To give some numbers, consider 10k sperm. We look in the embryo selection powers:
For 10k, i.e. e4.0, you get about 2.7 raw SDs, for embros; for sperm you multiply by the sperm SD scale, which is 1/√2 because sperm have half the variance of embryos. That’s about 1.9 SDs. Then you do a round of embryo selection, with say 10 embryos, for about .7 more raw SDs (because you’re mainly selecting on the maternal DNA). Total is 2.6 raw SDs. As IQ points this is like 13 IQ points. It’s pretty interesting, but not groundbreaking; and even with 10k sperm, could actually be pretty expensive, like would plausibly be at least several $100k. Another order of magnitude sperm would be another OOM expensiver, and only a very small bump in power. I haven’t run the numbers for disease reduction, and yeah it would probably be pretty nontrivial (because the marginal returns for diseases are high at the beginning, starting at a normal genome), but yeah it also does not qualify as strong GV.
To be clear, this would be awesome and someone should totally try, and it could plausibly synergize with other stuff. It’s just not a priority for getting to strong GV.
I think the numbers just kind of suck. I didn’t go into them much because gamete selection seems largely hypothetical. Like, the procedure here seems kinda expensive per-spermatocyte (guy who divides into 4 sperm). I gesture at how to compute it here: https://www.lesswrong.com/posts/2w6hjptanQ3cDyDw7/methods-for-strong-human-germline-engineering#The_power_of_single_gamete_selection
To give some numbers, consider 10k sperm. We look in the embryo selection powers:
For 10k, i.e. e4.0, you get about 2.7 raw SDs, for embros; for sperm you multiply by the sperm SD scale, which is 1/√2 because sperm have half the variance of embryos. That’s about 1.9 SDs. Then you do a round of embryo selection, with say 10 embryos, for about .7 more raw SDs (because you’re mainly selecting on the maternal DNA). Total is 2.6 raw SDs. As IQ points this is like 13 IQ points. It’s pretty interesting, but not groundbreaking; and even with 10k sperm, could actually be pretty expensive, like would plausibly be at least several $100k. Another order of magnitude sperm would be another OOM expensiver, and only a very small bump in power. I haven’t run the numbers for disease reduction, and yeah it would probably be pretty nontrivial (because the marginal returns for diseases are high at the beginning, starting at a normal genome), but yeah it also does not qualify as strong GV.
To be clear, this would be awesome and someone should totally try, and it could plausibly synergize with other stuff. It’s just not a priority for getting to strong GV.