How can you solve a problem that you can't adequately analyze or characterize? How will you know if you aren't trying a failed solution if you don't know what has been tried? How will you know if your simplifications and assumptions are realistic if you haven't seen how they've played out in the past?
Often they don't have the tech skills or desire/will to start something themselves, but are happy to join you if they recognize that it's a big enough problem.
If your expertise is organizational and can put together the right team with those that do have the domain knowledge, I'll concede the point. But domain knowledge is irreplaceable if you actually want to solve hard problems with scalable markets.
