I currently do Folding@Home (based on how proteins fold; research which has implications for numerous Human illnesses/diseases). It's quite good, although it doesn't currently support all GPUs.
It's no more "useless busywork" than security guards protecting banks from theft. In this case, the proof-of-work protects the network from double-spending.
Do you have a concrete suggestion for how the proof-of-work could be dual-purposed and still serve its function in the bitcoin network?
Wikipedia has a list of other worthwhile projects to consider: http://en.wikipedia.org/wiki/List_of_distributed_computing_p...