Gurkenglas comments on Awww, a Zebra - Less Wrong

23 Post author: Eliezer_Yudkowsky 01 October 2008 01:28AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (50)
Tags: humor hypocrisy personal

Comments (50)

Sort By: Old

You are viewing a single comment's thread. Show more comments above.

Comment author: Gurkenglas 19 May 2014 04:22:25PM 0 points [-]

Intelligence is roughly a function of genetics and some conglomeration of upbringing and chance. If you are looking for the human with the best genetics, picking from the most intelligent humans you can find is therefore not the optimal way to go.

Comment author: gwern 19 May 2014 10:25:38PM *  5 points [-]

Intelligence is roughly a function of genetics and some conglomeration of upbringing and chance.

Yes.

If you are looking for the human with the best genetics, picking from the most intelligent humans you can find is therefore not the optimal way to go.

No...? If scores on an IQ test are a joint product of genes & environment, then selecting the top scorer on an IQ test and using their genes will produce offspring with the highest scores on average compared to picking at random from the lower-scorers. To do better you'll need additional information we don't have (eg at the moment, no one can sequence a genome and extract an accurate predicted intelligence).

Comment author: TheOtherDave 19 May 2014 10:52:39PM 4 points [-]

Somewhat tangentially... couldn't you do better by identifying IQ-inhibiting and IQ-enhancing environmental factors and weighting IQ scores based on those factors? If Sam's IQ is 5% lower than Pat's but Sam has lived in an environment 5x worse for developing IQ, if I'm interested in genetics it seems I'd do better with Sam.

Comment author: gwern 19 May 2014 11:24:09PM *  4 points [-]

You'd need to know the elasticity of intelligence and environments (5x on what scale? And does -5% really indicate outperformance on the genetics side?), which I'm not actually sure we know, and much of the 'environment' contribution is nonshared - immeasurable, random, we don't know what it is. But hypothetically, you could do slightly better by trying to measure environment & control for it, yeah.

Comment author: Vaniver 21 May 2014 08:50:24PM 1 point [-]

To do better you'll need additional information we don't have (eg at the moment, no one can sequence a genome and extract an accurate predicted intelligence).

Actually, we could get some information that would help here- IQ up the ancestral tree. The correlation between grandparent and grandchild IQ is higher than one would expect from stacking two independent parent-child relationships.

Comment author: gwern 05 August 2014 09:45:28PM *  1 point [-]

Yes. I'm not sure how much they'd add; I ran into an interesting observation about this with regard to estimating cows' milk production based on their relatives and on SNPs, where the comparison runs the other direction:

The R^2 values were converted to realized reliabilities by dividing by mean reliability of 2008 daughter deviations and then adding the difference between published and observed reliabilities of 2003 parent averages. When averaged across all traits, combined genomic predictions had realized reliabilities that were 23% greater than reliabilities of parent averages (50 vs. 27%), and gains in information were equivalent to 11 additional daughter records.

So if an old SNP chip can add that much information in terms of family records, the family can't matter that much.

Powered by Reddit Powered by Reddit