How do you use Mechanical Turk?

I'm a big fan of mechanical turk and often leverage it with smartsheet. I've used mturk for basic writing, gathering content from multiple sources, and some minimal research. However, I don't think I've truly tapped into its potential.

How do you use Mechanical Turk to crowdsource effectively? What are some of the biggest problems or challenges you have solved with Mechanical Turk? Or are there failures you'd like to share?

I don't have the rep to add this as a comment. What is a mechanical turk?

Andy Baio has done a bunch of interesting stuff w/ MTurk:
http://www.google.com/search?q=site%3Awaxy.org+mechanical+turk See also:
http://stackoverflow.com/questions/193426/have-you-made-interesting-use-of-mechanical-turk

You should look into Crowdflower. They automatically determine the accuracy of workers and correct for bad workers, something that has always been difficult with Mechanical Turks.

You should also see the empirical evaluation of Mechanical Turk called "Cheap and Fast - But is it Good? Evaluating Non-Expert Annotations for Natural Language Tasks " by my colleague Rion Snow. There is a paper and easier to read blog post.