Statistical Challenges with Making Super IQ babies — LessWrong