This is true in some instances, but not generally. I think most advances are made gradually by very experienced people who understand exactly why things are "unsolvable" and what paths may be more fruitful.
Secondly, it is doubtful that a student will solve a millennium problem or make any other advance of that caliber.