Why try to guess what many people would want? Let them tell you more directly via a service for collaborative startup specification. Unlike implementation, specification -- at least at an informal level -- can be done by pretty well anyone. Of course, you will need to provide some way to end up with a coherent spec from all those user contributions. But if something like wikipedia can produce stellar results, then why not this?
And when you start working on something, why not keep all those people up to date with your progress? They would probably be very eager to see what you are doing with their spec contributions. And they may evolve the spec over time, again, in a collaborative way.
When your startup is ready to launch, you will then have all those people ready to try it out and tell their friends.
Wikipedia isn't the right mechanism. It's fine when you're collaborating on documenting something that's completely objective, but as soon as you get into something controversial it descends into chaos. There's good reason that some of the highest-quality articles on Wikipedia tend to be the ones about math.
Darcs/git might be a better model for this sort of thing: everybody gets their own working copy and can merge changes made to other copies into their own.
I suppose the sincerest form of voting would be merging others' ideas into your working copy. The system could track that and maintain a leaderboard of most-assimilated specs.
Amichail, there is a site that was a proto-version of what you're describing called The Business Experiment. It seems to be down right now, but here's an article about it:
"People know what they want, but people also want what they don't know." - Gilberto Gil
People know what they want- when they see it.
They may not be very good at telling you where their itch/pain is out of the blue. They may be taking it for granted and not even feeling it anymore. But if you just start scratching/massaging _somewhere_, you'll get useful "more to the left, no, a bit more up- THERE- AAaaaahh!" feedback. If you're so out of touch that they will push you away rather than steer you into the right spots, well, that's useful info too.
I don't imagine that feedback will be coming easy unless there's the prospect of an almost immediate solution (I'm less likely to discuss back itches on the phone). Your idea, as I understand it, offers a slim chance that each specific problem will be addressed, and even in that case the solution is months away.
Filtering people out isn't necessarily bad. This may help in identifying problems that hurt a lot. What you probably want, though, is problems that hurt lots of people.
Having something like what you propose would be better than not having it, and I don't want to discourage you from the idea. Just don't expect it to replace "trying to guess what many people would want."
Seems like you have to have a starting point in order to get direction for your project from users/customers. The initial idea doesn't have to be stellar, just something fairly specific. I hear posing the idea as a question helps.
Pasting a big bulletin on the web that says "What do you want me to make?" would likely be useless. (Or would it? There's something to try...) The domain is too big; trying to gathering the web into a room and giving them a blank whiteboard might generate a lot of ideas, but how many of them would you be interested in? In fact, isn't that kind of what the web is to begin with: a big whiteboard with a bunch of ideas written on it? So which ones get your attention?
Solve some problem that you're interested in, place it in public view, foster communication with the users/customers, and listen. Building something specific ("hey, here's a web-based classifieds system..."), getting people to use it, and giving them a place to critique it seems to be a good way to get people to tell you what they need built.
Building a prototype and then getting people to use it is like writing that initial phrase on a white board, circling it, and asking every one on the web to generate ideas. Interested people will cluster around the idea and talk about it. Some of the talk will be horrible, and some outstanding -- but it will likely be more focused than if you just ask the web "what should I make?"
I've had similar thoughts; a sort of business that is "open source" in the sense that the users determine what gets built and can look at (modify, even?) what's going on...
And when you start working on something, why not keep all those people up to date with your progress? They would probably be very eager to see what you are doing with their spec contributions. And they may evolve the spec over time, again, in a collaborative way.
When your startup is ready to launch, you will then have all those people ready to try it out and tell their friends.