It's only good for a jumping off place.
So totally true. But once you allow that 99% of modern life is jumping off, I'm not sure what you're griping about.
Just as one comparison, take every organization prominent enough to have it's own article in en.Wikipedia, go to their own websites (the vast majority will have one) and scrape all of the "about us" web pages these organizations authored about themselves, and imagine these as a collective "About Us"-apedia.
This "About Us"-apedia would make MySpace's worst year look like an exercise in design consistency. I for one can live without the metric fuckton of Flash-based incoherence as my standard point of departure on the agencies of the world.
It seems to me that all the people who hated Wikipedia on first sight share an underlying belief in knowledge as an authority network. The reason Wikipedia succeeded is that knowledge isn't what we thought it was. For the vast majority of purposes, authority is a boundary condition, not the thing itself.
The first step in assimilating a new body of knowledge is to survey the field's lexicon: What words are used and roughly how are they linked together? This cognitive process takes place long before factual assertions amount to a hill of beans. When the facts do begin to matter, most smart people are well aware that in this world we're all fed baloney 24 hours a day. Wikipedia is one of the places where it becomes especially clear how the baloney is made. That doesn't make it worse baloney than Superbowl Sunday—America's national slick-baloney celebration day. Is iOS somehow less Orwellian than the IBM PC? So we were told through a non-linguistic medium.
On Wikipedia, when I spot baloney, I click the magic button called "History" where I scan for edit wars and substantial discards. For the vast majority of articles, it's all there in plain view. The mythical, Orwellian-smashing parentage of iOS is harder to trace.
In the upcoming era of Deep Watson, those Wikipedia crumb trails of sturm und churn will suddenly become interesting resources to expose to automated data mining. Perhaps then the present surface form of the articles will begin to fade in importance. There's nothing stopping this, except for the will to go there, which is depressingly thin in the general public for the 99% of the time they're merely jumping off.