The prerequisite for mechanized sentience is AI. Hopefully those who develop it will incorporate Asimov's three rules of robotics. If not, we will get skynet.
Is true sentience required to get a Skynet situation? Couldn't a highly sophisticated, machine-learning system crunch the numbers and go through its algorithms and decide that wiping out humanity is the most logical outcome without requiring self-awareness? Was the WOPR in
War Games sentient? It learned, but I would be hesitant to say it was sentient.
Actually, the common denominator in all those stories seems to be the ability of said computer system to control our nuclear arsenal and then wipe us out. I was just reading an article a couple weeks ago about the technology in one of our Minutemen III launch silos. The technology is so old down there that they are still transferring data on 8-inch floppy disks! I don't think a sentient software program will be able to sneak into that computer system within the storage capacity of an 8-inch floppy.
Edit: Actually, come to think of it, the WOPR almost blew up the world and it didn't even realize that it wasn't playing a game!