An ultra-smart ultra-selfish hacker would sit down, do the math, and spend 5 years killing it on Wall-street doing quasi-legal machine trading in total safety. Rather than risk imprisonment and permanently damaging his career prospects over a few thousand dollars from Grandma's credit card.
Case in point:
Carl Gugasian, graduated with a degree in electrical engineering. Got a masters in systems analysis, and was a PhD student in statistics. His intelligent, methodical approach to robbing banks let him get away with it for 30 years. He earned about 2 million dollars doing it, which is about a million less than his undergraduate engineering degree alone would have earned him during that same period if he'd just gotten a regular job.
He also eventually got caught, through pure dumb chance he couldn't have planned for, and spent 20 years in prison. So not only did he make less money being a criminal, he got to go to federal pound me in the ass prison for his golden years.
http://www.ehbonline.org/article/S1090-5138%2814%2900077-4/a...
>Convicted criminal offenders had more children than individuals never convicted of a criminal offense. Criminal offenders also had more reproductive partners, were less often married, more likely to get remarried if ever married, and had more often contracted a sexually transmitted disease than non-offenders. Importantly, the increased reproductive success of criminals was explained by a fertility increase from having children with several different partners.
I think the bigger concern is how we select for intelligence. Doing it based on a few genes that correlate strongly with intelligence can lead to an increased vulnerability to health issues related to those genes we don't know about, but I think this applies to any gene level selection based on any characteristics.