I think both projects are pretty useless. I am running GIMPS out of curiosity and for a chance to win $50K from EFF (I'd rather it had financed my Newsfilter project). Explanations why GIMPS is useful on their own page are quite pathetic.
It's strange that nobody has found more productive tasks that could be distributed among multiple computers.
SETI does sound like it has at least some chance to find something of interest. I'd be willing to bet anything against aliens though. And I wouldn't want it to slow down my computer.
Anybody has suggestions for distributed computations that would actually be useful?
Do you know any practical tasks that require high computational power and can be easily parallelized?