Why not start with making something you want? If you don't know what other people want, you know exactly what pains you and just fix that. Why not even stop thinking about starting a startup when you start solving your own problem? If your product solves your own problem, find your niches to see if it bugs them also.

Someone put it nicely - facebook is Mark's side project 6, probably? Of course not all startups start that way, if you don't know what others want, that is a good way to start and at worst it is a good way to sharpen your programming/non-programming skills.