The machine learning algorithms I have been studying (in order to develop a robust cerebral cortex model) are also useful directly for real-world problems. For example, the Netflix Prize is a machine learning competition with the goal of improving the Netflix company's movie recommendation system.
I haven't spent too much time on this yet, but I tried applying a topographic clustering algorithm and managed to score 0.74% better than Netflix's own recommendation algorithm. (10% better wins $1,000,000.) I am ranked #1,364 out of 37,086 teams as of Jan 29, 2009.