Billion dollar baby
I would love to see algorithms researchers attack problems (not only longstanding ones but also ones) that are urgent and timely. Beyond the million dollar problems, we may now care about how the US govt (and may be other govts around the world in slow motion) will (may be) buy troubled assets from financial companies for $$$ Billions. The question is to design a suitable auction. Lawrence Ausubel and Peter Cramton are two Economists who have a working paper addressing this problem: check out the pages 1 and 2, and this paper. There are issues here a Computer Scientist can address.