3 ms·

It's map of the tested embryos projected on to the normal curve of some IQ distribution. Could be local to the parents or the company, or could be global averag

It's map of the tested embryos projected on to the normal curve of some IQ distribution. Could be local to the parents or the company, or could be global average. 68% of people fall within one standard deviation, 95% within two, 99.7% within three, etc. It's just a way of shorthanding the description of an individual result. A 1σ (sigma) result is in range of 68% of results, etc. The bottom axis is standard deviations where that ~3.6 result matches an individual scoring top among 10k individuals. Not sure how often high IQ embryos are produced with relation to high IQ births. I'd suspect that there are more high IQ embryos than high IQ births just given the sheer quantity of unborn embryos. That's the point of the screening, I'd imagine. If you really want a high IQ baby, and we can test for that, you'll have a higher chance if you use one of the embryos that test to the right of curve.

gus_massa 5mo ago
Thanks, the graphics now makes sense for me. Father ~117 IQ, Mother ~121 IQ. ~40 embryos, distribution μ=~117, σ=~7. Numbers read by eye, don't believe they are very accurate. Assuming my numbers are not very wrong, that embryo at ~135 IQ is like 3σ above in this family embryo distribution, that is 99.7% in this family embryo distribution. One in the 99.7% with only 40 samples is a little too lucky, but not unheard, specially if we assume that not everybody post the graphics. Assuming only the parents with a lucky case post the graphics. I suspect I'm underestimating σ in this family embryo distribution or under counting the embryos. It would be nice to have the raw data. --- Anyway, ignoring any ethical concern, I'm very skeptical in general, but in particular about some details. From a followup tweet https://x.com/GeneSmi96946389/status/2058290505004572762 https://x.com/GeneSmi96946389/status/2058290505004572762 > Herasight's IQ predictions correlate with real IQ at about 0.45 when you test them in real siblings: From https://herasight.substack.com/p/cogpgt https://herasight.substack.com/p/cogpgt > achieving a within-family correlation with latent general cognitive ability (‘g’) of 0.456 in UK Biobank adults and 0.448 when meta-analyzed with the adolescent ABCD cohort The problem with correlation is if you have a predictor, you add +10 to everyone or multiply it *1.1 and the new predictor has the same correlation, but with the new predictor everyone is a genius. So perhaps their predictor is good enough to pick that embryo inside that lot, but it may be very unreliable to know the real IQ. And "in real siblings" is very suspicions about how this predictor translates from a family lot to the general population. Also, they are using too many decimals, 0.456 and 0.448 is a red flag unless they have a huge lot of people in the test. And also, no embryo below 100% is extremely suspicious.

gus_massa 5mo ago

Thanks, the graphics now makes sense for me. Father ~117 IQ, Mother ~121 IQ. ~40 embryos, distribution μ=~117, σ=~7. Numbers read by eye, don't believe they are very accurate. Assuming my numbers are not very wrong, that embryo at ~135 IQ is like 3σ above in this family embryo distribution, that is 99.7% in this family embryo distribution. One in the 99.7% with only 40 samples is a little too lucky, but not unheard, specially if we assume that not everybody post the graphics. Assuming only the parents with a lucky case post the graphics. I suspect I'm underestimating σ in this family embryo distribution or under counting the embryos. It would be nice to have the raw data. --- Anyway, ignoring any ethical concern, I'm very skeptical in general, but in particular about some details. From a followup tweet https://x.com/GeneSmi96946389/status/2058290505004572762 https://x.com/GeneSmi96946389/status/2058290505004572762 > Herasight's IQ predictions correlate with real IQ at about 0.45 when you test them in real siblings: From https://herasight.substack.com/p/cogpgt https://herasight.substack.com/p/cogpgt > achieving a within-family correlation with latent general cognitive ability (‘g’) of 0.456 in UK Biobank adults and 0.448 when meta-analyzed with the adolescent ABCD cohort The problem with correlation is if you have a predictor, you add +10 to everyone or multiply it *1.1 and the new predictor has the same correlation, but with the new predictor everyone is a genius. So perhaps their predictor is good enough to pick that embryo inside that lot, but it may be very unreliable to know the real IQ. And "in real siblings" is very suspicions about how this predictor translates from a family lot to the general population. Also, they are using too many decimals, 0.456 and 0.448 is a red flag unless they have a huge lot of people in the test. And also, no embryo below 100% is extremely suspicious.