One of the great problems in designing an Artificial Intelligence we can control is our own lack of sufficient intelligence. Partial failure here (success as designing an AI, failure at control) would result in a future where we're all dead or pets to greater beings at best.
It would be helpful if biological humans were smarter. Nick Bostrom discuses a cognitive-enhancement process through repeated genetic selection. He argues that if the heritable genetic characteristics of intelligence can be assessed by large-scale correlation studies, a 1 in 100 selection pressure (select the "best" out of 100) towards correlates of higher cognitive function repeated over 10 gamete (sperm + egg) generations can result in massive IQ boosts in the selected individuals (upwards of 100 IQ points total gain at the end of the selection process compared to the baseline - i.e. us).
If Bostrom's scenario is realistic, what would be the social implications of such a selective process if there were, say, 1% adoption of this method among the population? In other words, how smart can we get, and still retain a viable society?