome assembly contiguity. The contig N50 value indicates that 50% of the assembly consists of contigs of at least that size. The nine newly sequenced birds are indicated with a blue circle. (D) Status of 15,133 ancestral bird genes, classified by TOGA as those with an intact reading frame (blue), with gene-inactivating mutations (premature stop codons, frameshifts, splice site disruptions, or deletions of exons or entire genes; orange), or with missing or incomplete coding sequences owing to assembly gaps or fragmentation (gray). (E) Status of 8338 near-universally conserved Aves genes (BUSCO, odb10) in annotations generated in this study (green circles) or NCBI RefSeq annotations. (F) Tissue-specific transcriptomes generated for pairs of sugar-consuming and non-sugar control species. Black arrows specify the direction of DESeq2 analysis (for details, see methods and table S11).](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ltfn4ycf2afx23cmmlfynfwn/bafkreibzhx7fuidph2njhmsrja2vzgnobkwzt3shlnmezccfskyiaiqnpy)
