Here are some data sets:

The UCI machine learning repository

http://archive.ics.uci.edu/ml/datasets.html?format=&task=&att=&area=&numAtt=&numIns=&type=&sort=instUp&view=table

Notes:

Lots of sets such as medical surveys, stock exchange, car sales,  ham/spam and US census data.

Google data:

http://www.google.com/publicdata/directory

Notes:

It also has tons of data. The only problem with it is that most of the interpretation is already done. However, someone could still combine several datasets to answer new questions.

Here are some cool examples of visualization techniques:

http://christopheviau.com/d3list/