But also people on the outer edges will be rare in the data this statement comes from, so it may not hold for them.
I think there should be more investigation somewhere[1] of whether generic cognitive improvements hold in the upper limits[2], cause it seems like the main way they could be high-impact. (by augmenting the best EAs/alignment researchers, under the view that impact scales closer to exponentially than linearly with intelligence.)
‘0.5 standard deviations’ seems very significant on on the outer edges of the distribution. Because of https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule#Table_of_numerical_values.
But also people on the outer edges will be rare in the data this statement comes from, so it may not hold for them.
I think there should be more investigation somewhere[1] of whether generic cognitive improvements hold in the upper limits[2], cause it seems like the main way they could be high-impact. (by augmenting the best EAs/alignment researchers, under the view that impact scales closer to exponentially than linearly with intelligence.)
In the sense of ‘by now, civilization should have done this’, but I’m not saying anything about whether it would be good to focus on at this point.
(and maybe it has received that focus, and I just don’t know where)
(eg whether fixing an iron deficiency in someone who is already at +3sd pushes them to +3.5sd)