A psychopath could fail to control his violent impulses if subjected to too many violent images. But I think it is a reasonable suggestion that that person may have been bound to crack at some point. Anyway, such people wouldn't just go from killing virtual people to killing real people. They tend to be the type that tortures animals and is sadistic at school. An odd liking for death and destruction in computer games doesn't indicate a potential psychopath. A disturbing liking for real pain and injury however is a classic red flag that someone is a psycho. That said, I suppose murdering digital people in GTA could somehow get tagged with killing real people if the person is fucked up enough... I don't know, I freely confess that I don't know enough on this subject. If I find something I'll post it.
Added later: I've had a fair look around. I can't find any hard scientific examples either way... damn.
You'd think there'd be some research on this! I can't find anything that isn't either irrelevant to the question or that you have to pay for, darn it!
There is this article. I wasn't very impressed by them including Crash Twinsanity as a violent game, but at least this reaches some conclusions... still nothing on psychopaths though... this isn't very helpful but for now its the best I've got.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2678173&tool=pmcentrez
Added later: I did find this article. It alludes to a possible link between video games and violence. Its just an opinion like mine though, albeit a more informed one. Here's the link.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1281391&tool=pmcentrez
Hey! on a more postitive note... this article hints at Tetris possibly being a way of combating violent flashbacks. Its important to measure the positive and not just look at the negative.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2607539&tool=pmcentrez
Well, thats all I have. My opinion, some sites, and a few positives... but no hard data. I think the juries out on this one. Sorry for the rambling, disjointed post, I put things down as I found them... not that the research found much...