OK. The Wisdom of Crowds meme is making it's way around the blogosphere. Just saw it for the 5th or 6th time so it's time to comment.
Extracting information by making private expectations and beliefs public is indeed extremely powerful, and something that the world has underutilized. It's intuitively obvious why it works: beliefs and expectation errors, are, in the aggregate, self-correcting.
Or are they? The truth, actually, is much murkier than this. The most obvious example is are financial markets - hardly bastions of what we might call 'accurate revelation'. In fact, our best theories to explain what happens in financial markets tell us that crowds aren't smart in the aggregate, but really, really dumb.
They're susceptible to mind viruses, via inductive expectations. They're vulnerable to cascades and herding behaviour, which can be triggered by seemingly innocuous phenomena, and which can cause aggregate irrationality even if individuals are rational.
Psychology is subject to a variety of cognitive biases. Kahneman and Tversky were a few of the guys that kicked this research off, years and years ago - now, it's the economic mainstream, and that's where concepts like hyperbolic discounting - which is completely irrational - come from.
So I think the answer is that crowds are good at picking numbers of jellybeans in a jar - because this is such an artificial exercise that it's isolated from all the problems above. But real-world crowds are much less intelligent and more fickle than they are 'wise' - at least when it comes to real-world situations.