Another lesson to take from Steve's experience: the bigger the impact your startup is likely to make, the greater the chance it will fail. That might seem quite counterintuitive at first. Certainly, if there's a market and the demand is there, then success should be forthcoming, right?
Except if the problem is unsolved, it's probably unsolved for a reason. I'm reminded of something I read (I believe it was from "Germs, Guns, and Steel", but I don't have it to reference): invention is the mother of necessity! Safe startups give people something they already want. Great startups give people something they didn't even know they needed.
But humans are dumb animals. The first three times you show them something better, they'll turn their backs. So the more important your work is, the more likely you are to fail, and fail, and fail again. It's not you, it's just human nature.