The reason that protein folding and other large-scale projects (e.g. SETI@Home) work is because there's enough data to go around. I think that these projects are likely bound by the fairly small bits of data they have available.
If we could get an open dataset of cetacean sounds along with a tagged assessment of the scenario (i.e. how many were around, were they agitated, etc) then this kind of project could be very interesting.
A better model is Kaggle. Kaggle releases data, and then challenges people to produce statistical models that can fit the data the best. Hundreds of people compete and best the results are usually very good.
However as far as I know there isn't that much data available for dolphins (and especially not labelled data, though the unlabelled data may be useful as well.)
If we could get an open dataset of cetacean sounds along with a tagged assessment of the scenario (i.e. how many were around, were they agitated, etc) then this kind of project could be very interesting.