I would beg to differ. It is not efficient from standpoint of labor allocation. Not experienced people are rarely efficient at completing complex tasks. On the other hand once problem captures imagination of the significant portion of population amount of progress made goes way up.
It depends on the problem. There are problems that are higly parallelizable and problems
that are not. There are problems where expertise is paramount and problems where it is not. What you said is true but what I said is also true, if you stretch enough the
definition of "fairly unreliable" :)