I think Silverman's argument is interesting - I don't know if she says it quite this way, but it seems like she's saying, either use the GAI or lower the bar for admission to a gifted program to 120 for full-scale IQ.
I kind of picked up from another page on the gdc site (I posted it on this thread already, I think) that the idea is to remove the speed question for most/all questions of gt *BUT* speed is important for good performance in school and other traditional settings so..?
(why not work kids this young into that speed, if their subtests which better correspond with 'g' are really nice looking? )
I am assuming that speed inevitably improves with age, and practice - but these tests are more of a statement of a 'static' percentile of speediness, right?